PAUL CUFFELECTRICAL ENGINEERING

PRINCETON UNIVERSITY

Secure Communication for Distributed Systems

Main Idea

Secrecy for distributed systems

Design encryption specifically for a system objective

Node A

Node BMessageInformation

Action

Adversary

Distributed System

Attack

Communication in Distributed Systems

“Smart Grid”

Image from http://www.solarshop.com.au

Cipher

Plaintext: Source of information: Example: English text: Information Theory

Ciphertext: Encrypted sequence: Example: Non-sense text: cu@sp4isit

Encipherer

Decipherer

Ciphertext

Key Key

Plaintext Plaintext

Example: Substitution Cipher

Alphabet A B C D E …Mixed Alphabet F Q S A R …

Simple Substitution

Example: Plaintext: …RANDOMLY GENERATED CODEB… Ciphertext: …DFLAUIPV WRLRDFNRA SXARQ…

Caesar CipherAlphabet A B C D E …Mixed Alphabet D E F G H …

Shannon Model

Schematic

Assumption Enemy knows everything about the system except the

keyRequirement

The decipherer accurately reconstructs the information

Encipherer

Decipherer

Ciphertext

Key Key

Plaintext Plaintext

Adversary

C. Shannon, "Communication Theory of Secrecy Systems," Bell Systems Technical Journal, vol. 28, pp. 656-715, Oct. 1949.

For simple substitution:

Shannon Analysis

Perfect Secrecy Adversary learns nothing about the information Only possible if the key is larger than the information

C. Shannon, "Communication Theory of Secrecy Systems," Bell Systems Technical Journal, vol. 28, pp. 656-715, Oct. 1949.

Shannon Analysis

Equivocation vs Redundancy Equivocation is conditional entropy: Redundancy is lack of entropy of the source: Equivocation reduces with redundancy:

C. Shannon, "Communication Theory of Secrecy Systems," Bell Systems Technical Journal, vol. 28, pp. 656-715, Oct. 1949.

Computational Secrecy

Assume limited computation resourcesPublic Key Encryption

Trapdoor Functions

Difficulty not proven Often “cat and mouse” game

Vulnerable to quantum computer attack

W. Diffie and M. Hellman, “New Directions in Cryptography,” IEEE Trans. on Info. Theory, 22(6), pp. 644-654, 1976.

1125897758 834 689524287

2147483647

X

Information Theoretic Secrecy

Achieve secrecy from randomness (key or channel), not from computational limit of adversary.

Physical layer secrecy Wyner’s Wiretap Channel [Wyner 1975]

Partial Secrecy Typically measured by “equivocation:” Other approaches:

Error exponent for guessing eavesdropper [Merhav 2003]

Cost inflicted by adversary [this talk]

Shannon Analysis

Equivocation vs Redundancy Equivocation is conditional entropy: Redundancy is lack of entropy of the source: Equivocation reduces with redundancy:

C. Shannon, "Communication Theory of Secrecy Systems," Bell Systems Technical Journal, vol. 28, pp. 656-715, Oct. 1949.

Equivocation

Not an operationally defined quantity

Bounds: List decoding Additional information needed for decryption

Not concerned with structure

Guessing Eavesdropper

[Merhav 03]:

X=0 X=1 X=20

0.20.40.6

Prior Distribution (X)

Prob. of correct guess

Guessing Eavesdropper

[Merhav 03]:

X=00X=01

X=02X=10

X=11X=12

X=20X=21

X=220

0.2

Prior Distribution (X1,X2)

Prob. of correct guess

Traditional View of Encryption

Information inside

Proposed View of Encryption

Information obscured

Images from albo.co.uk

Competitive Distributed System

Node A Node BMessage

Key

Information Action

Adversary

Attack

Encoder:

System payoff: .

Decoder:

Adversary:

Zero-Sum Game

Value obtained by system:Objective

Maximize payoff

Node A Node BMessage

Key

Information Action

Adversary

Attack

Secrecy-Distortion Literature

[Yamamoto 97]: Cause an eavesdropper to have high reconstruction

distortion Replace payoff (π) with distortion

[Yamamoto 88]: No secret key Lossy compression

Secrecy-Distortion Literature

[Theorem 3, Yamamoto 97]:

Theorem:

Choose Yields

How to Force High Distortion

Randomly assign binsSize of each bin is Adversary only knows bin

Reconstruction of only depends on the marginal posterior distribution of

Example:

INFORMATION THEORETIC RATE REGIONS

PROVABLE SECRECY

Theoretical Results

Lossless Transmission General Reward Function

Simplex interpretation Linear program

Hamming Distortion

Common Information Secret Key

Two Categories of Results

Competitive Distributed System

Node A Node BMessage

Key

Information Action

Adversary

Attack

Encoder:

System payoff: .

Decoder:

Adversary:

Zero-Sum Game

Value obtained by system:Objective

Maximize payoff

Node A Node BMessage

Key

Information Action

Adversary

Attack

Theorem:

[Cuff 10]

Lossless Case

Require Y=X Assume a payoff function

Related to Yamamoto’s work [97] Difference: Adversary is more capable with more

information

Also required:

Binary-Hamming Case

Binary Source:Hamming DistortionNaïve approach

Random hashing or time-sharingOptimal approach

Reveal excess 0’s or 1’s to condition the hidden bits

0 1 0 0 1 0 0 0 0 1* * 0 0 * * 0 * 0 *

SourcePublic message

(black line)

(orange line)

Linear Program on the Simplex

Category 1Category 2

Category 4

00.51

Series 1

Series 2

Series 3

Series 4

Series 5Series 1Series 2Series 3Series 4Series 5

Linear Program on the Simplex

Constraint:

Minimize:

Maximize:

U will only have mass at a small subset of points (extreme points)

Hamming Distortion – Low Key Rate

Arbitrary i.i.d. source (on finite alphabet)Hamming DistortionSmall Key Rate

Confuse with sets of size two Either reveal X or reveal that X is in {0,1} during

each instance ∏ = R0/2

General Payoff FunctionNo requirement for lossless transmission.

Any payoff function π(x,y,z)Any source distribution

(i.i.d.)

Adversary:

Payoff-Rate Function

Maximum achievable average payoff

Markov relationship:

Theorem:

Unlimited Public Communication

Maximum achievable average payoff

Conditional common information:

Theorem (R=∞):

Encoding Scheme

Coordination Strategies [Cuff-Permuter-Cover 10] Empirical coordination for U Strong coordination for Y

K

Converse

What the Adversary doesn’t know can hurt him.

[Yamamoto 97]

Knowledge of Adversary:

[Yamamoto 88]:

Summary

Framework for Encryption Average cost inflicted by adversary Dynamic settings where information is available

causally No use of “equivocation” Optimal performance uses both “strong” and

“empirical” coordination.