1
00:00:00,000 --> 00:00:04,422


2
00:00:04,422 --> 00:00:05,630
We've got a nice crowd today.

3
00:00:05,630 --> 00:00:08,390
I'd like to welcome you all
to the Distinguished Lecture

4
00:00:08,390 --> 00:00:09,230
series.

5
00:00:09,230 --> 00:00:12,170
We have the pleasure of
having Bob Morris here

6
00:00:12,170 --> 00:00:15,080
to talk to us about
the protection

7
00:00:15,080 --> 00:00:16,790
of valuable information.

8
00:00:16,790 --> 00:00:18,440
And before we get
into that, I'd like

9
00:00:18,440 --> 00:00:22,040
to continue with the
tradition of giving a somewhat

10
00:00:22,040 --> 00:00:27,423
extended bio to
our attended group

11
00:00:27,423 --> 00:00:29,090
here, particularly
the new students, who

12
00:00:29,090 --> 00:00:31,730
like to take our distinguished
speakers as role models.

13
00:00:31,730 --> 00:00:33,290
And so we'll do
a little of that.

14
00:00:33,290 --> 00:00:35,600
Mike [INAUDIBLE] is
unfortunately ill with the flu.

15
00:00:35,600 --> 00:00:40,010
Otherwise, we would have had
him play the traditional role.

16
00:00:40,010 --> 00:00:42,050
So Bob Morris is
a legendary figure

17
00:00:42,050 --> 00:00:44,610
in many areas of computer
security in particular.

18
00:00:44,610 --> 00:00:48,140
He was born in Boston, grew
up in northern Connecticut

19
00:00:48,140 --> 00:00:51,308
and went to school at
Exeter, and then went

20
00:00:51,308 --> 00:00:52,850
to this little school
down the street

21
00:00:52,850 --> 00:00:56,030
that sometimes has tried to an
excess but has always failed.

22
00:00:56,030 --> 00:00:59,420


23
00:00:59,420 --> 00:00:59,960
Majored in--

24
00:00:59,960 --> 00:01:01,940
[INAUDIBLE]?

25
00:01:01,940 --> 00:01:02,705
That direction.

26
00:01:02,705 --> 00:01:05,810


27
00:01:05,810 --> 00:01:10,760
Majored in chemistry,
but actually,

28
00:01:10,760 --> 00:01:13,790
during the middle of his
career at Harvard, he

29
00:01:13,790 --> 00:01:17,000
was drafted and spent a
couple of years in Germany

30
00:01:17,000 --> 00:01:19,190
as a signal
[INAUDIBLE] Heidelberg,

31
00:01:19,190 --> 00:01:21,890
working with the radio there.

32
00:01:21,890 --> 00:01:24,832
And was majoring in chemistry,
and then changed his mind

33
00:01:24,832 --> 00:01:27,040
just as he got near the end
of his career at Harvard,

34
00:01:27,040 --> 00:01:30,060
and decided he didn't want
to be a salesman for DuPont.

35
00:01:30,060 --> 00:01:34,088
And so he switched to a math
major his senior year, math

36
00:01:34,088 --> 00:01:35,630
being one of the
fields he recognized

37
00:01:35,630 --> 00:01:37,755
as being something you
could do almost anything you

38
00:01:37,755 --> 00:01:38,760
wanted afterwards.

39
00:01:38,760 --> 00:01:41,030
And as you'll see, he's
done many, many things

40
00:01:41,030 --> 00:01:43,400
in many different
directions ever since.

41
00:01:43,400 --> 00:01:47,217
He stayed on at Harvard, did
a master's in mathematics,

42
00:01:47,217 --> 00:01:49,050
studying areas of algebra
and number theory,

43
00:01:49,050 --> 00:01:51,350
particularly, looking
at class field theory

44
00:01:51,350 --> 00:01:54,710
and putting bounds on the size
of class groups for his thesis

45
00:01:54,710 --> 00:01:56,010
there.

46
00:01:56,010 --> 00:01:57,830
After that, he went
directly to Bell Labs

47
00:01:57,830 --> 00:02:00,920
as a mathematician
in 1960, then.

48
00:02:00,920 --> 00:02:03,260
And he was working on a very
interesting project, which

49
00:02:03,260 --> 00:02:05,100
didn't have the
term at the time,

50
00:02:05,100 --> 00:02:07,460
but now has become to
be known as modems.

51
00:02:07,460 --> 00:02:10,280
He was working on
the first modems.

52
00:02:10,280 --> 00:02:12,620
They didn't exist at the
time, and people at Bell Labs

53
00:02:12,620 --> 00:02:15,050
thought maybe that would be
a good idea, not realizing

54
00:02:15,050 --> 00:02:17,258
of course that the whole
internet and everything else

55
00:02:17,258 --> 00:02:19,850
is built around these
things at the time now.

56
00:02:19,850 --> 00:02:23,360
So they were aiming
for a high-speed modem

57
00:02:23,360 --> 00:02:25,028
of 48 bits per second.

58
00:02:25,028 --> 00:02:27,020
[LAUGHTER]

59
00:02:27,020 --> 00:02:30,380
With an error rate of 10 to
the minus 5th bit error that.

60
00:02:30,380 --> 00:02:33,590
So that was a couple of years
they were working on that,

61
00:02:33,590 --> 00:02:36,128
and he pointed out
that at the time,

62
00:02:36,128 --> 00:02:37,670
it seemed maybe they
would be getting

63
00:02:37,670 --> 00:02:38,795
near the theoretical limit.

64
00:02:38,795 --> 00:02:41,150
And the theoretical limits
have been reached and passed

65
00:02:41,150 --> 00:02:44,270
several times since
then at 1,200 and 9,600.

66
00:02:44,270 --> 00:02:47,053
Now we're at 56 kilobits,
but every time people

67
00:02:47,053 --> 00:02:48,720
say that we're at the
theoretical limit.

68
00:02:48,720 --> 00:02:50,137
But at the time,
they were working

69
00:02:50,137 --> 00:02:52,520
near the theoretical limit then.

70
00:02:52,520 --> 00:02:54,980
I asked him if he knew
Claude Shannon at the time.

71
00:02:54,980 --> 00:02:57,330
Claude had left Bell
Labs at that time,

72
00:02:57,330 --> 00:03:00,350
but Bob had known him in other
contexts, and used to love,

73
00:03:00,350 --> 00:03:02,390
he said, trampolining with them.

74
00:03:02,390 --> 00:03:07,460
[LAUGHTER]

75
00:03:07,460 --> 00:03:11,000
So then he moved into the
math department at Bell Labs.

76
00:03:11,000 --> 00:03:11,960
Ron Graham was there.

77
00:03:11,960 --> 00:03:14,210
The math department was
shifting into computer science

78
00:03:14,210 --> 00:03:17,570
at the time, and Bob got
interested in computer science

79
00:03:17,570 --> 00:03:19,170
a number of different ways.

80
00:03:19,170 --> 00:03:20,540
One of the interesting
things, and the connection

81
00:03:20,540 --> 00:03:22,665
with MIT here, was he got
involved with the Multics

82
00:03:22,665 --> 00:03:23,660
project.

83
00:03:23,660 --> 00:03:27,350
So he wrote the first PLO,
PL/I compiler for the Mutlics

84
00:03:27,350 --> 00:03:29,845
project, which was
around for, I think,

85
00:03:29,845 --> 00:03:31,970
he said an embarrassing,
long, long time afterwards

86
00:03:31,970 --> 00:03:34,190
or something.

87
00:03:34,190 --> 00:03:37,018
So he was involved with
that, and up here, he's

88
00:03:37,018 --> 00:03:38,310
used Multics quite a bit since.

89
00:03:38,310 --> 00:03:40,435
I'm surprised to see that
some of the design errors

90
00:03:40,435 --> 00:03:41,332
are still around.

91
00:03:41,332 --> 00:03:43,790
I guess there was a question
as to whether Multics is still

92
00:03:43,790 --> 00:03:44,938
running anywhere.

93
00:03:44,938 --> 00:03:45,980
We were talking with him.

94
00:03:45,980 --> 00:03:48,522
I don't think either of us knew
exactly, maybe up until quite

95
00:03:48,522 --> 00:03:50,570
recently, anyway.

96
00:03:50,570 --> 00:03:53,150
He did a lot of number theory,
and he kept his mathematics

97
00:03:53,150 --> 00:03:53,650
going too.

98
00:03:53,650 --> 00:03:56,630
He did a lot of number theory
at the same time he was there.

99
00:03:56,630 --> 00:03:59,690
He worked on implementations of
various mathematical functions

100
00:03:59,690 --> 00:04:02,222
for the 7090, and for
Unix, he did all the math

101
00:04:02,222 --> 00:04:04,430
implementations, sines,
cosines and things like that,

102
00:04:04,430 --> 00:04:06,740
for the Unix implementation.

103
00:04:06,740 --> 00:04:08,270
Also did the Unix
security scheme,

104
00:04:08,270 --> 00:04:11,720
the password scheme the way,
et cetera, a password works.

105
00:04:11,720 --> 00:04:13,040
Passwords get hashed into that.

106
00:04:13,040 --> 00:04:16,279


107
00:04:16,279 --> 00:04:18,959
In 1980-- so that was
software, some mathematics.

108
00:04:18,959 --> 00:04:20,660
Then he got into hardware.

109
00:04:20,660 --> 00:04:22,910
He started on a project
working with the Navy

110
00:04:22,910 --> 00:04:27,410
on very high-speed computer,
one of the fastest in time.

111
00:04:27,410 --> 00:04:31,230
They were aiming for $100
million operations per second.

112
00:04:31,230 --> 00:04:32,855
This was for sonar processing.

113
00:04:32,855 --> 00:04:33,980
He said it was interesting.

114
00:04:33,980 --> 00:04:36,140
He got into the project, and he
was talking to some Navy guys,

115
00:04:36,140 --> 00:04:38,040
and they realized he'd
never been on a boat.

116
00:04:38,040 --> 00:04:40,850
And they said some order
like, get his ass out to sea,

117
00:04:40,850 --> 00:04:43,640
and so they took him out
on a nuclear submarine

118
00:04:43,640 --> 00:04:46,601
for a week, which he said
was very interesting.

119
00:04:46,601 --> 00:04:48,960


120
00:04:48,960 --> 00:04:50,585
So that that machine
he was working on,

121
00:04:50,585 --> 00:04:53,183
it actually got built and was
in service for quite a while.

122
00:04:53,183 --> 00:04:54,350
The machine was interesting.

123
00:04:54,350 --> 00:04:56,420
It was primarily hardware design
he was involved with then.

124
00:04:56,420 --> 00:04:58,280
It had multiple processors,
and so the usual issues

125
00:04:58,280 --> 00:05:00,680
of synchronization and
coordination of the processors

126
00:05:00,680 --> 00:05:03,170
was a big issue.

127
00:05:03,170 --> 00:05:09,020
In 1984, he retired from Bell
Labs and went to work at NSA.

128
00:05:09,020 --> 00:05:11,900
I asked him if he'd done any
cryptography before he went

129
00:05:11,900 --> 00:05:13,910
over to NSA, and he
admitted that actually he

130
00:05:13,910 --> 00:05:17,630
had worked on a paper
on cryptanalysis

131
00:05:17,630 --> 00:05:19,670
that he thought was
kind of interesting.

132
00:05:19,670 --> 00:05:20,780
And he sent it off
to the NSA-- this

133
00:05:20,780 --> 00:05:21,650
is while he was at Bell Labs.

134
00:05:21,650 --> 00:05:23,270
He sent it off the
NSA saying, you

135
00:05:23,270 --> 00:05:25,670
don't mind if if
publish this, do you?

136
00:05:25,670 --> 00:05:27,990
And they came back
and said, yes.

137
00:05:27,990 --> 00:05:30,140
And [INAUDIBLE] several
high-level visitors

138
00:05:30,140 --> 00:05:31,640
from the agency
coming to Bell Labs,

139
00:05:31,640 --> 00:05:34,520
and it was agreed that
he wouldn't publish this.

140
00:05:34,520 --> 00:05:36,020
And it still has
not been published,

141
00:05:36,020 --> 00:05:38,270
and I look forward
to seeing it someday.

142
00:05:38,270 --> 00:05:40,280
But he says it was
perhaps the best

143
00:05:40,280 --> 00:05:41,679
paper he's ever written, so.

144
00:05:41,679 --> 00:05:45,431
[LAUGHTER]

145
00:05:45,431 --> 00:05:47,778


146
00:05:47,778 --> 00:05:49,820
But he also feels that
it's maybe not appropriate

147
00:05:49,820 --> 00:05:55,280
for it to be published
either, so, one of these days.

148
00:05:55,280 --> 00:05:57,147
At NSA, he started
off as chief scientist

149
00:05:57,147 --> 00:05:58,730
of the Computer
Security Center, which

150
00:05:58,730 --> 00:06:01,485
was a new center at the time,
working on computer security,

151
00:06:01,485 --> 00:06:03,110
both inside the
government and outside.

152
00:06:03,110 --> 00:06:06,770
Did consulting, a number
of different directions.

153
00:06:06,770 --> 00:06:09,410
He was promoted to chief
scientist of the Information

154
00:06:09,410 --> 00:06:11,363
Security Section
thereafter, working

155
00:06:11,363 --> 00:06:13,280
on protection of information
in general, which

156
00:06:13,280 --> 00:06:16,820
is the topic of today's talk.

157
00:06:16,820 --> 00:06:19,190
While he was there, he had
an interesting experience,

158
00:06:19,190 --> 00:06:20,690
as it's the destiny
of most children

159
00:06:20,690 --> 00:06:21,815
to embarrass their parents.

160
00:06:21,815 --> 00:06:24,230
In November of '88, the
internet worm came up.

161
00:06:24,230 --> 00:06:26,540
So again, interesting
experience with him

162
00:06:26,540 --> 00:06:31,520
to deal with that both in
his family and externally.

163
00:06:31,520 --> 00:06:36,020
In August of 1990, his career
took a very interesting turn

164
00:06:36,020 --> 00:06:38,510
with the invasion of Kuwait.

165
00:06:38,510 --> 00:06:40,820
He was pulled out
of his ordinary work

166
00:06:40,820 --> 00:06:45,860
and put on a special team whose
goal was to essentially nullify

167
00:06:45,860 --> 00:06:48,440
the Iraqi air defense system.

168
00:06:48,440 --> 00:06:50,480
And he spent eight
months on this project

169
00:06:50,480 --> 00:06:53,705
and effectively had the rank of
a two- or three-star general,

170
00:06:53,705 --> 00:06:55,580
and was getting everything
he needed in terms

171
00:06:55,580 --> 00:06:56,663
of cooperation, of course.

172
00:06:56,663 --> 00:06:58,880
And he didn't say a
lot about what he did

173
00:06:58,880 --> 00:07:01,280
or the details of that,
but I think we all

174
00:07:01,280 --> 00:07:02,970
know that what
happened in that war

175
00:07:02,970 --> 00:07:07,160
was astounding in a
number of regards.

176
00:07:07,160 --> 00:07:10,160
Then in 1994, he
retired from the NSA.

177
00:07:10,160 --> 00:07:14,750
He's now living up in Hanover
next to our sister institution,

178
00:07:14,750 --> 00:07:17,090
Dartmouth, where he has
an adjunct professorship.

179
00:07:17,090 --> 00:07:18,780
And he also has an
adjunct professor

180
00:07:18,780 --> 00:07:20,630
at Cambridge
University in England,

181
00:07:20,630 --> 00:07:24,837
so he goes back and visits
there substantially as well.

182
00:07:24,837 --> 00:07:25,670
So he's retired now.

183
00:07:25,670 --> 00:07:27,200
He says he likes to split wood.

184
00:07:27,200 --> 00:07:29,890
I know he's interested
in languages too.

185
00:07:29,890 --> 00:07:31,280
He speaks very many languages.

186
00:07:31,280 --> 00:07:34,530
I think if you came up and
challenge him with languages,

187
00:07:34,530 --> 00:07:36,245
ask him questions on
languages at the end

188
00:07:36,245 --> 00:07:37,880
or after the talk
or something, he's

189
00:07:37,880 --> 00:07:41,030
probably better able to
respond than anybody I know.

190
00:07:41,030 --> 00:07:42,770
He spends Tuesday
afternoons at lunch,

191
00:07:42,770 --> 00:07:45,265
he tells me, reading the
Bible in ancient Greek

192
00:07:45,265 --> 00:07:46,140
with a reading group.

193
00:07:46,140 --> 00:07:50,600
And so as usual, we
like to ask our speakers

194
00:07:50,600 --> 00:07:55,580
to sum up their life experiences
in an epigram to follow,

195
00:07:55,580 --> 00:07:58,940
and what kind of advice to
give very young students.

196
00:07:58,940 --> 00:08:01,670
And after thinking, he said,
if you're not having fun,

197
00:08:01,670 --> 00:08:03,420
you should be doing
something else.

198
00:08:03,420 --> 00:08:06,683
And that actually makes a lot
of other distinguished lectures

199
00:08:06,683 --> 00:08:07,350
[INAUDIBLE] too.

200
00:08:07,350 --> 00:08:08,150
You know, find your passion.

201
00:08:08,150 --> 00:08:08,670
Go after it.

202
00:08:08,670 --> 00:08:10,430
So without further ado,
I am pleased to introduce

203
00:08:10,430 --> 00:08:10,940
Bob Morris.

204
00:08:10,940 --> 00:08:15,780


205
00:08:15,780 --> 00:08:16,590
That's a yes.

206
00:08:16,590 --> 00:08:23,100


207
00:08:23,100 --> 00:08:23,600
Oh.

208
00:08:23,600 --> 00:08:26,620


209
00:08:26,620 --> 00:08:27,700
Yeah, high tech.

210
00:08:27,700 --> 00:08:33,210


211
00:08:33,210 --> 00:08:39,090
In order to deliver
what message I have,

212
00:08:39,090 --> 00:08:41,940
I actually have to
change the way you think

213
00:08:41,940 --> 00:08:43,260
about some kinds of things.

214
00:08:43,260 --> 00:08:46,000


215
00:08:46,000 --> 00:08:51,070
And when we're finished--

216
00:08:51,070 --> 00:08:52,820
well, when I'm finished
talking and you're

217
00:08:52,820 --> 00:08:58,220
finished listening to
me, we have to check back

218
00:08:58,220 --> 00:09:00,860
with these initial notions.

219
00:09:00,860 --> 00:09:05,150
I worked for an
organization, NSA,

220
00:09:05,150 --> 00:09:07,700
which is characterized
as being, let's

221
00:09:07,700 --> 00:09:15,940
say, somewhat paranoid,
to which I can respond,

222
00:09:15,940 --> 00:09:17,948
just because you're
paranoid doesn't mean

223
00:09:17,948 --> 00:09:19,240
the whole world is against you.

224
00:09:19,240 --> 00:09:22,930


225
00:09:22,930 --> 00:09:26,290
And even at NSA,
I was considered,

226
00:09:26,290 --> 00:09:29,490
in that organization,
unusually paranoid.

227
00:09:29,490 --> 00:09:33,190
[LAUGHTER]

228
00:09:33,190 --> 00:09:35,410
And let me give you
some examples of this.

229
00:09:35,410 --> 00:09:38,200


230
00:09:38,200 --> 00:09:42,040
These are, by the
way, random view graph

231
00:09:42,040 --> 00:09:45,340
that I'm putting up here with--

232
00:09:45,340 --> 00:09:54,190
perhaps relevant but not
necessarily very well ordered.

233
00:09:54,190 --> 00:09:57,400
I've given a talk
on this subject

234
00:09:57,400 --> 00:10:01,360
to many groups in the
past, some academic,

235
00:10:01,360 --> 00:10:04,690
as this is, some military,
some God knows what.

236
00:10:04,690 --> 00:10:08,150


237
00:10:08,150 --> 00:10:12,510
But the best abstract of
the talk that I've given

238
00:10:12,510 --> 00:10:17,800
is in, God help
us, WIRED magazine.

239
00:10:17,800 --> 00:10:18,910
Is there no dignity?

240
00:10:18,910 --> 00:10:20,404
[LAUGHTER]

241
00:10:20,404 --> 00:10:23,392
[APPLAUSE]

242
00:10:23,392 --> 00:10:26,380


243
00:10:26,380 --> 00:10:30,310
But it's the March 1996
issue of WIRED magazine,

244
00:10:30,310 --> 00:10:34,368
and actually, if you care more
about what I'm talking about,

245
00:10:34,368 --> 00:10:36,160
it's a good place to
look for it because it

246
00:10:36,160 --> 00:10:40,300
is a very good and
rather accurate

247
00:10:40,300 --> 00:10:43,450
summary of what I talked
about at a talk I gave just

248
00:10:43,450 --> 00:10:45,670
over two years ago.

249
00:10:45,670 --> 00:10:48,490
And it also goes
into some details

250
00:10:48,490 --> 00:10:50,740
of a conversation
I had with Phil

251
00:10:50,740 --> 00:10:52,840
Zimmerman at that
conference, and you

252
00:10:52,840 --> 00:10:55,510
can imagine, if you know
the people involved,

253
00:10:55,510 --> 00:10:58,431
that that must have been a hell
of an interesting conversation.

254
00:10:58,431 --> 00:11:03,350


255
00:11:03,350 --> 00:11:04,565
Now, paranoia.

256
00:11:04,565 --> 00:11:08,150


257
00:11:08,150 --> 00:11:12,300
This appears to be a telephone.

258
00:11:12,300 --> 00:11:18,390
This is commonly called a line
cord and this the handset cord.

259
00:11:18,390 --> 00:11:21,030
I have to tell you in
the strongest terms

260
00:11:21,030 --> 00:11:24,090
that I have different
names for this.

261
00:11:24,090 --> 00:11:25,320
This is called an antenna.

262
00:11:25,320 --> 00:11:28,560
[LAUGHTER]

263
00:11:28,560 --> 00:11:30,180
This is also called an antenna.

264
00:11:30,180 --> 00:11:32,900


265
00:11:32,900 --> 00:11:36,020
Many people refer to this
object as a microphone and this

266
00:11:36,020 --> 00:11:38,420
as an earphone.

267
00:11:38,420 --> 00:11:41,840
I think of this as a rather
high-quality microphone

268
00:11:41,840 --> 00:11:45,020
and this as a rather
low-quality microphone.

269
00:11:45,020 --> 00:11:48,350


270
00:11:48,350 --> 00:11:51,320
People think that I have a
tendency of seeing microphones

271
00:11:51,320 --> 00:11:52,940
and antennas everywhere.

272
00:11:52,940 --> 00:11:54,830
[LAUGHTER]

273
00:11:54,830 --> 00:11:59,480
I do, and that habit
is a necessary habit

274
00:11:59,480 --> 00:12:02,720
to be at all effective
in the kind of job

275
00:12:02,720 --> 00:12:05,150
that I was doing in NSA.

276
00:12:05,150 --> 00:12:07,500
I was involved in--

277
00:12:07,500 --> 00:12:10,340
I was the chief scientist
of the section of NSA

278
00:12:10,340 --> 00:12:14,480
that deals with the protection
of valuable or sensitive

279
00:12:14,480 --> 00:12:17,030
US information.

280
00:12:17,030 --> 00:12:19,640
And we did things
like, there was

281
00:12:19,640 --> 00:12:22,850
a crew of young people who
went over to the Pentagon,

282
00:12:22,850 --> 00:12:26,540
to the office of the
Chairman of the Joint Chiefs,

283
00:12:26,540 --> 00:12:31,670
and looked at that
office in great detail

284
00:12:31,670 --> 00:12:36,020
to assure themselves and
others that there were not

285
00:12:36,020 --> 00:12:42,170
any undue number of bugs
planted in his office.

286
00:12:42,170 --> 00:12:47,120
And they came to me many
times for advice and help

287
00:12:47,120 --> 00:12:50,060
and consultation, and
I recollect the day

288
00:12:50,060 --> 00:12:52,280
when they came back
and pointed out to me

289
00:12:52,280 --> 00:12:55,820
that in the office of the
Chairman of the Joint Chiefs,

290
00:12:55,820 --> 00:12:57,830
there was a bathroom.

291
00:12:57,830 --> 00:12:58,600
Oh my god.

292
00:12:58,600 --> 00:13:02,640


293
00:13:02,640 --> 00:13:05,040
One might have
noticed in a bathroom

294
00:13:05,040 --> 00:13:10,290
that if somebody opens
the door or closes

295
00:13:10,290 --> 00:13:13,650
the door of a bathroom, you
look at a sink with water in it,

296
00:13:13,650 --> 00:13:15,150
or the toilet bowl,
and you notice

297
00:13:15,150 --> 00:13:16,560
it actually changes level.

298
00:13:16,560 --> 00:13:20,790
I mean, there actually is some
connection between air pressure

299
00:13:20,790 --> 00:13:23,460
waves and water pressure
waves, and the fact

300
00:13:23,460 --> 00:13:25,050
that he had his
own bathroom caused

301
00:13:25,050 --> 00:13:29,340
us more cost, time and expense
than you would ordinarily

302
00:13:29,340 --> 00:13:29,970
believe.

303
00:13:29,970 --> 00:13:31,570
But that kind of
thing does happen.

304
00:13:31,570 --> 00:13:32,100
It was.

305
00:13:32,100 --> 00:13:34,680


306
00:13:34,680 --> 00:13:36,030
More paranoia.

307
00:13:36,030 --> 00:13:38,730
Now, this is a building without
an outside window, which

308
00:13:38,730 --> 00:13:41,730
means that I don't have
a prop for the next thing

309
00:13:41,730 --> 00:13:46,500
I was going to talk about,
but I might, with any luck--

310
00:13:46,500 --> 00:13:47,670
Yeah, I do have some chalk.

311
00:13:47,670 --> 00:13:49,128
I probably even
know how to use it.

312
00:13:49,128 --> 00:13:52,050


313
00:13:52,050 --> 00:13:53,180
Think of this as a window.

314
00:13:53,180 --> 00:13:56,270


315
00:13:56,270 --> 00:14:02,360
If you aim a laser
beam from a distance--

316
00:14:02,360 --> 00:14:07,520
and that distance can be quite
long, like easily half a mile,

317
00:14:07,520 --> 00:14:09,680
probably more--

318
00:14:09,680 --> 00:14:13,370
then that laser beam
will illuminate objects

319
00:14:13,370 --> 00:14:16,010
in the window.

320
00:14:16,010 --> 00:14:17,030
This is very convenient.

321
00:14:17,030 --> 00:14:19,080
This is the best example.

322
00:14:19,080 --> 00:14:20,780
This, by the way,
is going to turn out

323
00:14:20,780 --> 00:14:24,221
to be, no great
surprise, a microphone.

324
00:14:24,221 --> 00:14:27,110


325
00:14:27,110 --> 00:14:29,790
You put it on the windowsill.

326
00:14:29,790 --> 00:14:33,180
Now, a laser beam
coming in the window--

327
00:14:33,180 --> 00:14:34,950
hits that, reflected back--

328
00:14:34,950 --> 00:14:38,310
will be subject to
ordinary Doppler shift

329
00:14:38,310 --> 00:14:40,050
changes of frequency.

330
00:14:40,050 --> 00:14:43,980
You can detect those very
easily with very inexpensive

331
00:14:43,980 --> 00:14:47,070
equipment, and
then with the laser

332
00:14:47,070 --> 00:14:49,770
beam aimed through
the window at this,

333
00:14:49,770 --> 00:14:53,320
you can hear everything
that is spoken in a room.

334
00:14:53,320 --> 00:14:56,560
If this had-- well, I
wasn't set up to do that,

335
00:14:56,560 --> 00:14:58,590
but if this room had
an outside window,

336
00:14:58,590 --> 00:15:00,930
it would be a very
easy demonstration

337
00:15:00,930 --> 00:15:03,760
to show just how accurately
this can be done,

338
00:15:03,760 --> 00:15:05,790
and it can be done with
rather high quality.

339
00:15:05,790 --> 00:15:09,060
That's a high-quality
microphone.

340
00:15:09,060 --> 00:15:10,260
Plastic is much better.

341
00:15:10,260 --> 00:15:13,050
The cardboard cups
do a poor job.

342
00:15:13,050 --> 00:15:15,900
The plastic cups do
a rather good job.

343
00:15:15,900 --> 00:15:19,050


344
00:15:19,050 --> 00:15:21,900
Now, I'm talking about the
protection of information,

345
00:15:21,900 --> 00:15:24,510
and a couple of things
that I need to--

346
00:15:24,510 --> 00:15:28,500


347
00:15:28,500 --> 00:15:30,080
I had some notes here once.

348
00:15:30,080 --> 00:15:31,080
Well, it doesn't matter.

349
00:15:31,080 --> 00:15:36,470


350
00:15:36,470 --> 00:15:42,830
I ought to say a bit about what
I view as valuable information.

351
00:15:42,830 --> 00:15:45,740
In the past, the
valuable information

352
00:15:45,740 --> 00:15:47,690
that people most
wanted to protect

353
00:15:47,690 --> 00:15:50,180
or most wanted to
exploit were sort of

354
00:15:50,180 --> 00:15:56,210
both the same, were primarily
military and diplomatic

355
00:15:56,210 --> 00:15:58,340
messages.

356
00:15:58,340 --> 00:16:01,220
In peacetime, you wanted to know
what some foreign government

357
00:16:01,220 --> 00:16:02,510
was planning to do.

358
00:16:02,510 --> 00:16:06,800
Obvious the thing to do was to
tap the ambassador's telephone.

359
00:16:06,800 --> 00:16:09,500
In wartime, you wanted to know
what the military was planning

360
00:16:09,500 --> 00:16:13,340
to do, and so you
would intercept,

361
00:16:13,340 --> 00:16:18,370
in World War I and II,
their radio communications.

362
00:16:18,370 --> 00:16:20,510
Well, this is not new, then.

363
00:16:20,510 --> 00:16:23,150
I mean, you couldn't in the
United States during the Civil

364
00:16:23,150 --> 00:16:28,520
War find any telegraph line
anywhere in the country

365
00:16:28,520 --> 00:16:32,270
where there was fighting going
on that didn't have taps on it.

366
00:16:32,270 --> 00:16:35,030
I mean, that was a
really, really big deal

367
00:16:35,030 --> 00:16:37,130
during the Civil War.

368
00:16:37,130 --> 00:16:39,260
All telephone and
telegraph lines--

369
00:16:39,260 --> 00:16:41,570
in those days, it was
telegraph, not telephone.

370
00:16:41,570 --> 00:16:44,300
They were all tapped,
mostly encrypted

371
00:16:44,300 --> 00:16:46,460
or protected some
way or another,

372
00:16:46,460 --> 00:16:47,980
but people got the messages.

373
00:16:47,980 --> 00:16:51,410
So it's nothing
new, but military

374
00:16:51,410 --> 00:16:55,430
and diplomatic is sort of--

375
00:16:55,430 --> 00:16:57,180
it's becoming a
thing of the past.

376
00:16:57,180 --> 00:16:59,270
Not that it's less
important, but other things

377
00:16:59,270 --> 00:17:02,700
are gaining importance
to shift the balance.

378
00:17:02,700 --> 00:17:07,740
So currently, the most
important information

379
00:17:07,740 --> 00:17:13,410
that's floating around is
probably largely financial.

380
00:17:13,410 --> 00:17:16,770
Suppose that there is
a good deal of money

381
00:17:16,770 --> 00:17:18,839
in some Swiss bank
account, but it's

382
00:17:18,839 --> 00:17:23,369
the wrong account,
like not yours,

383
00:17:23,369 --> 00:17:25,394
and you want to fix that.

384
00:17:25,394 --> 00:17:28,050
[LAUGHTER]

385
00:17:28,050 --> 00:17:31,810
Well, OK.

386
00:17:31,810 --> 00:17:34,990
That means that the people who
are responsible for this system

387
00:17:34,990 --> 00:17:36,730
probably go to some lengths--

388
00:17:36,730 --> 00:17:38,080
in fact, they do--

389
00:17:38,080 --> 00:17:39,387
to protect the information.

390
00:17:39,387 --> 00:17:41,470
The people who want the
money go to some lengths--

391
00:17:41,470 --> 00:17:46,030
they do-- to extract
the information

392
00:17:46,030 --> 00:17:51,160
and engage in various
transfers of wealth

393
00:17:51,160 --> 00:17:53,720
from the wrong place
to the right place.

394
00:17:53,720 --> 00:17:58,660
And there have been a number
of very few of these incidents.

395
00:17:58,660 --> 00:18:02,810
About 5% are ever reported
even to the police

396
00:18:02,810 --> 00:18:06,070
so that it is difficult to
get a good measure of just how

397
00:18:06,070 --> 00:18:07,240
much of this is going on.

398
00:18:07,240 --> 00:18:09,100
There was a widely
publicized event

399
00:18:09,100 --> 00:18:12,160
at Citibank a few years back.

400
00:18:12,160 --> 00:18:15,650
Others, even larger ones,
have not become public,

401
00:18:15,650 --> 00:18:18,960
but they have in
any case happened.

402
00:18:18,960 --> 00:18:21,590
So that's the present.

403
00:18:21,590 --> 00:18:24,950
I'll talk about
ATMs in a minute.

404
00:18:24,950 --> 00:18:27,530
For the future, we're
getting, I think,

405
00:18:27,530 --> 00:18:33,110
into a number of areas
that are a bit different.

406
00:18:33,110 --> 00:18:35,900
Moving from financial
to generally

407
00:18:35,900 --> 00:18:39,800
commercial
transactions, the best

408
00:18:39,800 --> 00:18:43,310
figures I have, which are
good and relatively recent,

409
00:18:43,310 --> 00:18:47,570
from people who run their
businesses associated

410
00:18:47,570 --> 00:18:50,330
with the internet, so
that somebody orders

411
00:18:50,330 --> 00:18:55,040
something on the internet
and gets it shipped to them--

412
00:18:55,040 --> 00:18:57,290
the percentage of
transactions that turn out

413
00:18:57,290 --> 00:18:59,480
to be fraudulent
fraudulent in the sense

414
00:18:59,480 --> 00:19:02,750
that the person does not intend
to pay for what he's getting,

415
00:19:02,750 --> 00:19:05,170
is about 20%.

416
00:19:05,170 --> 00:19:06,250
Hey, that's low.

417
00:19:06,250 --> 00:19:09,280
Just wait.

418
00:19:09,280 --> 00:19:11,170
And so in the future,
I think we are

419
00:19:11,170 --> 00:19:13,630
going to see more
emphasis on privacy,

420
00:19:13,630 --> 00:19:17,680
more emphasis on
commercial transactions,

421
00:19:17,680 --> 00:19:20,830
and these will
shift the balance.

422
00:19:20,830 --> 00:19:25,460
If the World Wide Web turns
out to be a big success,

423
00:19:25,460 --> 00:19:28,870
then this i going to
vastly overweigh all

424
00:19:28,870 --> 00:19:30,640
of the other considerations.

425
00:19:30,640 --> 00:19:33,270
That hasn't happened yet,
but likely, it will happen.

426
00:19:33,270 --> 00:19:41,310


427
00:19:41,310 --> 00:19:41,810
OK.

428
00:19:41,810 --> 00:19:44,680


429
00:19:44,680 --> 00:19:51,500
We've seen enough of this, but
if you ask me, is cryptanalysis

430
00:19:51,500 --> 00:19:55,400
my my answer is, yes, it is.

431
00:19:55,400 --> 00:19:57,800
Sending things encrypted
with weak codes

432
00:19:57,800 --> 00:20:02,000
puts you at considerable
risk, and if you

433
00:20:02,000 --> 00:20:04,205
want to gather information
using cryptanalysis,

434
00:20:04,205 --> 00:20:08,390
it is a method that
pays off rather well.

435
00:20:08,390 --> 00:20:09,860
Is it the most important?

436
00:20:09,860 --> 00:20:12,930
No, it is not the
most important.

437
00:20:12,930 --> 00:20:18,400
Within the past 20 or 30
years, the other four--

438
00:20:18,400 --> 00:20:22,050
all of them begin
with B, by the way--

439
00:20:22,050 --> 00:20:24,510
burglary, blackmail
and bribery have

440
00:20:24,510 --> 00:20:27,810
become much more important
ways of gaining information

441
00:20:27,810 --> 00:20:29,640
for the best of all
possible reasons.

442
00:20:29,640 --> 00:20:31,830
They're cheaper.

443
00:20:31,830 --> 00:20:34,500
Cryptanalysis costs
a lot of money.

444
00:20:34,500 --> 00:20:38,400
And when you think in
terms of encryption,

445
00:20:38,400 --> 00:20:41,460
don't keep thinking in terms
of, oh, we can't use this code.

446
00:20:41,460 --> 00:20:42,390
It's been broken.

447
00:20:42,390 --> 00:20:43,630
Or go out and use that one.

448
00:20:43,630 --> 00:20:44,547
It hasn't been broken.

449
00:20:44,547 --> 00:20:48,170
That's not the way life is.

450
00:20:48,170 --> 00:20:52,970
People often ask me, can
you read such and such code.

451
00:20:52,970 --> 00:20:58,000
The answer is always yes, but
some are cheaper than others.

452
00:20:58,000 --> 00:21:00,390
Some are more
expensive than others.

453
00:21:00,390 --> 00:21:02,540
It's a dumb answer
to a dumb question.

454
00:21:02,540 --> 00:21:06,770
The core question was, can
you afford to read that code,

455
00:21:06,770 --> 00:21:09,590
and the answer to
that in many cases,

456
00:21:09,590 --> 00:21:12,890
yeah, we can read that
code, but it costs so much.

457
00:21:12,890 --> 00:21:14,580
So think in those terms.

458
00:21:14,580 --> 00:21:17,120
And I sort of want to
bend your brains a bit

459
00:21:17,120 --> 00:21:19,400
at this point to make
sure that you understand

460
00:21:19,400 --> 00:21:23,990
that the intelligence area is
very, very financially based,

461
00:21:23,990 --> 00:21:24,740
cost-based.

462
00:21:24,740 --> 00:21:27,320


463
00:21:27,320 --> 00:21:30,740
If you ask me a detailed
question like how much would

464
00:21:30,740 --> 00:21:32,990
the Soviets--

465
00:21:32,990 --> 00:21:35,210
well, I'm going
back a few years.

466
00:21:35,210 --> 00:21:37,880
We're talking now about
the former Soviet Union.

467
00:21:37,880 --> 00:21:40,070
How much would they
be willing to pay

468
00:21:40,070 --> 00:21:45,290
to get a good, long-term wiretap
on the president's telephone

469
00:21:45,290 --> 00:21:47,000
or on the ambassador's
telephone?

470
00:21:47,000 --> 00:21:49,220
The answer is, a
million dollars?

471
00:21:49,220 --> 00:21:49,943
Oh, yeah.

472
00:21:49,943 --> 00:21:51,860
They would be happy to
get something like that

473
00:21:51,860 --> 00:21:53,050
for a million dollars.

474
00:21:53,050 --> 00:21:53,870
$10 million?

475
00:21:53,870 --> 00:21:55,710
Nah, that's a bit high.

476
00:21:55,710 --> 00:21:58,820
So that's the range
we're working with.

477
00:21:58,820 --> 00:22:05,630
To get a good line on, well,
the US ambassador to Moscow,

478
00:22:05,630 --> 00:22:08,690
it's something that they would
spend a few million dollars,

479
00:22:08,690 --> 00:22:10,760
barely questioning the idea.

480
00:22:10,760 --> 00:22:12,150
So it is important.

481
00:22:12,150 --> 00:22:14,660
The other area--
people have been

482
00:22:14,660 --> 00:22:17,390
trying to work on me to change
the wording of this a bit,

483
00:22:17,390 --> 00:22:20,850
but it remains the
way it is, so it goes.

484
00:22:20,850 --> 00:22:26,330
It's become an important notion,
unathorized modifications

485
00:22:26,330 --> 00:22:27,450
of equipment.

486
00:22:27,450 --> 00:22:30,050
It's very difficult to
protect, for example,

487
00:22:30,050 --> 00:22:35,780
your cryptographic equipment at
night from the cleaning lady.

488
00:22:35,780 --> 00:22:38,660
I mean, embassies and other
military installations

489
00:22:38,660 --> 00:22:40,700
have cleaning ladies
and janitors and things

490
00:22:40,700 --> 00:22:42,830
of that sort.

491
00:22:42,830 --> 00:22:47,270
That's unavoidable, and
there is always the risk

492
00:22:47,270 --> 00:22:48,852
that your equipment
will be modified

493
00:22:48,852 --> 00:22:49,810
without you knowing it.

494
00:22:49,810 --> 00:22:54,150


495
00:22:54,150 --> 00:22:58,097
So I put this up here, and it
has become extremely important.

496
00:22:58,097 --> 00:23:03,180


497
00:23:03,180 --> 00:23:03,680
Why?

498
00:23:03,680 --> 00:23:04,610
Because it's cheap.

499
00:23:04,610 --> 00:23:13,020


500
00:23:13,020 --> 00:23:15,330
I should say a few things.

501
00:23:15,330 --> 00:23:18,030
I promised people that at
the beginning of this talk--

502
00:23:18,030 --> 00:23:20,280
and it's still the
beginning of the talk--

503
00:23:20,280 --> 00:23:24,330
that I would detail some
subjects that either I

504
00:23:24,330 --> 00:23:29,040
don't talk about much or at all,
or that you cannot expect me

505
00:23:29,040 --> 00:23:30,950
to tell the truth about.

506
00:23:30,950 --> 00:23:31,750
[LAUGHTER]

507
00:23:31,750 --> 00:23:34,080
I will detail those.

508
00:23:34,080 --> 00:23:37,860
You cannot expect me to tell the
truth about the vulnerabilities

509
00:23:37,860 --> 00:23:41,490
of any cryptographic equipment,
whether it's used by the US

510
00:23:41,490 --> 00:23:44,220
or our allies or
by our opponents.

511
00:23:44,220 --> 00:23:46,860
I don't talk about
the vulnerabilities

512
00:23:46,860 --> 00:23:48,570
of cryptographic equipment.

513
00:23:48,570 --> 00:23:51,990
My knowledge in that
area is world class,

514
00:23:51,990 --> 00:23:54,780
so much so the NSA
is not happy about me

515
00:23:54,780 --> 00:23:57,250
being loose on the streets.

516
00:23:57,250 --> 00:23:59,280
OK, that's one.

517
00:23:59,280 --> 00:24:03,160
Second, though I would tell
the truth if I talked about it,

518
00:24:03,160 --> 00:24:04,980
I don't talk about
cryptanalysis.

519
00:24:04,980 --> 00:24:08,070
That's mostly because even
among the bunch of people here,

520
00:24:08,070 --> 00:24:10,020
it's basically none
of your business.

521
00:24:10,020 --> 00:24:11,680
Cryptanalytic techniques?

522
00:24:11,680 --> 00:24:13,260
No.

523
00:24:13,260 --> 00:24:15,780
I'd be happy to help you
with a newspaper cryptogram,

524
00:24:15,780 --> 00:24:17,650
show you an easier
your way to solve them,

525
00:24:17,650 --> 00:24:19,067
but that's about
as far as I'd go.

526
00:24:19,067 --> 00:24:22,300


527
00:24:22,300 --> 00:24:26,310
The third thing is,
though, certain goings-on

528
00:24:26,310 --> 00:24:27,330
during the Gulf War.

529
00:24:27,330 --> 00:24:33,090
I was closer to the Gulf War
than most honest civilians

530
00:24:33,090 --> 00:24:34,560
were.

531
00:24:34,560 --> 00:24:39,330
Ron started to talk about
some of the aspects of that,

532
00:24:39,330 --> 00:24:43,050
but it got worse.

533
00:24:43,050 --> 00:24:46,760
Yeah, the worst day
was the day when

534
00:24:46,760 --> 00:24:49,860
I went to the general
counsel of NSA and said,

535
00:24:49,860 --> 00:24:53,090
Elizabeth, two or
three people and I

536
00:24:53,090 --> 00:24:57,480
are planning to assassinate
a foreign head of state.

537
00:24:57,480 --> 00:25:00,930
And she said, Bob.

538
00:25:00,930 --> 00:25:03,180
She was able to guess
pretty rapidly which

539
00:25:03,180 --> 00:25:06,060
foreign head of
state we had in mind,

540
00:25:06,060 --> 00:25:09,360
but still, it was not a
part of the normal business

541
00:25:09,360 --> 00:25:14,100
of the National Security
Agency to assassinate anyone,

542
00:25:14,100 --> 00:25:15,068
certainly--

543
00:25:15,068 --> 00:25:17,972
[LAUGHTER]

544
00:25:17,972 --> 00:25:19,770
And certainly not
foreign heads of state.

545
00:25:19,770 --> 00:25:24,850


546
00:25:24,850 --> 00:25:29,220
So I'm going to be
talking to some degree

547
00:25:29,220 --> 00:25:33,330
about cryptographic
protection and exploitation

548
00:25:33,330 --> 00:25:35,220
of encrypted things,
but to a greater

549
00:25:35,220 --> 00:25:37,690
degree about the
other aspects of it,

550
00:25:37,690 --> 00:25:39,550
because they are
far more important.

551
00:25:39,550 --> 00:25:42,120
And I really mean
to pound on that,

552
00:25:42,120 --> 00:25:46,620
but cryptanalysis is only one
of the weaknesses of information

553
00:25:46,620 --> 00:25:48,090
that allows exploitation.

554
00:25:48,090 --> 00:25:53,590
The others really exist,
more used, and are generally

555
00:25:53,590 --> 00:26:00,200
enough cheaper that are you
don't try to use cryptanalysis

556
00:26:00,200 --> 00:26:02,890
on some system unless the
other techniques have failed.

557
00:26:02,890 --> 00:26:06,360


558
00:26:06,360 --> 00:26:13,580
But if you're dealing
with cryptanalysis,

559
00:26:13,580 --> 00:26:20,480
now, here are some of the
things you need to think about,

560
00:26:20,480 --> 00:26:24,630
because I read in the
press, all over the place,

561
00:26:24,630 --> 00:26:30,260
in fact, all of the discussion
that goes on of people who

562
00:26:30,260 --> 00:26:33,050
say that such and
such a scheme doesn't

563
00:26:33,050 --> 00:26:35,300
have a long enough key length.

564
00:26:35,300 --> 00:26:36,830
Only 40-bits.

565
00:26:36,830 --> 00:26:40,070
Ought to have 60 or 100
or some other number.

566
00:26:40,070 --> 00:26:44,300
And I hear a good deal
of weeping and moaning

567
00:26:44,300 --> 00:26:50,610
about systems that have
inadequate key length, to which

568
00:26:50,610 --> 00:26:54,330
part of my response is,
what in the world difference

569
00:26:54,330 --> 00:26:58,110
does it make if you've got a
40-bit key or 60-bit key when

570
00:26:58,110 --> 00:27:01,060
I've already stolen
the information?

571
00:27:01,060 --> 00:27:02,810
And that's real.

572
00:27:02,810 --> 00:27:06,100
I mean that, that
a key length is

573
00:27:06,100 --> 00:27:09,310
just one factor of
cryptographic systems

574
00:27:09,310 --> 00:27:10,540
that needs to be looked at.

575
00:27:10,540 --> 00:27:11,160
Important?

576
00:27:11,160 --> 00:27:12,460
Yes, it's important.

577
00:27:12,460 --> 00:27:13,420
Most important?

578
00:27:13,420 --> 00:27:15,850
Not by any means.

579
00:27:15,850 --> 00:27:18,820
Proper custody of cryptographic
material is right there.

580
00:27:18,820 --> 00:27:21,700


581
00:27:21,700 --> 00:27:24,190
If one of your employees
is peddling this stuff

582
00:27:24,190 --> 00:27:30,790
to the Soviet Union,
that's not proper custody.

583
00:27:30,790 --> 00:27:35,500
The name Walker and
Whitworth should ring a bell.

584
00:27:35,500 --> 00:27:38,035
That was basically
a family operation.

585
00:27:38,035 --> 00:27:40,330
That's the American use
of the word "family,"

586
00:27:40,330 --> 00:27:41,905
not the Sicilian use.

587
00:27:41,905 --> 00:27:47,020
[LAUGHTER]

588
00:27:47,020 --> 00:27:50,380
Who were peddling
US Navy key material

589
00:27:50,380 --> 00:27:53,320
to the Soviets for 12 years
before they got caught.

590
00:27:53,320 --> 00:27:56,770
They got caught in the
early, early 1980s,

591
00:27:56,770 --> 00:28:00,700
and most of them at
this point are in jail

592
00:28:00,700 --> 00:28:03,580
under relatively
uncomfortable situations.

593
00:28:03,580 --> 00:28:05,770
So it's not a good
idea, but still, we

594
00:28:05,770 --> 00:28:08,350
had a tremendous
loss of information

595
00:28:08,350 --> 00:28:13,150
over that period of time because
the cryptographic material that

596
00:28:13,150 --> 00:28:15,850
was being used was also
being sold to the Soviets,

597
00:28:15,850 --> 00:28:20,610
and that's not the
way to do business.

598
00:28:20,610 --> 00:28:23,325
Now, choice in
distribution of keys.

599
00:28:23,325 --> 00:28:26,480


600
00:28:26,480 --> 00:28:31,610
I don't see this much directly
in terms of encryption keys.

601
00:28:31,610 --> 00:28:37,250
I do see it in terms of login
passwords on like Unix systems

602
00:28:37,250 --> 00:28:41,450
and systems like it that require
a password before you log in.

603
00:28:41,450 --> 00:28:46,170
You find on that in many
cases, in too many cases,

604
00:28:46,170 --> 00:28:48,140
people are using
names of girlfriends,

605
00:28:48,140 --> 00:28:51,650
names of dogs and things
like that for passwords.

606
00:28:51,650 --> 00:28:58,510
So if I know Ron's wife's
name and his dog's name,

607
00:28:58,510 --> 00:29:01,546
I'm halfway there.

608
00:29:01,546 --> 00:29:04,680
I mean, he will claim
that his password--

609
00:29:04,680 --> 00:29:05,180
[INAUDIBLE]

610
00:29:05,180 --> 00:29:05,710
Blind.

611
00:29:05,710 --> 00:29:07,323
OK, that's good.

612
00:29:07,323 --> 00:29:14,430
[LAUGHTER]

613
00:29:14,430 --> 00:29:17,430
Actually, I got a
percentage of that, a paper

614
00:29:17,430 --> 00:29:22,170
that I published
in the late 1970s

615
00:29:22,170 --> 00:29:27,600
where we came to the conclusion
that about 5% of all passwords

616
00:29:27,600 --> 00:29:28,830
were names of dogs.

617
00:29:28,830 --> 00:29:32,150


618
00:29:32,150 --> 00:29:35,443
Still, nothing much
I can do except just

619
00:29:35,443 --> 00:29:36,860
to keep pointing
to it and saying,

620
00:29:36,860 --> 00:29:38,910
actually, it is important.

621
00:29:38,910 --> 00:29:42,632
And if you let users
pick encryption keys,

622
00:29:42,632 --> 00:29:43,340
you had a patrol.

623
00:29:43,340 --> 00:29:45,960


624
00:29:45,960 --> 00:29:48,850
This happened in some ways
which were quite publicized

625
00:29:48,850 --> 00:29:50,200
during World War II.

626
00:29:50,200 --> 00:29:52,080
Adequate key space.

627
00:29:52,080 --> 00:29:56,700
40 bits is a fairly small key,
and you want something bigger

628
00:29:56,700 --> 00:29:59,770
than that, but it is only
one of a number of factors.

629
00:29:59,770 --> 00:30:01,980
Another thing which is
quite important people don't

630
00:30:01,980 --> 00:30:07,410
pay much attention
to is that that text

631
00:30:07,410 --> 00:30:12,280
that you're trying to produce is
typically awfully predictable--

632
00:30:12,280 --> 00:30:17,620
ASCII characters, perhaps,
where the letters come together

633
00:30:17,620 --> 00:30:20,710
to make words and the words
come together to make sentences.

634
00:30:20,710 --> 00:30:23,500
If you reduce that
predictability--

635
00:30:23,500 --> 00:30:26,110
and any good
compression scheme will,

636
00:30:26,110 --> 00:30:28,540
and the better
the compression is

637
00:30:28,540 --> 00:30:31,698
the better this works
to help cryptography--

638
00:30:31,698 --> 00:30:34,630


639
00:30:34,630 --> 00:30:37,060
it's a useful
thing, fairly cheap,

640
00:30:37,060 --> 00:30:40,660
and it may have even other
advantages that might help.

641
00:30:40,660 --> 00:30:45,220
You compress first
and then encrypt,

642
00:30:45,220 --> 00:30:49,120
and this actually makes it
more difficult to decrypt text.

643
00:30:49,120 --> 00:30:51,400
Now, when I say
system design where

644
00:30:51,400 --> 00:30:55,720
the entire key participates
in each encryption operation.

645
00:30:55,720 --> 00:30:58,660
I'm reminded of
newspaper cryptograms

646
00:30:58,660 --> 00:31:03,270
where all you have is some
permutation of the alphabet.

647
00:31:03,270 --> 00:31:05,650
A turns into X every
time A appears,

648
00:31:05,650 --> 00:31:09,490
and B turns into V
every time it appears.

649
00:31:09,490 --> 00:31:11,920
It was just a permutation.

650
00:31:11,920 --> 00:31:15,340
Now, in that case,
you've got a key length

651
00:31:15,340 --> 00:31:18,400
which is absolutely enormous.

652
00:31:18,400 --> 00:31:20,080
Anybody that wants
to work out what

653
00:31:20,080 --> 00:31:24,220
is 26 factorial, that's
the number of possible keys

654
00:31:24,220 --> 00:31:25,360
for the scheme.

655
00:31:25,360 --> 00:31:29,470
And newspaper cryptograms
choose a key, one key out

656
00:31:29,470 --> 00:31:32,320
of 26 factorial possible ones.

657
00:31:32,320 --> 00:31:36,520
But the entire key does not
participate in each encryption

658
00:31:36,520 --> 00:31:37,720
operation.

659
00:31:37,720 --> 00:31:39,200
It's quite the reverse of that.

660
00:31:39,200 --> 00:31:40,700
So as a matter of
fact, all you need

661
00:31:40,700 --> 00:31:44,860
is something on the
order of 6, 8, 10 words

662
00:31:44,860 --> 00:31:47,710
to be able to come
up with and know

663
00:31:47,710 --> 00:31:53,720
you've come up with the
plain text of the message.

664
00:31:53,720 --> 00:31:57,590
By the way, I'm a very
experienced and very relaxed

665
00:31:57,590 --> 00:31:58,370
speaker.

666
00:31:58,370 --> 00:32:02,340
And I welcome questions,
complaints, whatever.

667
00:32:02,340 --> 00:32:04,370
The more interaction,
in fact, the better.

668
00:32:04,370 --> 00:32:08,310
If anybody wants to say
something, just say it.

669
00:32:08,310 --> 00:32:16,220


670
00:32:16,220 --> 00:32:17,630
That's purely cryptographic.

671
00:32:17,630 --> 00:32:21,350
But it's only part of the story
because cryptography is only

672
00:32:21,350 --> 00:32:22,150
part of the story.

673
00:32:22,150 --> 00:32:32,545


674
00:32:32,545 --> 00:32:35,350
This gets into a
number of things--

675
00:32:35,350 --> 00:32:41,380
carelessness--
important in many cases,

676
00:32:41,380 --> 00:32:43,530
you find that if you
want some information,

677
00:32:43,530 --> 00:32:45,900
you don't need to
do cryptanalysis.

678
00:32:45,900 --> 00:32:49,290
You just go in the back of
the organization's parking lot

679
00:32:49,290 --> 00:32:51,810
and pick the stuff you want
out of the garbage can, which

680
00:32:51,810 --> 00:32:53,460
is where it got thrown.

681
00:32:53,460 --> 00:32:54,810
And it shouldn't happen.

682
00:32:54,810 --> 00:32:56,280
But it does happen.

683
00:32:56,280 --> 00:33:01,550
Overconfidence-- well,
overconfidence probably

684
00:33:01,550 --> 00:33:04,910
reached its peak
during World War II

685
00:33:04,910 --> 00:33:09,620
when the Germans had this
encryption machine called

686
00:33:09,620 --> 00:33:10,730
Enigma.

687
00:33:10,730 --> 00:33:13,970
And they thought, hey, it
was designed by a German.

688
00:33:13,970 --> 00:33:17,722


689
00:33:17,722 --> 00:33:22,880
By the way, who speaks
fluent German here?

690
00:33:22,880 --> 00:33:24,776
One, two.

691
00:33:24,776 --> 00:33:26,230
Yeah, OK.

692
00:33:26,230 --> 00:33:29,460
Oh, I'll apologize in advance.

693
00:33:29,460 --> 00:33:32,520
Probably I'll write some
things in German, some of which

694
00:33:32,520 --> 00:33:34,380
will be misspelled,
and some of which

695
00:33:34,380 --> 00:33:36,120
will not be politically correct.

696
00:33:36,120 --> 00:33:37,380
But you'll understand that.

697
00:33:37,380 --> 00:33:41,110


698
00:33:41,110 --> 00:33:44,770
One of the things that came out
of World War II that provided

699
00:33:44,770 --> 00:33:49,240
a long-term lesson
for the Allies

700
00:33:49,240 --> 00:33:55,440
was that the Germans drastically
underestimated the time, money,

701
00:33:55,440 --> 00:34:02,230
expense that the Allies would
put into decrypting the Enigma

702
00:34:02,230 --> 00:34:03,370
traffic.

703
00:34:03,370 --> 00:34:06,350
They totally
underestimated that.

704
00:34:06,350 --> 00:34:10,790
And since then, a good deal
of attention has gone--

705
00:34:10,790 --> 00:34:13,060
and they set up
committees several times

706
00:34:13,060 --> 00:34:18,070
during the war to reassess
the Enigma encryption scheme.

707
00:34:18,070 --> 00:34:23,630
And the answer always came
back from these committees

708
00:34:23,630 --> 00:34:26,480
that everything was fine.

709
00:34:26,480 --> 00:34:30,409
It's just too many
possibilities here

710
00:34:30,409 --> 00:34:32,150
for anyone to get
in any trouble.

711
00:34:32,150 --> 00:34:35,000
All the time, the Allies
were reading them.

712
00:34:35,000 --> 00:34:36,830
It makes you wonder
about committees.

713
00:34:36,830 --> 00:34:37,330
Sir?

714
00:34:37,330 --> 00:34:39,030
It makes you wonder
about committees.

715
00:34:39,030 --> 00:34:39,860
Yes.

716
00:34:39,860 --> 00:34:42,260
In David Cohen's
book, David Cohen

717
00:34:42,260 --> 00:34:45,739
talks about those
committees in some detail.

718
00:34:45,739 --> 00:34:48,710
But hey, the machine was
designed by a German.

719
00:34:48,710 --> 00:34:53,320


720
00:34:53,320 --> 00:34:57,280
But that overconfidence
can really kill you.

721
00:34:57,280 --> 00:34:59,950
And in many cases,
people nowadays

722
00:34:59,950 --> 00:35:03,490
are using, on the internet
and on the world wide web,

723
00:35:03,490 --> 00:35:06,580
encryption schemes which
are plainly too weak to do

724
00:35:06,580 --> 00:35:07,870
the job that needs to be done.

725
00:35:07,870 --> 00:35:11,720


726
00:35:11,720 --> 00:35:17,630
Information is most often
lost by theft or purchase.

727
00:35:17,630 --> 00:35:23,420
Most inside jobs-- now, I
have in my pocket a floppy--

728
00:35:23,420 --> 00:35:25,880
not a floppy disks, what
do you call those guys?

729
00:35:25,880 --> 00:35:32,030
A CD that is enough to
contain the whole Encyclopedia

730
00:35:32,030 --> 00:35:33,470
Britannica.

731
00:35:33,470 --> 00:35:35,760
It's a lot of information.

732
00:35:35,760 --> 00:35:38,030
And if I walk out
the door, no one's

733
00:35:38,030 --> 00:35:40,580
going to know that I have
one of these in my pocket.

734
00:35:40,580 --> 00:35:42,170
And that's true at
any installation,

735
00:35:42,170 --> 00:35:44,410
whether it's diplomatic,
or military, or whatever.

736
00:35:44,410 --> 00:35:47,380


737
00:35:47,380 --> 00:35:52,570
Information just gets
packaged into tiny forms.

738
00:35:52,570 --> 00:35:58,400
It started with floppy disks
and went from there to CDs.

739
00:35:58,400 --> 00:36:01,800


740
00:36:01,800 --> 00:36:04,290
Most jobs are, in
fact, inside jobs

741
00:36:04,290 --> 00:36:05,760
or have an inside component.

742
00:36:05,760 --> 00:36:10,320
I mentioned [INAUDIBLE]
before, but it keeps coming up.

743
00:36:10,320 --> 00:36:14,670
And these two--
microphones are everywhere,

744
00:36:14,670 --> 00:36:18,090
and antennas are everywhere.

745
00:36:18,090 --> 00:36:23,490
Just again reflects my paranoia
about that sort of thing.

746
00:36:23,490 --> 00:36:28,110
But that paranoia tends
to be very global.

747
00:36:28,110 --> 00:36:31,590
I mean, suppose I'm at
a military base working

748
00:36:31,590 --> 00:36:32,910
in a military base.

749
00:36:32,910 --> 00:36:37,350
And a piece of cryptographic
equipment that we ordered

750
00:36:37,350 --> 00:36:40,440
arrives on the loading
dock, perfectly good

751
00:36:40,440 --> 00:36:45,240
KG84 encryption machine
that we ordered.

752
00:36:45,240 --> 00:36:47,940
And hey, it arrived.

753
00:36:47,940 --> 00:36:53,430
My attitude towards it is,
even though it says KG84 on it

754
00:36:53,430 --> 00:36:57,390
and it looks like KG84, it looks
like the last KG84 that I saw,

755
00:36:57,390 --> 00:36:59,670
what is it?

756
00:36:59,670 --> 00:37:01,970
How did it get here?

757
00:37:01,970 --> 00:37:04,680
Who sent it here?

758
00:37:04,680 --> 00:37:08,700
What were the motives
of whoever sent it here?

759
00:37:08,700 --> 00:37:14,820
Now, there we've crossed the
line into clinical paranoia.

760
00:37:14,820 --> 00:37:17,862
But that is still the way that
I think about such a thing

761
00:37:17,862 --> 00:37:20,070
and that I recommend that
others think about a thing.

762
00:37:20,070 --> 00:37:23,970
You need to carry paranoia
about to that point.

763
00:37:23,970 --> 00:37:26,400
For the last several
years I was in NSA,

764
00:37:26,400 --> 00:37:29,730
I was one of the roadblocks
in the development

765
00:37:29,730 --> 00:37:33,390
and design of new US
cryptographic equipment.

766
00:37:33,390 --> 00:37:36,270
And I would get,
every few months,

767
00:37:36,270 --> 00:37:39,150
about any project that was
under development design

768
00:37:39,150 --> 00:37:43,620
a briefing by the designers
as to the details of what

769
00:37:43,620 --> 00:37:45,420
it was they were designing.

770
00:37:45,420 --> 00:37:48,010
And we would have
these conversations

771
00:37:48,010 --> 00:37:55,250
that went like, well, is your
machine resistant to this sort

772
00:37:55,250 --> 00:37:56,225
of attack.

773
00:37:56,225 --> 00:37:59,590
The response was, oh, no
one would ever do that.

774
00:37:59,590 --> 00:38:02,540
Uh huh.

775
00:38:02,540 --> 00:38:04,340
Well, you need to
get on my side.

776
00:38:04,340 --> 00:38:06,050
You need to get on
the paranoid side

777
00:38:06,050 --> 00:38:08,150
that indeed no one would
ever do that, sure.

778
00:38:08,150 --> 00:38:21,110


779
00:38:21,110 --> 00:38:23,210
I've run out of new
graphs at probably

780
00:38:23,210 --> 00:38:28,400
just the right time because
this is part of the important--

781
00:38:28,400 --> 00:38:33,375
the most important of the
things I have to address.

782
00:38:33,375 --> 00:38:35,910


783
00:38:35,910 --> 00:38:36,810
But I'll leave it up.

784
00:38:36,810 --> 00:38:39,477
Might as well stare at it.

785
00:38:39,477 --> 00:38:48,530


786
00:38:48,530 --> 00:38:49,030
OK.

787
00:38:49,030 --> 00:38:52,000
So World War II had
a number of effects

788
00:38:52,000 --> 00:38:57,670
on protection of information
that are old but still valid

789
00:38:57,670 --> 00:39:00,070
in very many important ways.

790
00:39:00,070 --> 00:39:03,470


791
00:39:03,470 --> 00:39:08,500
Certainly, this kind of thing
is an important part of it.

792
00:39:08,500 --> 00:39:12,340


793
00:39:12,340 --> 00:39:17,110
And the overconfidence
probably was the most important

794
00:39:17,110 --> 00:39:20,680
negative aspect
of the way people

795
00:39:20,680 --> 00:39:22,660
thought about protection
of information

796
00:39:22,660 --> 00:39:29,400
and actually did the protection.

797
00:39:29,400 --> 00:39:37,100
First off was loss of machines.

798
00:39:37,100 --> 00:39:40,500
During the war--

799
00:39:40,500 --> 00:39:41,970
I'm talking about World War II.

800
00:39:41,970 --> 00:39:45,760
And I'm talking about the
war that began in 1939.

801
00:39:45,760 --> 00:39:46,380
Remember that.

802
00:39:46,380 --> 00:39:47,940
Don't think about Pearl Harbor.

803
00:39:47,940 --> 00:39:52,380
Think about the actual
war, which began in 1939.

804
00:39:52,380 --> 00:39:58,230
The Germans lost quite a few
Enigmas that were currently

805
00:39:58,230 --> 00:40:01,410
in use to send messages.

806
00:40:01,410 --> 00:40:05,458
Lost them, that's a
dishonest way of putting.

807
00:40:05,458 --> 00:40:07,500
We stole a large number
of them from the Germans.

808
00:40:07,500 --> 00:40:13,376


809
00:40:13,376 --> 00:40:16,570
And as happens fairly
often, the Germans

810
00:40:16,570 --> 00:40:22,030
changed the internal wiring
of the Enigma machine.

811
00:40:22,030 --> 00:40:24,700
It made it very difficult--
and by difficult,

812
00:40:24,700 --> 00:40:26,200
I mean expensive--

813
00:40:26,200 --> 00:40:31,150
for us to continue cryptanalysis
with the modified Enigma.

814
00:40:31,150 --> 00:40:35,320
And what we would do--
we knew that the Germans,

815
00:40:35,320 --> 00:40:38,140
in order to get decent
weather forecasting,

816
00:40:38,140 --> 00:40:39,880
stationed a whole bunch of--

817
00:40:39,880 --> 00:40:43,390
well, not a whole bunch,
a few, two or three--

818
00:40:43,390 --> 00:40:47,020
ships in the Atlantic,
Central Atlantic,

819
00:40:47,020 --> 00:40:50,260
to get them reasonable
weather forecasts

820
00:40:50,260 --> 00:40:54,430
of Germany a few days later.

821
00:40:54,430 --> 00:40:59,110
Those ships had up-to-date
Enigmas on board

822
00:40:59,110 --> 00:41:02,440
and have the keying
material for them.

823
00:41:02,440 --> 00:41:07,630
We would send out a Naval
vessel to basically demolish

824
00:41:07,630 --> 00:41:10,960
that weather ship,
sink if necessary,

825
00:41:10,960 --> 00:41:13,600
and steal the Enigma
and the key material

826
00:41:13,600 --> 00:41:16,327
from it, which
would typically make

827
00:41:16,327 --> 00:41:18,410
all everything happy for
another couple of months.

828
00:41:18,410 --> 00:41:20,020
We did that constantly
during the war.

829
00:41:20,020 --> 00:41:23,430


830
00:41:23,430 --> 00:41:28,500
I remember sitting at
dinner with the Brit who

831
00:41:28,500 --> 00:41:31,110
was actually responsible
for sending those things up.

832
00:41:31,110 --> 00:41:34,640
And I said, you're
morally bankrupt.

833
00:41:34,640 --> 00:41:36,210
He smiled and said, you bet.

834
00:41:36,210 --> 00:41:40,180


835
00:41:40,180 --> 00:41:43,300
So there was a lot
of loss of machines,

836
00:41:43,300 --> 00:41:47,740
whereas the Brits had
actually a machine which

837
00:41:47,740 --> 00:41:50,230
was very similar to the Enigma.

838
00:41:50,230 --> 00:41:51,460
It had a different name.

839
00:41:51,460 --> 00:41:52,810
It was called Typex.

840
00:41:52,810 --> 00:41:57,700
And I don't know whether or not
they lost any during the war.

841
00:41:57,700 --> 00:41:59,050
I believe they did not.

842
00:41:59,050 --> 00:42:01,550
The US had also a
very small machine,

843
00:42:01,550 --> 00:42:06,370
which was called
the SIGABA, which

844
00:42:06,370 --> 00:42:10,030
one can be quite certain we did
not lose any during the war,

845
00:42:10,030 --> 00:42:13,090
did not lose custody of a
single one through the war.

846
00:42:13,090 --> 00:42:15,730
And that provided a
great deal in protection.

847
00:42:15,730 --> 00:42:19,750
Codes using the American machine
were never read during the war.

848
00:42:19,750 --> 00:42:22,630
And to a large extent,
that was because we

849
00:42:22,630 --> 00:42:24,280
didn't lose track of them.

850
00:42:24,280 --> 00:42:27,040


851
00:42:27,040 --> 00:42:38,650
Now, some habits of the
German signal authorities

852
00:42:38,650 --> 00:42:40,450
were a bit unfortunate.

853
00:42:40,450 --> 00:42:44,960


854
00:42:44,960 --> 00:42:48,920
For example, for
most of the war,

855
00:42:48,920 --> 00:42:52,980
every German military message
began with the three letters

856
00:42:52,980 --> 00:42:53,480
ANX.

857
00:42:53,480 --> 00:42:57,980


858
00:42:57,980 --> 00:42:58,520
Why?

859
00:42:58,520 --> 00:43:00,710
Well, I'm not going to
teach you much German.

860
00:43:00,710 --> 00:43:03,410
But this is just a
way you would normally

861
00:43:03,410 --> 00:43:05,450
expect a message to begin.

862
00:43:05,450 --> 00:43:07,400
The X stands for space.

863
00:43:07,400 --> 00:43:11,540
It just says to
somebody, and who cares.

864
00:43:11,540 --> 00:43:18,290
But having the message begin
with a known set of characters

865
00:43:18,290 --> 00:43:21,770
is just asking for
trouble because when

866
00:43:21,770 --> 00:43:24,500
you get into key search now,
you can speed the thing up

867
00:43:24,500 --> 00:43:31,740
enormously by knowing
that the message begins

868
00:43:31,740 --> 00:43:32,610
with these letters.

869
00:43:32,610 --> 00:43:34,230
This was changed
later in the war.

870
00:43:34,230 --> 00:43:36,400
But for the first two or
three years of the war,

871
00:43:36,400 --> 00:43:37,470
this was always used.

872
00:43:37,470 --> 00:43:42,300
Another habit which the
Allies found very useful

873
00:43:42,300 --> 00:43:46,750
was the habit of, particularly
with a high ranking

874
00:43:46,750 --> 00:43:50,460
and important officer,
never abbreviating

875
00:43:50,460 --> 00:43:51,645
his name or position.

876
00:43:51,645 --> 00:43:54,360


877
00:43:54,360 --> 00:43:59,580
And that's combined with a
particular characteristic

878
00:43:59,580 --> 00:44:05,760
of the Enigma that--

879
00:44:05,760 --> 00:44:07,380
it doesn't seem very important.

880
00:44:07,380 --> 00:44:10,080
But it never
encrypted a character

881
00:44:10,080 --> 00:44:11,640
into the same character.

882
00:44:11,640 --> 00:44:13,920
If the plain text
character was A,

883
00:44:13,920 --> 00:44:16,890
the ciphertext
character was never A.

884
00:44:16,890 --> 00:44:22,260
Now, that means that after
you get a few dozen characters

885
00:44:22,260 --> 00:44:25,800
and you have a suspected
text might have appeared

886
00:44:25,800 --> 00:44:27,630
in the plain text,
you will simply

887
00:44:27,630 --> 00:44:30,990
run it through until you
don't get any letter going

888
00:44:30,990 --> 00:44:34,150
into the same letter and
with some probability, which

889
00:44:34,150 --> 00:44:38,670
I can compute, place that text.

890
00:44:38,670 --> 00:44:42,090
Now, the kind of thing that
would happen would be--

891
00:44:42,090 --> 00:44:44,340
I'm going to write down
the beginning of a message

892
00:44:44,340 --> 00:44:46,020
to a very high ranking officer.

893
00:44:46,020 --> 00:44:49,320
And here, I will have
both the misspelling

894
00:44:49,320 --> 00:44:51,120
and the politically
incorrect material

895
00:44:51,120 --> 00:44:52,200
that I spoke of earlier.

896
00:44:52,200 --> 00:44:55,040


897
00:44:55,040 --> 00:44:59,055
ANX, that's just
to, to somebody.

898
00:44:59,055 --> 00:45:05,300


899
00:45:05,300 --> 00:45:10,760
I'm going to write the rank of
a very high ranking SS officer,

900
00:45:10,760 --> 00:45:12,650
partly just so you
see how long it is.

901
00:45:12,650 --> 00:45:24,680


902
00:45:24,680 --> 00:45:25,340
Oh, hey.

903
00:45:25,340 --> 00:45:30,640


904
00:45:30,640 --> 00:45:31,360
That's the title.

905
00:45:31,360 --> 00:45:34,450


906
00:45:34,450 --> 00:45:36,575
By the way, did I
spell it incorrectly?

907
00:45:36,575 --> 00:45:37,640
That's correct.

908
00:45:37,640 --> 00:45:38,320
That is correct?

909
00:45:38,320 --> 00:45:38,820
Yeah.

910
00:45:38,820 --> 00:45:40,010
OK, fine.

911
00:45:40,010 --> 00:45:44,260
Other people have not been
sure whether they should be HN.

912
00:45:44,260 --> 00:45:45,906
I believe this is correct.

913
00:45:45,906 --> 00:45:46,860
[INAUDIBLE]

914
00:45:46,860 --> 00:45:48,180
Yeah.

915
00:45:48,180 --> 00:45:50,760
But that's not the end of
it because you would never

916
00:45:50,760 --> 00:45:52,860
omit or abbreviate his name.

917
00:45:52,860 --> 00:45:55,740
It just would not be a
habit that would ever occur

918
00:45:55,740 --> 00:45:57,008
in the German signal service.

919
00:45:57,008 --> 00:45:58,800
So we have to continue
with the guy's name.

920
00:45:58,800 --> 00:46:03,410


921
00:46:03,410 --> 00:46:05,660
And this will go on forever.

922
00:46:05,660 --> 00:46:19,070


923
00:46:19,070 --> 00:46:20,660
And I'm not exaggerating.

924
00:46:20,660 --> 00:46:24,170
This kind of thing at these
lengths actually occurred.

925
00:46:24,170 --> 00:46:26,785


926
00:46:26,785 --> 00:46:28,160
I haven't gotten
to his name yet.

927
00:46:28,160 --> 00:46:38,700


928
00:46:38,700 --> 00:46:42,913
And now, well, for the
Germans, the apologies

929
00:46:42,913 --> 00:46:43,830
apply here, of course.

930
00:46:43,830 --> 00:46:47,690


931
00:46:47,690 --> 00:46:49,255
Pardon?

932
00:46:49,255 --> 00:46:50,030
[INAUDIBLE]

933
00:46:50,030 --> 00:46:50,530
Oh.

934
00:46:50,530 --> 00:46:53,770
This is not
particularly untypical.

935
00:46:53,770 --> 00:46:58,370
This kind of thing happened in
actual messages all the time.

936
00:46:58,370 --> 00:47:00,340
And when you have the--

937
00:47:00,340 --> 00:47:02,800
and with a high
ranking like this,

938
00:47:02,800 --> 00:47:06,070
you knew he was traveling
around, planning

939
00:47:06,070 --> 00:47:07,240
to visit North Africa.

940
00:47:07,240 --> 00:47:10,750
You could be absolutely sure
that his name and title, not

941
00:47:10,750 --> 00:47:15,070
abbreviated, would occur in
a message where he was going.

942
00:47:15,070 --> 00:47:18,280
So you knew that this was
highly likely to appear.

943
00:47:18,280 --> 00:47:20,680
So you could put this
as a probable text

944
00:47:20,680 --> 00:47:22,240
against the encrypted text.

945
00:47:22,240 --> 00:47:26,950
And it meant that key search
became almost a triviality

946
00:47:26,950 --> 00:47:29,240
once you knew that
kind of thing.

947
00:47:29,240 --> 00:47:31,060
The other kinds of
things which would

948
00:47:31,060 --> 00:47:34,510
be sigma carelessness
of one kind or another

949
00:47:34,510 --> 00:47:39,220
that some German unit
that had an Enigma

950
00:47:39,220 --> 00:47:42,280
and was not doing anything
of the slightest importance

951
00:47:42,280 --> 00:47:45,910
was ordered to report every day.

952
00:47:45,910 --> 00:47:47,500
OK, they reported every day.

953
00:47:47,500 --> 00:48:03,570


954
00:48:03,570 --> 00:48:06,870
And that's the precise German
that would, in fact, be

955
00:48:06,870 --> 00:48:10,140
used to say, nothing to report.

956
00:48:10,140 --> 00:48:14,610
And there was one unit that
throughout the war dependably,

957
00:48:14,610 --> 00:48:17,100
every day nothing to report.

958
00:48:17,100 --> 00:48:20,130


959
00:48:20,130 --> 00:48:21,360
OK.

960
00:48:21,360 --> 00:48:24,120
But don't imagine the
Germans were dumb or dumber

961
00:48:24,120 --> 00:48:25,890
than others.

962
00:48:25,890 --> 00:48:27,120
The Allies were dumber.

963
00:48:27,120 --> 00:48:30,560
And the dumbest was the US.

964
00:48:30,560 --> 00:48:37,220
And the use of a code in World
War II that the Brits knew

965
00:48:37,220 --> 00:48:41,610
had been read by the Germans
during World War I, Admiralty

966
00:48:41,610 --> 00:48:42,110
Cipher.

967
00:48:42,110 --> 00:48:46,190


968
00:48:46,190 --> 00:48:48,320
I have a copy that
I didn't bring.

969
00:48:48,320 --> 00:48:50,930
It's called Admiralty
Cipher Number 2.

970
00:48:50,930 --> 00:48:52,765
I keep it because if
there's another war,

971
00:48:52,765 --> 00:48:53,765
it might come in useful.

972
00:48:53,765 --> 00:49:03,460


973
00:49:03,460 --> 00:49:07,810
So that kind of thing goes
under the general category

974
00:49:07,810 --> 00:49:12,550
of carelessness, which it
is, inappropriate procedures,

975
00:49:12,550 --> 00:49:14,650
the predictability.

976
00:49:14,650 --> 00:49:16,960
The cases where
identical messages were

977
00:49:16,960 --> 00:49:19,330
sent over different systems--

978
00:49:19,330 --> 00:49:23,350
what used to happen as
regularly as clockwork

979
00:49:23,350 --> 00:49:25,600
is that the German
military would

980
00:49:25,600 --> 00:49:30,520
go to a new system, a
complete new scheme.

981
00:49:30,520 --> 00:49:32,590
And they'd say,
hey, you start using

982
00:49:32,590 --> 00:49:36,430
the scheme on November 20.

983
00:49:36,430 --> 00:49:40,060
And what would
happen would be Unit

984
00:49:40,060 --> 00:49:43,210
A would send a message
to Unit B. Unit B

985
00:49:43,210 --> 00:49:46,330
would send something back that
says, hey, we can't read that.

986
00:49:46,330 --> 00:49:49,260
We don't have the new code yet.

987
00:49:49,260 --> 00:49:49,770
OK.

988
00:49:49,770 --> 00:49:52,530
So they would send the
message again in the old code.

989
00:49:52,530 --> 00:49:56,480


990
00:49:56,480 --> 00:49:59,370
That happened on all parties.

991
00:49:59,370 --> 00:50:00,360
The Germans did it.

992
00:50:00,360 --> 00:50:02,000
But also, the Allies did that.

993
00:50:02,000 --> 00:50:03,530
And it was reasonably common.

994
00:50:03,530 --> 00:50:07,330


995
00:50:07,330 --> 00:50:09,730
Some of those problems
were very serious and not

996
00:50:09,730 --> 00:50:12,700
very widely publicized,
particularly the fact

997
00:50:12,700 --> 00:50:15,970
that Admiralty Code and
Admiralty Cipher Number

998
00:50:15,970 --> 00:50:18,400
2 was used throughout
the war, from beginning

999
00:50:18,400 --> 00:50:24,160
to end, to order the track of
convoys across the Atlantic.

1000
00:50:24,160 --> 00:50:26,950
So if the German
submarine service

1001
00:50:26,950 --> 00:50:33,190
wanted to know where the
current convoy was going,

1002
00:50:33,190 --> 00:50:35,940
it was in Admiralty
Cipher, encoded

1003
00:50:35,940 --> 00:50:38,470
in Admiralty Cipher Number 2.

1004
00:50:38,470 --> 00:50:42,370
And one tends to read more about
all the wonderful things that

1005
00:50:42,370 --> 00:50:46,960
happened at Bletchley Park,
rather than some of the equally

1006
00:50:46,960 --> 00:50:50,020
wonderful, long-term
of course, things

1007
00:50:50,020 --> 00:50:53,750
that the Allies did that
were comparable or worse.

1008
00:50:53,750 --> 00:50:55,170
I shudder to think of that.

1009
00:50:55,170 --> 00:50:55,970
But it did happen.

1010
00:50:55,970 --> 00:50:59,260


1011
00:50:59,260 --> 00:51:00,063
Speak.

1012
00:51:00,063 --> 00:51:02,230
Why did they [INAUDIBLE]
three years of code if they

1013
00:51:02,230 --> 00:51:04,180
knew it had been broken?

1014
00:51:04,180 --> 00:51:07,798
Yeah, that's my question, too.

1015
00:51:07,798 --> 00:51:10,300
[INAUDIBLE] policy [INAUDIBLE].

1016
00:51:10,300 --> 00:51:13,150


1017
00:51:13,150 --> 00:51:15,490
Maybe that's the reason.

1018
00:51:15,490 --> 00:51:17,380
I can't answer the question.

1019
00:51:17,380 --> 00:51:20,350
I mean, this was
done by the Brits,

1020
00:51:20,350 --> 00:51:22,070
done by Naval intelligence.

1021
00:51:22,070 --> 00:51:24,520
And that's what they did.

1022
00:51:24,520 --> 00:51:27,070
Why they did it beats me.

1023
00:51:27,070 --> 00:51:28,570
It's not widely publicized.

1024
00:51:28,570 --> 00:51:31,570
I don't think that appears, for
example, in Cohen's book, which

1025
00:51:31,570 --> 00:51:34,388
talks mostly about Bletchley
Park kinds of things.

1026
00:51:34,388 --> 00:51:36,430
But Bletchley Park was
not the only kind of thing

1027
00:51:36,430 --> 00:51:38,293
that happened.

1028
00:51:38,293 --> 00:51:41,060
[INAUDIBLE]

1029
00:51:41,060 --> 00:51:44,060
No, we lost convoys
because of that.

1030
00:51:44,060 --> 00:51:46,680
It's bad, actually very bad.

1031
00:51:46,680 --> 00:51:51,680


1032
00:51:51,680 --> 00:51:57,230
Now, after World War II,
the intelligence business

1033
00:51:57,230 --> 00:51:59,780
changed in a number of
fairly important ways.

1034
00:51:59,780 --> 00:52:03,580


1035
00:52:03,580 --> 00:52:11,980
Because of the downfall of the
Enigma and machines like it,

1036
00:52:11,980 --> 00:52:13,930
the intelligence
services of countries

1037
00:52:13,930 --> 00:52:16,430
decided to go to
different schemes,

1038
00:52:16,430 --> 00:52:26,920
go away from these rotating,
highly mechanical cipher

1039
00:52:26,920 --> 00:52:31,090
machines we used in World War
II to go to better things.

1040
00:52:31,090 --> 00:52:34,810
By the way, I brought with me
a cipher machine was actually

1041
00:52:34,810 --> 00:52:38,650
used all over the place
by the US military

1042
00:52:38,650 --> 00:52:39,730
during World War II.

1043
00:52:39,730 --> 00:52:41,260
It's called the M-209.

1044
00:52:41,260 --> 00:52:45,258
It's a lovely machine, in
the sense that the Army was

1045
00:52:45,258 --> 00:52:46,800
able to bring this
into the trenches,

1046
00:52:46,800 --> 00:52:50,740
the frontline units
in the trenches,

1047
00:52:50,740 --> 00:52:53,650
for example in the
desert in North Africa.

1048
00:52:53,650 --> 00:52:55,960
And after six months,
it would still

1049
00:52:55,960 --> 00:53:00,400
be working, just beautiful
construction and a fairly

1050
00:53:00,400 --> 00:53:02,840
decent encryption scheme.

1051
00:53:02,840 --> 00:53:05,410
The instructions that went
with it during the war

1052
00:53:05,410 --> 00:53:07,990
said, you can expect
that the enemy will not

1053
00:53:07,990 --> 00:53:12,490
read something enciphered
with this machine in less

1054
00:53:12,490 --> 00:53:14,110
than 24 hours.

1055
00:53:14,110 --> 00:53:15,860
And it may not seem like much.

1056
00:53:15,860 --> 00:53:23,050
But for frontline activities,
often a day just fine,

1057
00:53:23,050 --> 00:53:25,300
particularly if you
know it's just a day.

1058
00:53:25,300 --> 00:53:29,380
So it was a good machine
and beautifully constructed.

1059
00:53:29,380 --> 00:53:33,860
And its secrecy was
perfectly adequate for what

1060
00:53:33,860 --> 00:53:36,680
it was being used for.

1061
00:53:36,680 --> 00:53:38,480
Not much better
than that, though.

1062
00:53:38,480 --> 00:53:44,920


1063
00:53:44,920 --> 00:53:49,330
Other kinds of things
began to show up

1064
00:53:49,330 --> 00:53:51,760
beginning about 10
years after the war.

1065
00:53:51,760 --> 00:53:55,540
For example, if the
Russian government

1066
00:53:55,540 --> 00:54:00,790
gives a beautiful
four foot in diameter

1067
00:54:00,790 --> 00:54:03,520
great seal of the United
States to the US ambassador

1068
00:54:03,520 --> 00:54:04,645
to hang up in the embassy--

1069
00:54:04,645 --> 00:54:07,870


1070
00:54:07,870 --> 00:54:10,580
I'm going back to
what I said earlier.

1071
00:54:10,580 --> 00:54:13,750
My question is not, is
there a microphone in it?

1072
00:54:13,750 --> 00:54:16,690
My question is, where
is the microphone in it?

1073
00:54:16,690 --> 00:54:19,970
And where is the antenna in it?

1074
00:54:19,970 --> 00:54:22,240
But then the
ambassador at the time

1075
00:54:22,240 --> 00:54:24,070
was not sufficiently paranoid.

1076
00:54:24,070 --> 00:54:29,906
And it was, in fact,
installed at the embassy

1077
00:54:29,906 --> 00:54:33,044
and certainly some
harm came of it.

1078
00:54:33,044 --> 00:54:36,740


1079
00:54:36,740 --> 00:54:37,430
That's one.

1080
00:54:37,430 --> 00:54:41,540


1081
00:54:41,540 --> 00:54:45,200
As time went on, during
the '50s and '60s,

1082
00:54:45,200 --> 00:54:50,600
small computers began to
be used for cryptography

1083
00:54:50,600 --> 00:54:54,830
instead of these kinds
of mechanical monsters.

1084
00:54:54,830 --> 00:54:59,180
And the kinds of errors
that used to occur gradually

1085
00:54:59,180 --> 00:55:05,390
changed from unwise
operator actions, unwise

1086
00:55:05,390 --> 00:55:10,820
choice of key material
to machine failure.

1087
00:55:10,820 --> 00:55:13,740
And machine failure
became quite important.

1088
00:55:13,740 --> 00:55:17,335
I need to keep track of time.

1089
00:55:17,335 --> 00:55:18,814
10:15.

1090
00:55:18,814 --> 00:55:20,160
10:15.

1091
00:55:20,160 --> 00:55:20,660
OK.

1092
00:55:20,660 --> 00:55:25,470


1093
00:55:25,470 --> 00:55:29,400
So that gets me to a rule
which became very important

1094
00:55:29,400 --> 00:55:31,230
during the '50s and '60s.

1095
00:55:31,230 --> 00:55:33,390
The first thing to do
with a message that

1096
00:55:33,390 --> 00:55:35,830
comes over one of these
important channels

1097
00:55:35,830 --> 00:55:40,470
is check for plain text because
one of the most common error

1098
00:55:40,470 --> 00:55:42,870
modes of the kind of
equipment being used

1099
00:55:42,870 --> 00:55:47,187
was, in fact, that it would
mistakenly send plain text out

1100
00:55:47,187 --> 00:55:48,270
instead of encrypted text.

1101
00:55:48,270 --> 00:55:55,700


1102
00:55:55,700 --> 00:55:59,330
When the typewriters at
the US embassy in Moscow

1103
00:55:59,330 --> 00:56:01,790
broke and needed to
be repaired, they

1104
00:56:01,790 --> 00:56:08,870
were sent out to a local Soviet
typewriter repair company.

1105
00:56:08,870 --> 00:56:11,630
Do I have to go on with
the rest of the story?

1106
00:56:11,630 --> 00:56:13,970
Not really.

1107
00:56:13,970 --> 00:56:17,630
And what you expecting would
happen in fact did happen.

1108
00:56:17,630 --> 00:56:20,510
And it was important.

1109
00:56:20,510 --> 00:56:21,770
Where was the microphone?

1110
00:56:21,770 --> 00:56:23,600
Well, the microphone
was where it

1111
00:56:23,600 --> 00:56:26,660
needed to be to detect
what was happening

1112
00:56:26,660 --> 00:56:28,280
when the typewriter typed.

1113
00:56:28,280 --> 00:56:32,570
And the message being sent out--

1114
00:56:32,570 --> 00:56:35,450
if you listen for it, all
you'd hear would be some

1115
00:56:35,450 --> 00:56:37,550
occasional clicks, about
one click a second.

1116
00:56:37,550 --> 00:56:40,250


1117
00:56:40,250 --> 00:56:42,240
But there's nothing
unusual about that.

1118
00:56:42,240 --> 00:56:44,930
If you look at the power
line in this building,

1119
00:56:44,930 --> 00:56:48,710
looking for noise on it, you
would find zillions of places

1120
00:56:48,710 --> 00:56:53,180
where there were clicks and
various kinds of signals

1121
00:56:53,180 --> 00:56:54,960
going into the power.

1122
00:56:54,960 --> 00:56:56,570
And in fact, in the
average kitchen--

1123
00:56:56,570 --> 00:56:58,760
what with microwave
ovens, toasters ovens,

1124
00:56:58,760 --> 00:57:00,890
and things of that sort--

1125
00:57:00,890 --> 00:57:05,000
various kinds of
noise and clicks

1126
00:57:05,000 --> 00:57:06,320
occur all over the place.

1127
00:57:06,320 --> 00:57:08,990


1128
00:57:08,990 --> 00:57:11,720
But I have a problem for you.

1129
00:57:11,720 --> 00:57:14,810
In the case that
actually happened,

1130
00:57:14,810 --> 00:57:17,390
all that was noticed
when people bore in on it

1131
00:57:17,390 --> 00:57:21,170
was that the typewriter,
the repaired typewriter

1132
00:57:21,170 --> 00:57:25,410
was sending out what appeared to
be a click about once a second.

1133
00:57:25,410 --> 00:57:27,920
That's all, not very suspicious.

1134
00:57:27,920 --> 00:57:30,490


1135
00:57:30,490 --> 00:57:33,980
The clicks appeared to
be nearly identical.

1136
00:57:33,980 --> 00:57:36,160
The timing between
the clicks appeared

1137
00:57:36,160 --> 00:57:38,380
to be very close to one second.

1138
00:57:38,380 --> 00:57:40,900


1139
00:57:40,900 --> 00:57:44,380
As far as one can determine,
the clicks were the same.

1140
00:57:44,380 --> 00:57:47,620
And they were sent
once per second.

1141
00:57:47,620 --> 00:57:53,150
I have a problem for
you, a homework exercise.

1142
00:57:53,150 --> 00:57:55,700
Putting reasonable
assumptions on all this,

1143
00:57:55,700 --> 00:58:00,230
how much information could be
sent with one click per second

1144
00:58:00,230 --> 00:58:03,050
where the clicks pretty
much looked all the same

1145
00:58:03,050 --> 00:58:06,500
and they were pretty much
all one second apart?

1146
00:58:06,500 --> 00:58:10,220
If you look at the
data rate and work out

1147
00:58:10,220 --> 00:58:12,950
with just some reasonable
assumptions about this,

1148
00:58:12,950 --> 00:58:16,610
you might be surprised at
what the data rate could be.

1149
00:58:16,610 --> 00:58:19,370
The question is, would it
keep up with the typewriter?

1150
00:58:19,370 --> 00:58:21,020
I'll tell you the
answer to that,

1151
00:58:21,020 --> 00:58:25,370
yes, with essentially
identical clicks

1152
00:58:25,370 --> 00:58:27,860
being sent out essentially
once per second.

1153
00:58:27,860 --> 00:58:32,190


1154
00:58:32,190 --> 00:58:33,970
There's lots of other
things like that.

1155
00:58:33,970 --> 00:58:37,160
Both us, and our allies,
and our opponents

1156
00:58:37,160 --> 00:58:41,960
have many times sent out
communication equipment, even

1157
00:58:41,960 --> 00:58:46,040
cryptographic equipment, to
basically the local Radio

1158
00:58:46,040 --> 00:58:48,170
Shack to get them repaired.

1159
00:58:48,170 --> 00:58:50,495
When they come back,
they're nearly the same

1160
00:58:50,495 --> 00:58:51,620
as when they were sent out.

1161
00:58:51,620 --> 00:58:54,960


1162
00:58:54,960 --> 00:58:55,460
OK.

1163
00:58:55,460 --> 00:58:57,080
Again, we're getting
into my paranoia.

1164
00:58:57,080 --> 00:59:00,500
But I think I'm putting
a few underpinnings

1165
00:59:00,500 --> 00:59:04,790
under the paranoia to convince
you that this kind of thing

1166
00:59:04,790 --> 00:59:05,390
does happen.

1167
00:59:05,390 --> 00:59:06,110
And that's real.

1168
00:59:06,110 --> 00:59:09,660


1169
00:59:09,660 --> 00:59:12,930
I'll tell a story
on the other side.

1170
00:59:12,930 --> 00:59:23,430
The Russians-- no, Soviets, had
some equipment that sent out

1171
00:59:23,430 --> 00:59:28,320
encrypted traffic by FM radio.

1172
00:59:28,320 --> 00:59:33,810
There also happened to be a very
small modulation of this signal

1173
00:59:33,810 --> 00:59:35,670
AM.

1174
00:59:35,670 --> 00:59:37,490
Though the main
signal itself was FM,

1175
00:59:37,490 --> 00:59:40,920
there was a tiny
bit, 30 decibels down

1176
00:59:40,920 --> 00:59:44,820
from the real signal,
was an AM signal

1177
00:59:44,820 --> 00:59:48,660
that everybody viewed
was too weak to hear.

1178
00:59:48,660 --> 00:59:50,010
OK, fine.

1179
00:59:50,010 --> 00:59:54,030
The CIA rented space in the
first floor of the building--

1180
00:59:54,030 --> 00:59:57,570
this is in Vienna in the '50s--

1181
00:59:57,570 --> 01:00:04,940
and listened to the
low, very dim AM

1182
01:00:04,940 --> 01:00:07,390
signal, which they could do
because they were very close.

1183
01:00:07,390 --> 01:00:09,100
They just basically
set up a shop

1184
01:00:09,100 --> 01:00:12,145
to sell lederhosen or
something like that.

1185
01:00:12,145 --> 01:00:14,270
In fact, they were quite
embarrassed that the store

1186
01:00:14,270 --> 01:00:19,480
they set up, which was really
just a fake CIA store--

1187
01:00:19,480 --> 01:00:21,850
was selling sufficiently
high quality goods

1188
01:00:21,850 --> 01:00:23,380
that they had too
many customers.

1189
01:00:23,380 --> 01:00:27,070


1190
01:00:27,070 --> 01:00:29,860
That went on for
over five years when

1191
01:00:29,860 --> 01:00:34,480
we were, in fact, intercepting
traffic from the principle

1192
01:00:34,480 --> 01:00:39,430
Soviet officer in Vienna
back to headquarters

1193
01:00:39,430 --> 01:00:48,130
from this lederhosen store on
the first floor of Austria.

1194
01:00:48,130 --> 01:00:51,340
Again, OK, so there's
microphones everywhere.

1195
01:00:51,340 --> 01:00:52,960
That was the one.

1196
01:00:52,960 --> 01:00:59,060


1197
01:00:59,060 --> 01:01:02,180
But even in cryptography,
one doesn't necessarily

1198
01:01:02,180 --> 01:01:03,470
have to do poorly.

1199
01:01:03,470 --> 01:01:07,100
I'm going to finish up on
the bright side of things.

1200
01:01:07,100 --> 01:01:11,000
Suppose I have a
cryptographic setup.

1201
01:01:11,000 --> 01:01:18,190


1202
01:01:18,190 --> 01:01:21,460
I have a box
attached to a person,

1203
01:01:21,460 --> 01:01:25,960
and then a communication
line someplace,

1204
01:01:25,960 --> 01:01:29,510
going to someplace else
also attached to the person.

1205
01:01:29,510 --> 01:01:31,660
This is a long
distance line that

1206
01:01:31,660 --> 01:01:35,120
causes these two cryptographic
devices to communicate.

1207
01:01:35,120 --> 01:01:37,458
So somebody puts something
in, in clear text.

1208
01:01:37,458 --> 01:01:39,000
The machine changes
it to ciphertext,

1209
01:01:39,000 --> 01:01:40,840
sends it over the circuit.

1210
01:01:40,840 --> 01:01:44,290
It's decrypted here,
and then the plain text.

1211
01:01:44,290 --> 01:01:48,800
And one can suppose that
this line is tapped probably

1212
01:01:48,800 --> 01:01:50,010
more than once.

1213
01:01:50,010 --> 01:01:51,710
And the tap may
be either passive,

1214
01:01:51,710 --> 01:01:54,230
in the sense that it just
listens to the traffic going

1215
01:01:54,230 --> 01:01:57,410
over it, or it may be
active, in that it changes

1216
01:01:57,410 --> 01:02:01,210
the traffic or inserts its own.

1217
01:02:01,210 --> 01:02:02,530
That's what I want.

1218
01:02:02,530 --> 01:02:04,810
I want secrecy of communication.

1219
01:02:04,810 --> 01:02:07,960
I want bad guys not to
be able to find out what

1220
01:02:07,960 --> 01:02:09,310
this message actually says.

1221
01:02:09,310 --> 01:02:11,620
I want them to get
only the cipher text.

1222
01:02:11,620 --> 01:02:14,240


1223
01:02:14,240 --> 01:02:16,710
Is it possible to
design a system--

1224
01:02:16,710 --> 01:02:21,350
and this is another homework
exercise for you, too.

1225
01:02:21,350 --> 01:02:25,250


1226
01:02:25,250 --> 01:02:29,480
Is it possible to design
a system of this sort that

1227
01:02:29,480 --> 01:02:33,950
protects the secrecy of the
information in such a way

1228
01:02:33,950 --> 01:02:38,660
that the enemy, in the form
of a cleaning lady or janitor,

1229
01:02:38,660 --> 01:02:42,380
can get all the information
out of one of these boxes,

1230
01:02:42,380 --> 01:02:45,740
like this box, every
piece of information?

1231
01:02:45,740 --> 01:02:48,440
And I mean that in the broadest
possible sense-- all data,

1232
01:02:48,440 --> 01:02:52,740
all programs, all keys, anything
that's in that machine--

1233
01:02:52,740 --> 01:02:57,680
but still not be able to later
read the encrypted traffic that

1234
01:02:57,680 --> 01:02:59,180
goes across the line.

1235
01:02:59,180 --> 01:03:02,350
Is it possible to
design such a system?

1236
01:03:02,350 --> 01:03:04,100
It's a difficult
question because you have

1237
01:03:04,100 --> 01:03:05,392
to think about a lot of things.

1238
01:03:05,392 --> 01:03:07,340
But again, I'll
tell you the answer.

1239
01:03:07,340 --> 01:03:09,330
The answer is, yes,
it is possible.

1240
01:03:09,330 --> 01:03:12,020
And in fact, it is
done on some systems

1241
01:03:12,020 --> 01:03:14,480
that I know about and use.

1242
01:03:14,480 --> 01:03:17,000
But still, it's a
difficult exercise

1243
01:03:17,000 --> 01:03:19,770
to work out just
how do you do that.

1244
01:03:19,770 --> 01:03:21,280
But it can be done.

1245
01:03:21,280 --> 01:03:22,650
Sir?

1246
01:03:22,650 --> 01:03:27,830
[INAUDIBLE] they
then been able to--

1247
01:03:27,830 --> 01:03:28,720
[INAUDIBLE]

1248
01:03:28,720 --> 01:03:32,010
[INAUDIBLE]

1249
01:03:32,010 --> 01:03:34,830
Yeah, if both cleaning
ladies were there

1250
01:03:34,830 --> 01:03:37,260
and the cleaning ladies
could communicate,

1251
01:03:37,260 --> 01:03:39,900
then the system is broken.

1252
01:03:39,900 --> 01:03:42,460
I mean this only
goes so far it goes.

1253
01:03:42,460 --> 01:03:45,960
But from on of the
two machines, you

1254
01:03:45,960 --> 01:03:48,200
can get all the
information in the broadest

1255
01:03:48,200 --> 01:03:49,560
sense that's in them.

1256
01:03:49,560 --> 01:03:54,720
But if you get two, then you
can later decrypt any traffic

1257
01:03:54,720 --> 01:03:59,070
between those two stations,
but not between one of them

1258
01:03:59,070 --> 01:04:02,970
and some third station someplace
that wasn't broken into.

1259
01:04:02,970 --> 01:04:05,700
So that's about as
far as you can go.

1260
01:04:05,700 --> 01:04:12,740


1261
01:04:12,740 --> 01:04:15,830
I'm going to quite
while I'm ahead.

1262
01:04:15,830 --> 01:04:18,080
Questions, comments, whatever.

1263
01:04:18,080 --> 01:04:19,220
Speak.

1264
01:04:19,220 --> 01:04:21,098
Sir?

1265
01:04:21,098 --> 01:04:25,066
[APPLAUSE]

1266
01:04:25,066 --> 01:04:31,030


1267
01:04:31,030 --> 01:04:32,100
Speak.

1268
01:04:32,100 --> 01:04:34,360
[INAUDIBLE] about the
protection of information

1269
01:04:34,360 --> 01:04:37,210
that is protected
from knowing it

1270
01:04:37,210 --> 01:04:41,260
exists-- that is,
you go to lengths

1271
01:04:41,260 --> 01:04:44,650
to ensure that no
one knows if that's

1272
01:04:44,650 --> 01:04:46,630
that's the right
telephone to tap.

1273
01:04:46,630 --> 01:04:49,450
How successful or
problematic would that be?

1274
01:04:49,450 --> 01:04:50,440
It is done.

1275
01:04:50,440 --> 01:04:52,120
It is of some importance.

1276
01:04:52,120 --> 01:04:55,210
And it is done.

1277
01:04:55,210 --> 01:04:58,630
For example, in the intelligence
business, we have sensors--

1278
01:04:58,630 --> 01:05:00,820
these are gatherers of
information-- in some

1279
01:05:00,820 --> 01:05:03,310
of the goddamnedest place.

1280
01:05:03,310 --> 01:05:05,800
How do you send that
information out?

1281
01:05:05,800 --> 01:05:08,410
A priori, you don't.

1282
01:05:08,410 --> 01:05:12,140
You have to work out
where you can send it out.

1283
01:05:12,140 --> 01:05:17,650
What good does it do to tap
a fiber optic cable going

1284
01:05:17,650 --> 01:05:22,720
across Siberia since you have
no conceivable way of getting

1285
01:05:22,720 --> 01:05:25,620
that rate of information out?

1286
01:05:25,620 --> 01:05:27,120
You've got to
somehow conceal it.

1287
01:05:27,120 --> 01:05:28,830
Yes, that is an important issue.

1288
01:05:28,830 --> 01:05:31,050
And people do worry about that.

1289
01:05:31,050 --> 01:05:32,940
And it's not been well-solved.

1290
01:05:32,940 --> 01:05:35,610
But there are partial
solutions to the problem.

1291
01:05:35,610 --> 01:05:38,450
Fair?

1292
01:05:38,450 --> 01:05:39,840
yes? ?

1293
01:05:39,840 --> 01:05:41,970
Suppose that some big
financial institution

1294
01:05:41,970 --> 01:05:44,850
hires you to be an
advisor and asks you,

1295
01:05:44,850 --> 01:05:49,410
what do you think about
the risks of key escrow

1296
01:05:49,410 --> 01:05:50,390
key recovery systems?

1297
01:05:50,390 --> 01:05:54,510
Should we let our keys be
escrowed by a third party?

1298
01:05:54,510 --> 01:05:55,620
How would you advise that?

1299
01:05:55,620 --> 01:05:57,720
Do you trust your
government, sir?

1300
01:05:57,720 --> 01:05:58,710
No.

1301
01:05:58,710 --> 01:06:02,268
You and I don't have
to pursue it then.

1302
01:06:02,268 --> 01:06:04,440
I asked your opinion.

1303
01:06:04,440 --> 01:06:06,690
Well, my opinion is
I would first ask,

1304
01:06:06,690 --> 01:06:09,690
do you trust your government?

1305
01:06:09,690 --> 01:06:13,710
And the answer to
that would come out

1306
01:06:13,710 --> 01:06:16,530
mostly somewhat negative--

1307
01:06:16,530 --> 01:06:19,830
no, or not much, or
whatever like that.

1308
01:06:19,830 --> 01:06:21,130
Do I trust my government?

1309
01:06:21,130 --> 01:06:24,420
I don't feel much differently
from the way you feel.

1310
01:06:24,420 --> 01:06:28,110
I mean, there's good
guys there and bad guys.

1311
01:06:28,110 --> 01:06:31,290
I won't characterize myself as
being either side of that one.

1312
01:06:31,290 --> 01:06:38,290


1313
01:06:38,290 --> 01:06:41,740
But I can understand
the FBI's position

1314
01:06:41,740 --> 01:06:45,160
in pushing for a
key escrow scheme.

1315
01:06:45,160 --> 01:06:50,830
What do you do to counter
the communications

1316
01:06:50,830 --> 01:06:58,040
abilities of the druggies
or potential terrorists?

1317
01:06:58,040 --> 01:06:59,290
Yeah, it's an important issue.

1318
01:06:59,290 --> 01:07:04,810


1319
01:07:04,810 --> 01:07:06,160
Sir?

1320
01:07:06,160 --> 01:07:09,136
You mentioned that
there were [INAUDIBLE]

1321
01:07:09,136 --> 01:07:10,930
compromised the
financial security

1322
01:07:10,930 --> 01:07:15,010
of financial institutions which
dwarf the Citibank episode

1323
01:07:15,010 --> 01:07:17,783
but that were largely unknown.

1324
01:07:17,783 --> 01:07:19,755
Could you give us a
couple of examples?

1325
01:07:19,755 --> 01:07:21,730
[LAUGHTER]

1326
01:07:21,730 --> 01:07:25,590
Hey, suppose I want to
rob a bank between now

1327
01:07:25,590 --> 01:07:29,560
and the end of the evening.

1328
01:07:29,560 --> 01:07:31,420
I don't walk into
a bank with a gun.

1329
01:07:31,420 --> 01:07:33,160
That's dumb.

1330
01:07:33,160 --> 01:07:35,410
Because all I'm going
to do, if I successfully

1331
01:07:35,410 --> 01:07:38,320
rob the bank is get enough money
for a pleasant weekend in Las

1332
01:07:38,320 --> 01:07:39,760
Vegas.

1333
01:07:39,760 --> 01:07:42,110
If I start walking into
the bank with a gun,

1334
01:07:42,110 --> 01:07:45,970
I'll spend an unpleasant weekend
in the county jail followed

1335
01:07:45,970 --> 01:07:50,290
by an unpleasant couple of
years in the state prison.

1336
01:07:50,290 --> 01:07:53,370
Go for the ATMs.

1337
01:07:53,370 --> 01:07:57,280
Consider, when I was
last in Norway, indeed,

1338
01:07:57,280 --> 01:08:01,440
I have a card which I inserted
into an ATM in Norway.

1339
01:08:01,440 --> 01:08:09,480
And 1,000 kroner, about $200,
came out of the machine.

1340
01:08:09,480 --> 01:08:12,240
What actually took place
in that transaction?

1341
01:08:12,240 --> 01:08:14,060
There are a number
of parties to this.

1342
01:08:14,060 --> 01:08:14,910
There's me.

1343
01:08:14,910 --> 01:08:18,790
I don't want my pin spread
all across the country,

1344
01:08:18,790 --> 01:08:21,930
particularly when it's
combined with bank

1345
01:08:21,930 --> 01:08:28,040
identification and the
account number at that bank.

1346
01:08:28,040 --> 01:08:30,920
That combined with a pin
let's anybody get my money.

1347
01:08:30,920 --> 01:08:32,300
So I don't want that.

1348
01:08:32,300 --> 01:08:37,010
The bank that owns
the ATM in Norway,

1349
01:08:37,010 --> 01:08:39,890
before it gets me
the 1,000 kroner,

1350
01:08:39,890 --> 01:08:43,160
wants to make sure that
it will get the money back

1351
01:08:43,160 --> 01:08:49,880
from the local bank in
northern New Hampshire, out

1352
01:08:49,880 --> 01:08:53,180
in the boondocks of
northern New Hampshire.

1353
01:08:53,180 --> 01:08:54,710
And the bank in
New Hampshire wants

1354
01:08:54,710 --> 01:09:01,569
to make sure that you're talking
about a real customer that

1355
01:09:01,569 --> 01:09:03,609
had some money in the
bank and hasn't withdrawn

1356
01:09:03,609 --> 01:09:05,350
at all in the last day.

1357
01:09:05,350 --> 01:09:10,899
Now, there are both high tech
and low tech ways of extracting

1358
01:09:10,899 --> 01:09:12,500
the money from an ATM.

1359
01:09:12,500 --> 01:09:16,359
The most common method
in the UK and Europe

1360
01:09:16,359 --> 01:09:20,050
is to drive a forklift
truck up to the ATM

1361
01:09:20,050 --> 01:09:23,200
and pull it out of the
wall and take it home

1362
01:09:23,200 --> 01:09:25,657
where you can work on it in
the comfort of your living

1363
01:09:25,657 --> 01:09:26,740
room with a cutting torch.

1364
01:09:26,740 --> 01:09:29,590


1365
01:09:29,590 --> 01:09:33,700
There are more
sophisticated approaches.

1366
01:09:33,700 --> 01:09:36,850
The most common in the
US, also fairly low tech,

1367
01:09:36,850 --> 01:09:38,859
is to go and buy an ATM--

1368
01:09:38,859 --> 01:09:41,950
and I assure you that can be
bought rather readily and not

1369
01:09:41,950 --> 01:09:43,630
terribly expensive--

1370
01:09:43,630 --> 01:09:48,430
go and install it in the
mall down the street.

1371
01:09:48,430 --> 01:09:51,354
When somebody comes up
and puts a pin in into it,

1372
01:09:51,354 --> 01:09:52,854
you pass a message
that says, sorry,

1373
01:09:52,854 --> 01:09:54,370
the machine is out of order.

1374
01:09:54,370 --> 01:09:55,150
Try again later.

1375
01:09:55,150 --> 01:09:58,240
But you've done is collected a
whole bunch of pins and account

1376
01:09:58,240 --> 01:09:59,950
numbers.

1377
01:09:59,950 --> 01:10:01,960
Not very high tech,
but very profitable.

1378
01:10:01,960 --> 01:10:03,460
It turns out that,
in the weekend,

1379
01:10:03,460 --> 01:10:05,610
it's fairly easy to
come up with, say,

1380
01:10:05,610 --> 01:10:07,760
$100,000 by this technique.

1381
01:10:07,760 --> 01:10:09,760
That has been done
many times over.

1382
01:10:09,760 --> 01:10:12,700


1383
01:10:12,700 --> 01:10:15,370
I know it's only a partial
answer to your question.

1384
01:10:15,370 --> 01:10:17,170
But it is, at least
in part, an answer.

1385
01:10:17,170 --> 01:10:18,265
That does happen.

1386
01:10:18,265 --> 01:10:19,380
It's fairly common.

1387
01:10:19,380 --> 01:10:21,280
It's almost daily in the US.

1388
01:10:21,280 --> 01:10:22,430
Sir?

1389
01:10:22,430 --> 01:10:25,720
Do you think that a centralized
financial system based

1390
01:10:25,720 --> 01:10:28,660
on book entry settlement
is more vulnerable--

1391
01:10:28,660 --> 01:10:29,590
Based on what?

1392
01:10:29,590 --> 01:10:31,330
On book entry settlement.

1393
01:10:31,330 --> 01:10:34,885
That is [INAUDIBLE].

1394
01:10:34,885 --> 01:10:37,450
No, it takes too long.

1395
01:10:37,450 --> 01:10:39,100
[INAUDIBLE]

1396
01:10:39,100 --> 01:10:41,740
[INAUDIBLE] the
Federal Reserve System

1397
01:10:41,740 --> 01:10:45,370
for something on the
order of $100,000 a day.

1398
01:10:45,370 --> 01:10:47,380
And they won't notice
for about a month.

1399
01:10:47,380 --> 01:10:48,280
Oh, I'm sorry.

1400
01:10:48,280 --> 01:10:49,720
Let me rephrase the question.

1401
01:10:49,720 --> 01:10:51,910
Do you think that a
centralized system

1402
01:10:51,910 --> 01:10:55,000
with a central
clearinghouse is more

1403
01:10:55,000 --> 01:10:57,700
insecure than a
decentralized system

1404
01:10:57,700 --> 01:11:00,790
where people are making
peer to peer payments?

1405
01:11:00,790 --> 01:11:02,686
Yes, I do.

1406
01:11:02,686 --> 01:11:07,570
And it has, in practice, not
the worked out very well.

1407
01:11:07,570 --> 01:11:08,230
Yeah.

1408
01:11:08,230 --> 01:11:09,320
Sir?

1409
01:11:09,320 --> 01:11:13,105
Do you think that the
gap between best military

1410
01:11:13,105 --> 01:11:14,771
intelligence practice
and cryptography

1411
01:11:14,771 --> 01:11:19,700
and cryptologists that
publish academic commercial

1412
01:11:19,700 --> 01:11:22,532
cryptography and cryptoanalysis
is widening or narrowing

1413
01:11:22,532 --> 01:11:25,260
over time?

1414
01:11:25,260 --> 01:11:27,860
I don't notice that it's
gotten wider narrower.

1415
01:11:27,860 --> 01:11:30,730
Actually, compared to when I
first knew about these things,

1416
01:11:30,730 --> 01:11:33,080
the gap has gotten
somewhat narrower.

1417
01:11:33,080 --> 01:11:36,260
We mostly used one-time
pads in the '50s,

1418
01:11:36,260 --> 01:11:39,350
went to machines in the late
'50s, and on into the '60s

1419
01:11:39,350 --> 01:11:41,190
and beyond.

1420
01:11:41,190 --> 01:11:43,460
So I don't see a gap
that's widened there.

1421
01:11:43,460 --> 01:11:45,440
I see a gap that's shrunk a bit.

1422
01:11:45,440 --> 01:11:47,510
Yet, the gap between
them is there.

1423
01:11:47,510 --> 01:11:49,940
And it's big.

1424
01:11:49,940 --> 01:11:57,110
My notion is that the encryption
systems of the US and allied

1425
01:11:57,110 --> 01:11:59,420
countries is pretty darn--

1426
01:11:59,420 --> 01:12:02,870
no, pretty good is
one thing to say.

1427
01:12:02,870 --> 01:12:09,950


1428
01:12:09,950 --> 01:12:12,454
Adequate for the information
they're protecting.

1429
01:12:12,454 --> 01:12:21,460


1430
01:12:21,460 --> 01:12:24,431
What are the factors that
affect the size of that gap?

1431
01:12:24,431 --> 01:12:27,868
Do you foresee a change in them?

1432
01:12:27,868 --> 01:12:41,622


1433
01:12:41,622 --> 01:12:44,600
Nah.

1434
01:12:44,600 --> 01:12:45,820
Yeah, go ahead.

1435
01:12:45,820 --> 01:12:49,260


1436
01:12:49,260 --> 01:12:53,830
Suppose the key escrow
system does get into place.

1437
01:12:53,830 --> 01:12:56,800
A key escrow database with
a few billion keys in it

1438
01:12:56,800 --> 01:13:00,520
is a few cubic inches
on proper storage media.

1439
01:13:00,520 --> 01:13:05,740
Could you estimate the dollar
value of that information?

1440
01:13:05,740 --> 01:13:07,090
Probably very large.

1441
01:13:07,090 --> 01:13:10,330


1442
01:13:10,330 --> 01:13:13,340
Well, it depends--
dollar value to who.

1443
01:13:13,340 --> 01:13:15,820
I mean, how much would
it cost me to find out

1444
01:13:15,820 --> 01:13:18,820
your social security number,
credit card number for all

1445
01:13:18,820 --> 01:13:22,150
your cards, bank account number,
and how well your payment

1446
01:13:22,150 --> 01:13:22,990
record is?

1447
01:13:22,990 --> 01:13:26,200
How much do you think that
would cost me to find out?

1448
01:13:26,200 --> 01:13:27,520
Pardon?

1449
01:13:27,520 --> 01:13:28,450
$20.

1450
01:13:28,450 --> 01:13:31,210
Yeah, they got it.

1451
01:13:31,210 --> 01:13:33,010
It doesn't matter,
so long as they were

1452
01:13:33,010 --> 01:13:35,440
able to say the number in it.

1453
01:13:35,440 --> 01:13:37,640
I mean, if they said
$20, that's fine.

1454
01:13:37,640 --> 01:13:40,000
If they said $200,
that's fine, too.

1455
01:13:40,000 --> 01:13:42,160
It actually is
somewhere in that range.

1456
01:13:42,160 --> 01:13:46,480
To get what I asked for probably
costs something like $200.

1457
01:13:46,480 --> 01:13:48,105
But the person that
got it for them,

1458
01:13:48,105 --> 01:13:49,480
it doesn't cost
that much at all.

1459
01:13:49,480 --> 01:13:50,550
It only costs $5.

1460
01:13:50,550 --> 01:13:53,200


1461
01:13:53,200 --> 01:13:55,330
So value to who, right?

1462
01:13:55,330 --> 01:13:59,170


1463
01:13:59,170 --> 01:14:00,904
[INAUDIBLE]

1464
01:14:00,904 --> 01:14:05,190


1465
01:14:05,190 --> 01:14:07,070
Well, I have two answers.

1466
01:14:07,070 --> 01:14:08,000
One is not much.

1467
01:14:08,000 --> 01:14:12,080
The other is I've delegated all
responsibility for that issue

1468
01:14:12,080 --> 01:14:15,500
to the vice president.

1469
01:14:15,500 --> 01:14:24,180
[APPLAUSE]

1470
01:14:24,180 --> 01:14:31,135
[INAUDIBLE]

1471
01:14:31,135 --> 01:14:31,760
Yeah, OK, what?

1472
01:14:31,760 --> 01:14:35,510


1473
01:14:35,510 --> 01:14:37,490
OK.

1474
01:14:37,490 --> 01:14:40,670
Given the NSA's alleged
budget and the fact

1475
01:14:40,670 --> 01:14:41,720
that you seem to say--

1476
01:14:41,720 --> 01:14:42,970
Does anybody obey this budget?

1477
01:14:42,970 --> 01:14:46,030


1478
01:14:46,030 --> 01:14:47,736
[INAUDIBLE]

1479
01:14:47,736 --> 01:14:49,410
And they even could be wrong.

1480
01:14:49,410 --> 01:14:53,510
And given the fact that you say
that military and government is

1481
01:14:53,510 --> 01:14:56,020
so far ahead, why
make such a big deal

1482
01:14:56,020 --> 01:14:58,460
about key escrow and all that?

1483
01:14:58,460 --> 01:15:01,160
Is it possible that current
cryptography is actually

1484
01:15:01,160 --> 01:15:03,510
of some value and is
actually difficult to break

1485
01:15:03,510 --> 01:15:04,260
by the government?

1486
01:15:04,260 --> 01:15:06,540
It could be linked, so yes.

1487
01:15:06,540 --> 01:15:09,185
It would not be
difficult to construct

1488
01:15:09,185 --> 01:15:13,100
from things that
are widely known

1489
01:15:13,100 --> 01:15:19,730
to commercial people in the
US to produce an encryption

1490
01:15:19,730 --> 01:15:22,853
system that is quite
expensive to break.

1491
01:15:22,853 --> 01:15:24,270
Now, I'm not saying
hard to break.

1492
01:15:24,270 --> 01:15:26,180
I'm saying expensive to break.

1493
01:15:26,180 --> 01:15:34,670
And overseas-- well, there might
be one or two Israeli citizens

1494
01:15:34,670 --> 01:15:37,130
in the audience.

1495
01:15:37,130 --> 01:15:40,040
They are better than us.

1496
01:15:40,040 --> 01:15:44,720
So our expectation is if we see
an encryption system coming out

1497
01:15:44,720 --> 01:15:48,350
of Israel, we say things--
well, I don't want

1498
01:15:48,350 --> 01:15:50,360
to say it in Aramaic or Hebrew.

1499
01:15:50,360 --> 01:15:55,310
But that is quite
worrisome because without--

1500
01:15:55,310 --> 01:15:58,950
well, using the expertise
that they have in hand,

1501
01:15:58,950 --> 01:16:01,040
they could come up with
a system that would be

1502
01:16:01,040 --> 01:16:03,230
very expensive for us to break.

1503
01:16:03,230 --> 01:16:08,150


1504
01:16:08,150 --> 01:16:11,102
That it?

1505
01:16:11,102 --> 01:16:12,086
Over there.

1506
01:16:12,086 --> 01:16:16,770


1507
01:16:16,770 --> 01:16:25,440
[INAUDIBLE] All right, do you
think that that enough of these

1508
01:16:25,440 --> 01:16:29,580
authentication issues, like the
bank machine and the internet

1509
01:16:29,580 --> 01:16:33,492
purchases, have happened that
people are becoming aware--

1510
01:16:33,492 --> 01:16:34,950
people running
these businesses are

1511
01:16:34,950 --> 01:16:36,540
becoming aware of these issues?

1512
01:16:36,540 --> 01:16:42,270
Or are they continuing to just
blindly stumble along and not

1513
01:16:42,270 --> 01:16:44,160
pay much attention to--

1514
01:16:44,160 --> 01:16:47,660
I have not detected any
awareness of ambition.

1515
01:16:47,660 --> 01:16:48,883
No, I'm serious.

1516
01:16:48,883 --> 01:16:51,050
And people come up with
encryption methods which are

1517
01:16:51,050 --> 01:16:53,438
plainly inadequate to the job.

1518
01:16:53,438 --> 01:16:54,980
But you've got to
remember that there

1519
01:16:54,980 --> 01:17:00,790
are other kinds
of cultural issues

1520
01:17:00,790 --> 01:17:03,880
that are not unimportant.

1521
01:17:03,880 --> 01:17:09,070
I have a card, which is issued
by the South African Power

1522
01:17:09,070 --> 01:17:10,330
Company.

1523
01:17:10,330 --> 01:17:14,080
It's a card that you take to
the power company's office.

1524
01:17:14,080 --> 01:17:17,020
And they change the
magnetic bits on it

1525
01:17:17,020 --> 01:17:19,660
that, in effect, add to
the amount of money on it.

1526
01:17:19,660 --> 01:17:23,200
The one I have is good
for 1,000 kilowatt hours.

1527
01:17:23,200 --> 01:17:27,730
But it is an organization
where everyone

1528
01:17:27,730 --> 01:17:32,312
associated with the power
company is dishonest.

1529
01:17:32,312 --> 01:17:35,800


1530
01:17:35,800 --> 01:17:38,800
In South Africa,
normally a customer

1531
01:17:38,800 --> 01:17:42,110
does not pay the electric bill.

1532
01:17:42,110 --> 01:17:42,740
Why not?

1533
01:17:42,740 --> 01:17:45,200
Well, why pay it?

1534
01:17:45,200 --> 01:17:46,790
Do you think that
the power company

1535
01:17:46,790 --> 01:17:48,920
is going to send someone
out to disconnect

1536
01:17:48,920 --> 01:17:51,410
somebody's power in one of
the townships in South Africa?

1537
01:17:51,410 --> 01:17:52,010
Not likely.

1538
01:17:52,010 --> 01:17:53,135
You'd never see them again.

1539
01:17:53,135 --> 01:17:58,110


1540
01:17:58,110 --> 01:18:00,920
In the South African
yellow pages,

1541
01:18:00,920 --> 01:18:03,440
you will find a
kind of business,

1542
01:18:03,440 --> 01:18:05,840
people who are in
business, people

1543
01:18:05,840 --> 01:18:09,630
wire around electrical meters.

1544
01:18:09,630 --> 01:18:14,790
And I mean, there are
cultural differences here

1545
01:18:14,790 --> 01:18:18,140
which take the whole
subject matter away

1546
01:18:18,140 --> 01:18:19,760
from what you and I
have been thinking

1547
01:18:19,760 --> 01:18:22,160
about in a very important way.

1548
01:18:22,160 --> 01:18:26,480
And that kind of thing
has to be tracked.

1549
01:18:26,480 --> 01:18:29,930
If you intercept a
message in Arabic

1550
01:18:29,930 --> 01:18:33,485
and it does not begin
by saying God is great,

1551
01:18:33,485 --> 01:18:35,360
then you missed the
beginning of the message.

1552
01:18:35,360 --> 01:18:39,090


1553
01:18:39,090 --> 01:18:43,300
And so we have to
track the whole thing.

1554
01:18:43,300 --> 01:18:46,040
And what you're asking
about is, to some degree,

1555
01:18:46,040 --> 01:18:50,060
a cultural question, much
along the lines of do

1556
01:18:50,060 --> 01:18:52,740
you trust your government.

1557
01:18:52,740 --> 01:18:56,510
The majority answer here
would be not very much.

1558
01:18:56,510 --> 01:19:00,605
Or the South African example--

1559
01:19:00,605 --> 01:19:04,820


1560
01:19:04,820 --> 01:19:06,920
well, I've talked
around your question.

1561
01:19:06,920 --> 01:19:09,530
I guess I didn't answer it.

1562
01:19:09,530 --> 01:19:12,200
No, I don't perceive
any awareness

1563
01:19:12,200 --> 01:19:14,120
on the part of the US.

1564
01:19:14,120 --> 01:19:18,140
That's the direct answer
to the direct question.

1565
01:19:18,140 --> 01:19:19,740
One last question?

1566
01:19:19,740 --> 01:19:22,030
This is a bit more broad.

1567
01:19:22,030 --> 01:19:23,930
Recently, the Cold War is over.

1568
01:19:23,930 --> 01:19:25,580
There seem to be less enemies.

1569
01:19:25,580 --> 01:19:28,340
But also, there's
a lot more enemies

1570
01:19:28,340 --> 01:19:31,860
that have more powerful ways
of encrypting information

1571
01:19:31,860 --> 01:19:33,620
and more powerful
ways of protecting it.

1572
01:19:33,620 --> 01:19:37,100
And information has just
been put into a lot more use.

1573
01:19:37,100 --> 01:19:38,900
There used to be Pravda.

1574
01:19:38,900 --> 01:19:40,730
And now, there's 400
newspapers that we

1575
01:19:40,730 --> 01:19:42,590
have to get out of Russia.

1576
01:19:42,590 --> 01:19:45,200
Is the NSA able to keep up?

1577
01:19:45,200 --> 01:19:46,095
It costs.

1578
01:19:46,095 --> 01:19:48,200
But that's not a high expense.

1579
01:19:48,200 --> 01:19:51,500
We certainly track all
unclassified publications,

1580
01:19:51,500 --> 01:19:52,700
for sure.

1581
01:19:52,700 --> 01:19:55,340
We read all newspapers.

1582
01:19:55,340 --> 01:20:01,410
And the number has
increased because--

1583
01:20:01,410 --> 01:20:06,920
I mean, the Kazakhstan Times
or the Uzbekistan Journal,

1584
01:20:06,920 --> 01:20:08,520
yeah we have to
track that stuff.

1585
01:20:08,520 --> 01:20:09,870
And it does cost.

1586
01:20:09,870 --> 01:20:11,950
It is not a
particularly high cost.

1587
01:20:11,950 --> 01:20:12,630
But it is done.

1588
01:20:12,630 --> 01:20:14,070
And it does cost money.

1589
01:20:14,070 --> 01:20:17,350
And it's done in the other way,
too, done in both directions.

1590
01:20:17,350 --> 01:20:21,210
But your statement,
the Cold War is over--

1591
01:20:21,210 --> 01:20:26,700
I'd want to-- yeah, I'd want
to modulate that just a bit,

1592
01:20:26,700 --> 01:20:28,668
not much.

1593
01:20:28,668 --> 01:20:29,580
[INAUDIBLE]

1594
01:20:29,580 --> 01:20:31,080
Yeah.

1595
01:20:31,080 --> 01:20:37,440
Well, among all those
little countries

1596
01:20:37,440 --> 01:20:40,560
that you see along the southern
boundary of the former Soviet

1597
01:20:40,560 --> 01:20:45,600
Union, 8 or 10 little countries
nobody's ever heard of,

1598
01:20:45,600 --> 01:20:49,980
One of them,
Kazakhstan, is larger

1599
01:20:49,980 --> 01:20:51,530
than all of Western Europe.

1600
01:20:51,530 --> 01:20:54,420


1601
01:20:54,420 --> 01:20:58,620
So one has to keep some kind of
balance in this kind of thing.

1602
01:20:58,620 --> 01:21:00,762
I don't even know what
the capital it is anymore

1603
01:21:00,762 --> 01:21:02,220
because what used
to be the capital

1604
01:21:02,220 --> 01:21:09,240
five years ago is
politically incorrect now.

1605
01:21:09,240 --> 01:21:11,970
It's been changed
to a Kazak word.

1606
01:21:11,970 --> 01:21:13,440
But then as soon
as they did that,

1607
01:21:13,440 --> 01:21:16,130
they moved the capital
about 50 miles to the north.

1608
01:21:16,130 --> 01:21:18,015
So it's hard to keep track.

1609
01:21:18,015 --> 01:21:21,330
But again, this is
something that does happen.

1610
01:21:21,330 --> 01:21:22,492
And it does cost.

1611
01:21:22,492 --> 01:21:23,700
But it's not a major expense.

1612
01:21:23,700 --> 01:21:25,560
It's something we have to do.

1613
01:21:25,560 --> 01:21:30,030
Well, I'd like to thank you
for a very interesting talk.

1614
01:21:30,030 --> 01:21:33,680
[APPLAUSE]

1615
01:21:33,680 --> 01:21:40,000