Comprehensive Exam Ph.D. in Electrical Engineering

1. The Theory of Concurrent Codes with Application to Omnidirectional Jam-Resistant Communications without Shared Secrets Comprehensive ExamPh.D. in Electrical Engineering William L. Bahn 14 May 2007

2. The future of warfare:Net-centric, Joint, and Coalition

3. What’s the point of this work? Concurrent codes address one component of secure communications: The availability of the communications link in situations where high directionality and/or shared secrets are not feasible.

4. This problem involves several disciplines and needs more overlap than usually exists.

5. Each goal is achievable when the good guys share secrets that the bad guys don’t know. Secure communications occur only when all four goals are achieved. Secure communications involve four distinct security goals. Can the bad guys read my message? Encryption using symmetric cryptography Can the bad guys change my message? Hash functions, message digests, MACs Integrity Confidentiality Security Authenticity Availability Can the bad guys forge my message? Passwords Can the bad guys jam my message? Spread spectrum

6. Narrowband communications work fine in an nice, friendly, ideal world. NB

7. But they are easily jammed by any competing signal of similar power. NB

8. Spread spectrum provides protection against a competing signal. SS

9. In Frequency Hop Spread Spectrum (FH/SS), Sender and Receiver change frequencies according to a schedule. • Time Freq • 14 • 45 • 23 • 12 • 19 • 31 • 42 • Time Freq • 14 • 45 • 23 • 12 • 19 • 31 • 42 FH/SS

10. Jammer doesn’t know schedule, so… they jam random frequencies. • Time Freq • 14 • 45 • 23 • 12 • 19 • 31 • 42 • Time Freq • 38, 27, 24 • 19, 26, 45 • 18, 33, 37 • 15, 25, 29 • 13, 28, 44 • 29, 31, 49 • 22, 30, 42 • Time Freq • 14 • 45 • 23 • 12 • 19 • 31 • 42 FH/SS Problem: Jammer increases bit error rate (BER) Solution: Error correcting codes

11. Frequency sequence exchanged using a “secure alternate channel.” • Time Freq • 14 • 45 • 23 • 12 • 19 • 31 • 42 • Time Freq • 14 • 45 • 23 • 12 • 19 • 31 • 42 The symmetric key is any and all information that must be kept from the jammer but that both the sender and the receiver must have access to.

12. What if the alternate channel isn’t so secure? • Time Freq • 14 • 45 • 23 • 12 • 19 • 31 • 42 • Time Freq • 14 • 45 • 23 • 12 • 19 • 31 • 42 An informed jammer knows (or somehow obtains) the symmetric key. They do not know any private keys - information that only the sender knows or that only the receiver knows.

13. An informed jammer DOES know schedule, so they jam the right frequencies. • Time Freq • 14 • 45 • 23 • 12 • 19 • 31 • 42 • Time Freq • 14 • 45 • 23 • 12 • 19 • 31 • 42 • Time Freq • 14 • 45 • 23 • 12 • 19 • 31 • 42 FH/SS Problem: An “informed jammer” can reduce processing gain to unity. Solution: ????

14. Traditional Spread Spectrum relies on shared secrets staying secret. Private Secrets Shared Secrets Private Secrets Receiver Sender Public Information Jammer

15. The management of symmetric keys is not scaleable and cannot meet the requirements of the GIG. • Very small unit level (10 people) • Key Pairs: 45 • Medium unit level (1,000 people) • Key Pairs: ~500,000 • Small theater-scale: (100,000 people) • Pair Keys: 5 billion • Coalition-scale: (1,000,000 people) • Pair Keys: 500 billion

16. An “informed jammer” can exploit all of the shared secrets. Private Secrets Shared Secrets Private Secrets Receiver Sender Public Information Jammer

17. But how can we communicate securely without shared secrets?First, how do we do it with shared secrets? Symmetric Cryptography Attack at dawn! Attack at dawn! K K SENDER U&3ro0+wn@”}EJn RECEIVER A single key both encrypts and decrypts a message. Both sender and receiver must possess it. Attacker must NOT possess it. An attacker can compromise the distribution process.

18. Asymmetric Cryptography simply uses two keys! Asymmetric Cryptography Attack at dawn! Attack at dawn! A B SENDER kO$7*jfMsi@4ifnnY RECEIVER Anything encrypted with one key can only be decrypted with the other key: P = T(T(P,A),B); P = T(T(P,B),A) Receiver generates A and B. Key A is distributed – to everyone (Public Key). Key B is kept secret – from everyone (Private Key).

19. Three of the four security goals can be achieved using PKI. NOTE: This is a highly simplified description of how PKI works in the real world.

20. If a shared secret is not available, a hole emerges for omnidirectional links. Each goal is achievable when the good guys share secrets that the bad guys don’t know. Encryption using asymmetric cryptography Encryption using symmetric cryptography Can the bad guys read my message? Can the bad guys change my message? Hash functions, message digests, MACs Digital Signatures Integrity Confidentiality Security Authenticity Availability Digital Signatures Can the bad guys forge my message? Passwords Omnidirectional SS links jammed as easily as NB Highly directional links or spread spectrum Can the bad guys jam my message?

21. What’s the point of this work? Concurrent codes address one component of secure communications: The availability of the communications link in situations where high directionality and/or shared secrets are not feasible.

22. Error detecting and correcting codes are great for dealing with random noise – concurrent codes are designed to deal with malicious non-random noise.

25. So how is it done?BBC Algorithm 101 • Encode by placing “indelible marks” at locations dictated by progressively longer prefixes of the message. • Decode by looking for “indelible marks” at locations dictated by progressively longer prefixes of the message.

26. An “indelible mark” is a transmission that is very difficult for an attacker to suppress. • UWB • Short of noise at a specific time. • FH • Noise at a specific carrier frequency. • DS • Random data at a given code/offset. The mark is not data modulated – it is data placed. No data is present in the mark – the presence of the mark is the data. The attacker can distort the mark – as long as we can still detect it. The attacker can add additional marks – we can deal with that.

27. Checksum bits appended to message eliminate terminal hallucinations. • Appended 0-bits act as checksum bits. • Terminal hallucinations survive each checksum bit at a rate equal to the packet mark density. • Overall rate for k checksum bits: • k = 19 => 1ppb at 33% density.

28. Impulse-based UWB Implementation.

29. Simple receiver leaves little for attacker to attack.

30. BBC: Sequential decoding performs depth first search in linear time.

31. Exponential Receiver Blow-up does not occur below 50% mark density. • Steady-state hallucination level: • Receiver effort doubled at 33% density. • Receiver effort 10x at 47% density. • If attacker can afford to broadcast 33%, they can likely afford to broadcast 100%.

32. Actual and predicted receiver workload in very close agreement.

33. Audio BBC recordings of 1 through 4 concurrent messages. 1 3 2 4

34. Actual workload at 99% packet density oscillates in close agreement with predicted bounds.

35. Concurrent codes have potential applications beyond hostile jam-resistance. • RFID • Jamming an issue for item-level tagging. • MAC-less networks • Wired or wireless. • No collision detection/avoidance – just transmit! • To prevent self-jamming, devices monitor mark density and adjust data rate accordingly. • Information Retrieval • Can perform more powerful searches than present techniques.

36. Concurrent codes are NOT Nirvana! • The system can still be jammed. • As can all the others. • There is a penalty to be paid. • As there is with the others. • The goal is to not to be more jam-resistant than uncompromised spread spectrum. • It isn’t. • The goal is to retain a comparable level of jam-resistance without a shared secret. • It retains roughly half of the data rate.

37. Demo Programs • JAVA Image Demo • BBC Image Demo • JAVA Audio Demo • BBC Audio Demo