Inaugural Lecture - February 19 th 2008 1 From Fish to Phishing Kenny Paterson Information Security Group Mathematics Department Royal Holloway, University of London
Mar 31, 2015
Inaugural Lecture - February 19th 2008 1
From Fish to Phishing
Kenny Paterson
Information Security Group
Mathematics Department
Royal Holloway, University of London
Inaugural Lecture - February 19th 2008 2
CINS/F1-01
Overview
1. What is Cryptography?
2. Fish and Colossus
3. WEP and GSM
4. IPsec
5. Phishing
6. Concluding remarks
Inaugural Lecture - February 19th 2008 3
1. What is Cryptography?
• Historically: making (and breaking) codes and ciphers.– Designed to scramble messages so they cannot be
read by an enemy. – The preserve of emperors and generals.– Archetypes: the Caesar cipher; Kama Sutra code.
• Today: a range of techniques for ensuring the confidentiality, integrity and origin of data.– Mobile phones, chip and pin cards, Internet e-
commerce.– Industrial cryptography.
Inaugural Lecture - February 19th 2008 4
What is Cryptography?
• And a thriving academic discipline involving a blend of mathematics, statistics and computer science. – Advanced encryption, signature, key exchange
primitives.– Secure multi-party computation.– Private information retrieval from databases.– Anonymous handshake protocols.– Electronic elections and auctions.– ….
Inaugural Lecture - February 19th 2008 5
This Talk
• Cryptography is a powerful tool.– Instrumental in increasing security and confidence in
the digital age.
• But cryptography has many limitations.– Human involvement.– Changing adversaries.– Difficulties of key management.– Widening chasm between theory and practice.
• Our aim:– To illustrate some of these problems using a mixture
of historical and current examples.
Inaugural Lecture - February 19th 2008 6
2. Fish and Colossus
• Usual assumption: interceptor knows everything about the system.
• So security depends entirely on the secrecy of the key K.
• Kerckhoffs’ Principle.
CiphertextC
Key K
EncryptionAlgorithm
MessageM
DecryptionAlgorithm
MessageM
Interceptor
Key K
Inaugural Lecture - February 19th 2008 7
Fish
• 1941: Germans begin to build pan-European wireless communications network.– Linking Wehrmacht commands with general staff in Berlin.– Using directional antennae and high-speed, non-Morse signalling for
teleprinter traffic.– Encrypted using Geheimschreiber machine.
• Lorenz SZ40/42 teletype attachment.
– Careful traffic analysis indicated possible high value of traffic.
• Traffic named “Fish” by Bletchley Park staff.– Each link named after a different species: Bream, Codfish,…
• 1942: British start to systematically intercept Fish signals.– And Bletchley Park begins to analyse ciphertext.– But with virtually no information about the encryption method being used!
• Jan-May 1945: British decrypt 22 million characters of Fish traffic. – Without ever having seen a Lorenz machine!
8
,21
Breaking Fish
7
1219
7
12
,21
,8,3
,8
• Initial analysis suggested Fish traffic was being encrypted using a stream cipher.– Message converted into numbers, A=0, B=1,…, Z=25.– Message added character-by-character to keystream.
MessageM
MessageM
Key K
+
KeystreamKeystreamGenerator
Key K
Key K
-
KeystreamGenerator
CiphertextC=K+M mod 26
Encryption Decryption
Inaugural Lecture - February 19th 2008 9
Breaking Fish
• In theory: stream cipher known to be unbreakable if keystream is a truly random sequence of characters.– Shannon (1949): H(M|C)=H(M).– Ciphertext reveals nothing (statistically) about the message.
• In practice: sender and receiver have to generate a pseudo-random keystream using a deterministic algorithm and a short key.– Introducing statistical imperfections exploitable by cryptanalyst…
Inaugural Lecture - February 19th 2008 10
Fishing at a Depth
• Fish message indicators preceding encrypted data were presumed to indicate initial setting of keystream generator.
• Equality of indicators would imply equality of keystreams.– Known as a depth at Bletchley Park.
• So what if a depth occurred for two closely related messages?– Should never be permitted because known to introduce security
weakness.– But operators make mistakes….
• With some inspired guess-work, this could allow the two related messages to be recovered…
Inaugural Lecture - February 19th 2008 11
Fishing at a Depth
K
C1 5 3 20 23 23 7 1 16 14 5 12 3 0 14 12 14 23 6 17
M1
Text1
Text2
M2
C2 5 3 20 23 23 7 3 17 19 10 18 6 22 24 7 14 2 9 21
K
Inaugural Lecture - February 19th 2008 12
Fishing at a Depth
K
C1 5 3 20 23 23 7 1 16 14 5 12 3 0 14 12 14 23 6 17
M1
Text1
Text2
M2
C2 5 3 20 23 23 7 3 17 19 10 18 6 22 24 7 14 2 9 21
K
Inaugural Lecture - February 19th 2008 13
Fishing at a Depth
K
C1 5 3 20 23 23 7 1 16 14 5 12 3 0 14 12 14 23 6 17
M1
Text1 C R Y P T OText2 C R Y P T O
M2
C2 5 3 20 23 23 7 3 17 19 10 18 6 22 24 7 14 2 9 21
K
Inaugural Lecture - February 19th 2008 14
Fishing at a Depth
K
C1 5 3 20 23 23 7 1 16 14 5 12 3 0 14 12 14 23 6 17
M1 2 17 24 15 19 14
Text1 C R Y P T OText2 C R Y P T O
M2 2 17 24 15 19 14
C2 5 3 20 23 23 7 3 17 19 10 18 6 22 24 7 14 2 9 21
K
Inaugural Lecture - February 19th 2008 15
Fishing at a Depth
K 3 12 22 8 4 19
C1 5 3 20 23 23 7 1 16 14 5 12 3 0 14 12 14 23 6 17
M1 2 17 24 15 19 14
Text1 C R Y P T OText2 C R Y P T O
M2 2 17 24 15 19 14
C2 5 3 20 23 23 7 3 17 19 10 18 6 22 24 7 14 2 9 21
K 3 12 22 8 4 19
C=K+M mod 26
Inaugural Lecture - February 19th 2008 16
Fishing at a Depth
K 3 12 22 8 4 19
C1 5 3 20 23 23 7 1 16 14 5 12 3 0 14 12 14 23 6 17
M1 2 17 24 15 19 14
Text1 C R Y P T O G R A P H YText2 C R Y P T O
M2 2 17 24 15 19 14
C2 5 3 20 23 23 7 3 17 19 10 18 6 22 24 7 14 2 9 21
K 3 12 22 8 4 19
Inaugural Lecture - February 19th 2008 17
Fishing at a Depth
K 3 12 22 8 4 19
C1 5 3 20 23 23 7 1 16 14 5 12 3 0 14 12 14 23 6 17
M1 2 17 24 15 19 14 6 17 0 15 7 24
Text1 C R Y P T O G R A P H YText2 C R Y P T O
M2 2 17 24 15 19 14
C2 8 15 16 5 1 7 3 17 19 10 18 6 22 24 7 14 2 9 21
K 3 12 22 8 4 19
Inaugural Lecture - February 19th 2008 18
Fishing at a Depth
K 3 12 22 8 4 19 21 25 14 16 5 5
C1 5 3 20 23 23 7 1 16 14 5 12 3 0 14 12 14 23 6 17
M1 2 17 24 15 19 14 6 17 0 15 7 24
Text1 C R Y P T O G R A P H YText2 C R Y P T O
M2 2 17 24 15 19 14
C2 8 15 16 5 1 7 3 17 19 10 18 6 22 24 7 14 2 9 21
K 3 12 22 8 4 19
C=K+M mod 26
Inaugural Lecture - February 19th 2008 19
Fishing at a Depth
K 3 12 22 8 4 19 21 25 14 16 5 5
C1 5 3 20 23 23 7 1 16 14 5 12 3 0 14 12 14 23 6 17
M1 2 17 24 15 19 14 6 17 0 15 7 24
Text1 C R Y P T O G R A P H YText2 C R Y P T O
M2 2 17 24 15 19 14
C2 8 15 16 5 1 7 3 17 19 10 18 6 22 24 7 14 2 9 21
K 3 12 22 8 4 19 21 25 14 16 5 5
Equality of Keysteams
Inaugural Lecture - February 19th 2008 20
Fishing at a Depth
K 3 12 22 8 4 19 21 25 14 16 5 5
C1 5 3 20 23 23 7 1 16 14 5 12 3 0 14 12 14 23 6 17
M1 2 17 24 15 19 14 6 17 0 15 7 24
Text1 C R Y P T O G R A P H YText2 C R Y P T O
M2 2 17 24 15 19 14 8 18 5 20 13 1
C2 8 15 16 5 1 7 3 17 19 10 18 6 22 24 7 14 2 9 21
K 3 12 22 8 4 19 21 25 14 16 5 5
C=K+M mod 26
Inaugural Lecture - February 19th 2008 21
Fishing at a Depth
K 3 12 22 8 4 19 21 25 14 16 5 5
C1 5 3 20 23 23 7 1 16 14 5 12 3 0 14 12 14 23 6 17
M1 2 17 24 15 19 14 6 17 0 15 7 24
Text1 C R Y P T O G R A P H YText2 C R Y P T O I S F U N B
M2 2 17 24 15 19 14 8 18 5 20 13 1
C2 8 15 16 5 1 7 3 17 19 10 18 6 22 24 7 14 2 9 21
K 3 12 22 8 4 19 21 25 14 16 5 5
Inaugural Lecture - February 19th 2008 22
Fishing at a Depth
K 3 12 22 8 4 19 21 25 14 16 5 5
C1 5 3 20 23 23 7 1 16 14 5 12 3 0 14 12 14 23 6 17
M1 2 17 24 15 19 14 6 17 0 15 7 24
Text1 C R Y P T O G R A P H Y I S F U N BText2 C R Y P T O I S F U N B
M2 2 17 24 15 19 14 8 18 5 20 13 1
C2 8 15 16 5 1 7 3 17 19 10 18 6 22 24 7 14 2 9 21
K 3 12 22 8 4 19 21 25 14 16 5 5
Inaugural Lecture - February 19th 2008 23
Fishing at a Depth
K 3 12 22 8 4 19 21 25 14 16 5 5
C1 5 3 20 23 23 7 1 16 14 5 12 3 0 14 12 14 23 6 17
M1 2 17 24 15 19 14 6 17 0 15 7 24 8 18 5 29 13 1
Text1 C R Y P T O G R A P H Y I S F U N BText2 C R Y P T O I S F U N B
M2 2 17 24 15 19 14 8 18 5 20 13 1
C2 8 15 16 5 1 7 3 17 19 10 18 6 22 24 7 14 2 9 21
K 3 12 22 8 4 19 21 25 14 16 5 5
Related messages
Inaugural Lecture - February 19th 2008 24
Fishing at a Depth
K 3 12 22 8 4 19 21 25 14 16 5 5 18 22 7 20 10 5
C1 5 3 20 23 23 7 1 16 14 5 12 3 0 14 12 14 23 6 17
M1 2 17 24 15 19 14 6 17 0 15 7 24 8 18 5 29 13 1
Text1 C R Y P T O G R A P H Y I S F U N BText2 C R Y P T O I S F U N B
M2 2 17 24 15 19 14 8 18 5 20 13 1
C2 8 15 16 5 1 7 3 17 19 10 18 6 22 24 7 14 2 9 21
K 3 12 22 8 4 19 21 25 14 16 5 5
C=K+M mod 26
Inaugural Lecture - February 19th 2008 25
Fishing at a Depth
K 3 12 22 8 4 19 21 25 14 16 5 5 18 22 7 20 10 5
C1 5 3 20 23 23 7 1 16 14 5 12 3 0 14 12 14 23 6 17
M1 2 17 24 15 19 14 6 17 0 15 7 24 8 18 5 29 13 1
Text1 C R Y P T O G R A P H Y I S F U N BText2 C R Y P T O I S F U N B
M2 2 17 24 15 19 14 8 18 5 20 13 1
C2 8 15 16 5 1 7 3 17 19 10 18 6 22 24 7 14 2 9 21
K 3 12 22 8 4 19 21 25 14 16 5 5 18 22 7 20 10 5
Equality of Keysteams
Inaugural Lecture - February 19th 2008 26
Fishing at a Depth
K 3 12 22 8 4 19 21 25 14 16 5 5 18 22 7 20 10 5
C1 5 3 20 23 23 7 1 16 14 5 12 3 0 14 12 14 23 6 17
M1 2 17 24 15 19 14 6 17 0 15 7 24 8 18 5 29 13 1
Text1 C R Y P T O G R A P H Y I S F U N BText2 C R Y P T O I S F U N B
M2 2 17 24 15 19 14 8 18 5 20 13 1 4 2 0 20 18 4
C2 8 15 16 5 1 7 3 17 19 10 18 6 22 24 7 14 2 9 21
K 3 12 22 8 4 19 21 25 14 16 5 5 18 22 7 20 10 5
C=K+M mod 26
Inaugural Lecture - February 19th 2008 27
Fishing at a Depth
K 3 12 22 8 4 19 21 25 14 16 5 5 18 22 7 20 10 5
C1 5 3 20 23 23 7 1 16 14 5 12 3 0 14 12 14 23 6 17
M1 2 17 24 15 19 14 6 17 0 15 7 24 8 18 5 29 13 1
Text1 C R Y P T O G R A P H Y I S F U N BText2 C R Y P T O I S F U N B E C A U S E
M2 2 17 24 15 19 14 8 18 5 20 13 1 4 2 0 20 18 4
C2 8 15 16 5 1 7 3 17 19 10 18 6 22 24 7 14 2 9 21
K 3 12 22 8 4 19 21 25 14 16 5 5 18 22 7 20 10 5
Inaugural Lecture - February 19th 2008 28
Fishing at a Depth
K 3 12 22 8 4 19 21 25 14 16 5 5 18 22 7 20 10 5
C1 5 3 20 23 23 7 1 16 14 5 12 3 0 14 12 14 23 6 17
M1 2 17 24 15 19 14 6 17 0 15 7 24 8 18 5 29 13 1
Text1 C R Y P T O G R A P H Y I S F U N B EText2 C R Y P T O I S F U N B E C A U S E
M2 2 17 24 15 19 14 8 18 5 20 13 1 4 2 0 20 18 4
C2 8 15 16 5 1 7 3 17 19 10 18 6 22 24 7 14 2 9 21
K 3 12 22 8 4 19 21 25 14 16 5 5 18 22 7 20 10 5
Related messages
Inaugural Lecture - February 19th 2008 29
Fishing at a Depth
K 3 12 22 8 4 19 21 25 14 16 5 5 18 22 7 20 10 5
C1 5 3 20 23 23 7 1 16 14 5 12 3 0 14 12 14 23 6 17
M1 2 17 24 15 19 14 6 17 0 15 7 24 8 18 5 29 13 1 4
Text1 C R Y P T O G R A P H Y I S F U N B EText2 C R Y P T O I S F U N B E C A U S E
M2 2 17 24 15 19 14 8 18 5 20 13 1 4 2 0 20 18 4
C2 8 15 16 5 1 7 3 17 19 10 18 6 22 24 7 14 2 9 21
K 3 12 22 8 4 19 21 25 14 16 5 5 18 22 7 20 10 5
Inaugural Lecture - February 19th 2008 30
Fishing at a Depth
K 3 12 22 8 4 19 21 25 14 16 5 5 18 22 7 20 10 5 13
C1 5 3 20 23 23 7 1 16 14 5 12 3 0 14 12 14 23 6 17
M1 2 17 24 15 19 14 6 17 0 15 7 24 8 18 5 29 13 1 4
Text1 C R Y P T O G R A P H Y I S F U N B EText2 C R Y P T O I S F U N B E C A U S E
M2 2 17 24 15 19 14 8 18 5 20 13 1 4 2 0 20 18 4
C2 8 15 16 5 1 7 3 17 19 10 18 6 22 24 7 14 2 9 21
K 3 12 22 8 4 19 21 25 14 16 5 5 18 22 7 20 10 5
C=K+M mod 26
Inaugural Lecture - February 19th 2008 31
Fishing at a Depth
K 3 12 22 8 4 19 21 25 14 16 5 5 18 22 7 20 10 5 13
C1 5 3 20 23 23 7 1 16 14 5 12 3 0 14 12 14 23 6 17
M1 2 17 24 15 19 14 6 17 0 15 7 24 8 18 5 29 13 1 4
Text1 C R Y P T O G R A P H Y I S F U N B EText2 C R Y P T O I S F U N B E C A U S E
M2 2 17 24 15 19 14 8 18 5 20 13 1 4 2 0 20 18 4
C2 8 15 16 5 1 7 3 17 19 10 18 6 22 24 7 14 2 9 21
K 3 12 22 8 4 19 21 25 14 16 5 5 18 22 7 20 10 5 13
Equality of Keysteams
Inaugural Lecture - February 19th 2008 32
Fishing at a Depth
K 3 12 22 8 4 19 21 25 14 16 5 5 18 22 7 20 10 5 13
C1 5 3 20 23 23 7 1 16 14 5 12 3 0 14 12 14 23 6 17
M1 2 17 24 15 19 14 6 17 0 15 7 24 8 18 5 29 13 1 4
Text1 C R Y P T O G R A P H Y I S F U N B EText2 C R Y P T O I S F U N B E C A U S E
M2 2 17 24 15 19 14 8 18 5 20 13 1 4 2 0 20 18 4 8
C2 8 15 16 5 1 7 3 17 19 10 18 6 22 24 7 14 2 9 21
K 3 12 22 8 4 19 21 25 14 16 5 5 18 22 7 20 10 5 13
C=K+M mod 26
Inaugural Lecture - February 19th 2008 33
Fishing at a Depth
K 3 12 22 8 4 19 21 25 14 16 5 5 18 22 7 20 10 5 13
C1 5 3 20 23 23 7 1 16 14 5 12 3 0 14 12 14 23 6 17
M1 2 17 24 15 19 14 6 17 0 15 7 24 8 18 5 29 13 1 4
Text1 C R Y P T O G R A P H Y I S F U N B EText2 C R Y P T O I S F U N B E C A U S E I
M2 2 17 24 15 19 14 8 18 5 20 13 1 4 2 0 20 18 4 8
C2 8 15 16 5 1 7 3 17 19 10 18 6 22 24 7 14 2 9 21
K 3 12 22 8 4 19 21 25 14 16 5 5 18 22 7 20 10 5 13
Inaugural Lecture - February 19th 2008 34
Deducing Fish’s Structure
• Just such a depth was intercepted on 30th August 1941.– Two messages with same indicator HQIBPEXEZMUG.– Abbreviations, misspellings and corrections were inserted by
wireless operator when forced to retransmit a long message.– Operator should have chosen new message indicator, but did
not.
• Analysis by Tiltman then recovered the two messages.
• More importantly a sequence of nearly 4000 keystream letters was obtained.
• From this sequence, Tutte (later assisted by others) determined the entire structure of the Lorenz machine.
Inaugural Lecture - February 19th 2008 35
Lorenz SZ40 Structure
43
61
37
47 5351 59
Motor Wheels
Chi Wheels
Psi Wheels
Clock
41 31 29 2326
Keystream bits
Inaugural Lecture - February 19th 2008 36
Lorenz SZ40 Structure
• 5 parallel bits of keystream produced per clock pulse.– Bit-by-bit combined with message in Baudot coded form.
• 12 pinwheels, arranged in two groups of five (chi and psi) plus two motor wheels, M1 and M2.– Output bits taken from XOR sums of chi and psi wheels.– Chi wheels of lengths 41, 31, 29, 26, 23, clocked regularly.– Psi wheels of lengths 43, 47, 51, 53, 59, clocked irregularly,
according to output of M1.– M1 of length 37 clocked irregularly according to output of M2.– M2 of length 61 clocked regularly.
• Modern interpretation: irregularly clocked circulating shift registers.
• 2501 possible keys.– Monthly (later daily) setting of pins on each wheel.– Per message key: initial rotational offset of each wheel.
Inaugural Lecture - February 19th 2008 37
Lorenz SZ40
Size:51cm × 46cm × 46cm (20in × 18in × 18in)
Inaugural Lecture - February 19th 2008 38
Fish and Colossus
• In 1943, Max Newman raised the possibility of using a machine to automate the breaking of Fish.– Ideally suited to repetitive calculations involved in statistical
analysis developed by Tutte, Turing, and many others.– But initial all-mechanical machines were slow and unreliable.
• Tommy Flowers proposed and led the build of a rival electro-mechanical design, Colossus. – Based at Post Office Research Station, Dollis Hill, London.– Using 1500 state-of-the-art thermionic valves, thyratrons, and
photomultipliers.– Implementing shift registers, systolic arrays, configurable
Boolean operations on data,…– But not a Turing-complete machine.
Inaugural Lecture - February 19th 2008 39
Mechanised Cryptanalysis of Fish
• Colossus Mark I delivered 18th January 1944.
• Rapidly followed by first Colossus Mark II (2400 valves and 5 times as fast).
• Eventually 10 Colossi in 24-hour operation at Bletchley Park, with 11th in production.
Inaugural Lecture - February 19th 2008 40
The Value of Fish Traffic
• By 8th May 1945, Bletchley Park had broken 13508 messages on 718 keys, obtaining 63 million plaintext characters.
• Fish yielded information of great strategic value:– Strategic appreciations, order of battle, strength of individual
Wehermacht divisions.– German situation reports for the entire Russian front.– German strategic plans to hold on to Italy.– Information about likely success of D-Day landings:
• 8th May 1944, Field Marshall von Rundstedt to general staff, Berlin: an Allied assault on Normandy would “be the enemy’s pre-requisite condition for a subsequent descent on the Channel coast’’.
– Revelation of plans for counter-attack at Anzio beach-head.– Insight into Hitler’s mental state.
Inaugural Lecture - February 19th 2008 41
Other Aspects of the Fish Story
• Destruction of Colossi at the war’s end.– Colossus re-build project recently completed.
• Wartime work gave British scientists and engineers a head-start in the fledgling computer industry.
• Fish/Colossus story only began to emerge in the mid-1970s.– Several key documents only recently
declassified.• Including “General report on Tunny”.
– Whole story masterfully told in Paul Gannon’s “Colossus – Bletchley Park’s Greatest Secret” (Atlantic Press, 2006).
Tommy Flowers MBE
1905-1998
Inaugural Lecture - February 19th 2008 42
Fishing Lessons
• Kerckhoffs’ Principle not applicable, but lack of system knowledge only delayed the breaking of Fish.
• A single human error provided the key to unlocking Fish.– Keystream repetition for two closely related messages.
• At least three major intellectual achievements:– Initial decryption from a depth (Tiltman).– Deriving the Lorenz machine’s structure from keystream alone
(Tutte et al.).– Development of mechanised cryptanalysis (Newman, Flowers).
Inaugural Lecture - February 19th 2008 43
3. WEP and GSM
• In the late 1990’s, wireless equipment became cheap enough to be used in mass-market networking equipment.
• IEEE developed 802.11 family of WirelessLAN standards.– Operating in “free for all” unregulated frequencies.
• Recognition that encryption is needed because of broadcast nature of signals.
• IEEE 802.11b&g included WEP (Wired Equivalent Privacy) mechanisms.– Encryption.– Integrity protection for data.– Authentication of network nodes.
Inaugural Lecture - February 19th 2008 44
WEP (In)security
• World War Drive 2004 – Survey of 228,537 networks – 140,890 (60%) configured to use Open System
Authentication.– Meaning no encryption or authentication enabled.
• Demonstration of vulnerability.
• Legality of demo doubtful!
Inaugural Lecture - February 19th 2008 45
WEP (In)security
• WEP requires end-user to configure a shared key in every communicating device.– Easy in a small home network of 2 or 3 devices.– More difficult in a corporate environment with many devices.– Updating keys a major headache.– A classic key management problem.
• Worse still, the entire WEP design is seriously flawed.– Authentication is trivial to defeat.– Encryption shown to be weak by Fluhrer, Mantin and Shamir.– Cracking tools (Airsnort, WEPcrack) are widely available on
Internet.• Can recover WEP key in a matter on minutes.
• What went wrong?
Inaugural Lecture - February 19th 2008 46
GSM Security
• GSM = second generation mobile phone system.– 1.9 billion customers.– GSM networks in over 210 countries.
• Cryptography integrated as part of GSM from the start.– Algorithms and architecture designed by experts.– Security almost entirely hidden from end-users.– This security (especially key management) is not cost-free.
• Operators had a strong economic incentive to get the GSM security design right.– Protect revenue stream so as to recoup investment in licences
purchased from national governments.– Desire to avoid embarrassing breaches of personal privacy
occurring in first generation networks.
Inaugural Lecture - February 19th 2008 47
Lessons from WEP
• Economic incentives are often a major driver for adoption of security measures.– GSM using paid-for frequencies, 802.11 using free-
for-all frequencies.– Lack of incentive led to sloppy design in WEP.
• Employ security experts to design security systems, not enthusiasts.
• Good key management is hard and best not left to end-users.
Inaugural Lecture - February 19th 2008 48
Lessons from WEP
But: designers of WiMAX have recently repeated most of the same errors made in WEP design…
Those who cannot learn from history are doomed to repeat it.
George Santayana, Reason in Common Sense, The Life of Reason, Vol. 1.
You must learn from the mistakes of others. You can't possibly live long enough to make them all yourself.Sam Levenson
Inaugural Lecture - February 19th 2008 49
4. IPsec
• IPsec provides cryptographic protection for IP packets.– Encryption and integrity protection.
• An important system for protecting Internet traffic.– e.g. widely used in Virtual Private Networking
applications.
• Specified in IETF RFCs 4301-4309 and related documents.– RFCs are (essentially) standards for the Internet.– Very complex set of documents with many options.– 300+ pages of very technical text.
Inaugural Lecture - February 19th 2008 50
IPsec
• IPsec uses industrial-strength cryptography.
• Yet we still managed to break IPsec in certain encryption-only configurations.– Ciphertext-only attacks.– Attacks demonstrated in the lab.– Paterson and Yau (Eurocrypt 2006), Degabriele and
Paterson (IEEE Security and Privacy 2007).
Inaugural Lecture - February 19th 2008 51
Breaking IPsec
• Capture ciphertexts from the network.• Modify ciphertexts so as to produce predictable
changes to underlying messages.– Bit flipping weakness of CBC mode encryption.– Messages now have small, attacker-induced faults.
• Inject modified ciphertexts into the network.• IPsec decryption results in faulty IP packets.
– IP produces ICMP error messages when these faulty packets are further processed.
– ICMP messages are not encrypted and carry portions of faulty IP packets.
– These can be intercepted.
Inaugural Lecture - February 19th 2008 52
Breaking IPsec
CiphertextC
Key K
EncryptionAlgorithm
MessageM
DecryptionAlgorithm
MessageM
Key K
Interceptor
Active attacker
Reactive System
Inaugural Lecture - February 19th 2008 53
Breaking IPsec
• The encryption-only configurations that we broke were already known to have theoretical weaknesses.– Bellovin (1995, 1996), using ideas of Wagner.
• So why were they still allowed in the standards?
Inaugural Lecture - February 19th 2008 54
Breaking IPsec
RFC 4303:
“Using encryption-only for confidentiality is allowed by ESP. However, it should be noted that in general, this will provide defense only against passive attackers.”
“ESP allows encryption-only … because this may offer considerably better performance and still provide adequate security, e.g., when higher layer authentication/integrity protection is offered independently.”
Inaugural Lecture - February 19th 2008 55
Breaking IPsec
• From the IPsec administrator's guide of a well-known vendor:
“If you require data confidentiality only in your IPSec tunnel implementation, you should use ESP without authentication. By leaving off the authentication service, you gain some performance speed but lose the authentication service.”
http://www.cisco.com/en/US/docs/security/security_management/vms/router_mc/1.3.x/user/guide/U13_bldg.html#wp1068306 (last accessed 16/2/2008).
Inaugural Lecture - February 19th 2008 56
IPsec Lessons
• Cryptography is only ever a component in a secure system and should not be viewed in isolation.
• Encryption on its own is not sufficient to provide confidentiality.
• Be aware of shifts in the adversary’s capabilities.
• Complexity and flexibility are the enemies of security.
• Sacrifice backward compatibility if security is the primary objective.
• Gulf in understanding between theoreticians, standards writers, implementers, and users.– Security message gets lost in translation.
Inaugural Lecture - February 19th 2008 57
5. Phishing
• Demonstration: let’s take an on-line test.http://www.sonicwall.com/phishing/
• An attack of this general type is known as a phishing attack.
• 6 Billion phishing e-mails are sent world-wide each month.
• Average loss per successful attack is estimated at $1200 (Federal Trade Commission).– Junk e-mail is a lot cheaper to send than junk mail.– So even if only a tiny fraction are successful, it’s still
economically viable for the attacker.
Inaugural Lecture - February 19th 2008 58
Phishing
• Phishing exploits a mixture of human gullibility, technological naivety, fear, and sometimes greed.– Users trust that “From” address in e-mail is a
guarantee of origin, and that link in e-mail is a guarantee of destination for their sensitive data.
• Arguably, cryptography is of no use at all in preventing this form of attack.– Unless we had a global authentication infrastructure
that is used universally to prove the origin of all e-mails.
Inaugural Lecture - February 19th 2008 59
Phishing Lessons
• Cryptography has its limitations.• Don’t rely on a technology to do a job for which
it was never designed.– Smart banks never use e-mail to ask their
customers to do anything sensitive.– Unfortunately, their customers don’t all know this
yet.
• Much more research is needed in the area of humans and security.– How humans take security-sensitive decisions, and
how they can be guided towards making better ones.
Inaugural Lecture - February 19th 2008 60
6. Concluding Remarks
• Cryptography is one of the most powerful tools we have in our security armoury.
• Implementing, deploying and managing effective cryptography is difficult and expensive.– Key management may be hardest of all.
• In theory, theory and practice are the same. In practice, they are not.
• Eliminate humans (and human error).
• Watch out for changing adversaries.
• Recognise the limitations of cryptography.
• Learn from history.
Inaugural Lecture - February 19th 2008 61
Thanks
• Thanks to Marta Baker and her staff.
• Many thanks to colleagues and students for making the ISG such a special place to work.
• Many, many thanks to Fred Piper for his immeasurable and constant support over the years.
• And thank you all for coming.