secrecy capacities and multiterminal source coding
play

Secrecy Capacities and Multiterminal Source Coding Prakash Narayan - PowerPoint PPT Presentation

Secrecy Capacities and Multiterminal Source Coding Prakash Narayan Joint work with Imre Csisz ar and Chunxuan Ye Multiterminal Source Coding The Model n X 2 x 2 n x n x X X 1 3 1 3 x m n X m m 2 terminals. X 1


  1. Secrecy Capacities and Multiterminal Source Coding Prakash Narayan Joint work with Imre Csisz´ ar and Chunxuan Ye

  2. Multiterminal Source Coding

  3. The Model n X 2 x 2 n x n x X X 1 3 1 3 x m n X m • m ≥ 2 terminals. • X 1 , . . . , X m , m ≥ 2, are rvs with finite alphabets X 1 , . . . , X m . • Consider a discrete memoryless multiple source with components X n 1 = ( X 11 , . . . , X 1 n ) , . . . , X n m = ( X m 1 , . . . , X mn ). • Terminal X i observes the component X n i = ( X i 1 , . . . , X in ).

  4. The Model x 2 F F 2 m+2 F 3 F x x F m+3 3 m+1 1 F 1 F F F rm 2m m x m • The terminals are allowed to communicate over a noiseless channel, possibly interactively in several rounds. • All the transmissions are observed by all the terminals. • No rate constraints on the communication. • Assume w.l.o.g that transmissions occur in consecutive time slots in r rounds. △ • Communication depicted by rvs F = F 1 , . . . F rm , where ∗ F ν = transmission in time slot ν by terminal i ≡ ν mod m . ∗ F ν is a function of X n i and ( F 1 , . . . , F ν − 1 ).

  5. Communication for Omniscience x 2 F F 2 m+2 F 3 F x x F m+3 3 m+1 1 F 1 F F F rm 2m m x m • Each terminal wishes to become “omniscient,” i.e., recover ( X n 1 , . . . , X n m ) with probability ≥ 1 − ε . • What is the smallest achievable rate of communication for omniscience (CO-rate), lim n 1 n H ( F 1 , . . . , F rm )?

  6. Minimum Communication for Omniscience Proposition [I. Csisz´ ar - P. N., ’02]: The smallest achievable CO-rate, n H ( F ( n ) , . . . , F ( n ) lim n 1 rm ), which enables ( X n 1 , . . . , X n m ) to be ε n -recoverable at all the 1 terminals with communication ( F ( n ) , . . . , F ( n ) rm ) (with the number of rounds possibly 1 depending on n ), with ε n → 0, is m � R min = min R i , ( R 1 ,... ,R m ) ∈R SW i =1 � � ′ ′ ′ where R SW = ( R 1 , · · · , R m ) : � i ≥ H ( X B | X B c ) , B ⊂ { 1 , . . . , m } i ∈ B R . Remark : The region R SW , if stated for all B ⊆ { 1 , . . . , m } , gives the achievable rate region for the multiterminal version of the Slepian-Wolf source coding theorem. Case: m = 2; R min = H ( X 1 | X 2 ) + H ( X 2 | X 1 ).

  7. Communication for Omniscience Proof of Proposition: The proposition is a source coding theorem of the “Slepian-Wolf” type, with the additional element that interactive communication is not a priori excluded. Achievability: Straightforward extension of the multiterminal Slepian-Wolf source coding theorem; the CO-rates can be achieved with noninteractive communication. Converse: Nontrivial; consequence of the following “Main Lemma.”

  8. Common Randomness n F K = K (X , ) 2 2 2 x 2 n F n K = K (X , ) K = K (X , ) F x x 1 1 3 3 3 1 1 3 x m n F K = K (X , ) m m m Common Randomness (CR): A function K of ( X n 1 , · · · , X n m ) is ε - CR , achievable with communication F , if Pr { K = K 1 = · · · = K m } ≥ 1 − ε. Thus, CR consists of random variables generated by different terminals, based on – local measurements or observations – transmissions or exchanges of information such that the random variables agree with probability ∼ = 1.

  9. Main Lemma . . . B . . . m 1 Lemma [I. Csisz´ ar - P. N., ’02]: If K is ε -CR for the terminals X 1 , · · · , X m , achievable with communication F = ( F 1 , · · · , F rm ), then m 1 R i + m ( ε log |K| + 1) � nH ( K | F ) = H ( X 1 , · · · , X m ) − n i =1 for some numbers ( R 1 , · · · , R m ) ∈ R SW where � � ′ ′ � ′ R SW = ( R 1 , · · · , R m ) : i ≥ H ( X B | X B c ) , B ⊂ { 1 , . . . , m } R . i ∈ B Remark : Decomposition of total joint entropy H ( X 1 , . . . , X m ) into the normalized conditional entropy of any achievable ε -CR conditioned on the communication with which it is achieved, and a sum of rates which satisfy the SW conditions.

  10. Secrecy Capacities

  11. The General Model (X ,...,X ) User 2 21 2n (X ,...,X ) (X ,...,X ) 31 3n 11 1n User 1 User 3 Wiretapper User m (Z ,...,Z ) 1 n (X ,...,X ) m1 mn The user terminals wish to generate CR which is effectively concealed from an eavesdropper with access to the public interterminal communication or from a wiretapper.

  12. Secret Key n F K = K (X , ) 2 2 2 x 2 n F n K = K (X , ) K = K (X , ) F x x 1 1 3 3 3 1 1 3 x m n F K = K (X , ) m m m Secret Key (SK): A function K of ( X n 1 , · · · , X n m ) is an ε - SK , achievable with communication F , if • Pr { K = K 1 = · · · = K m } ≥ 1 − ε (“ ε -common randomness”) 1 • n I ( K ∧ F ) ≤ ε (“secrecy”) 1 n H ( K ) ≥ 1 • n log |K| − ε (“uniformity”) where K = set of all possible values of K . Thus, a secret key is effectively concealed from an eavesdropper with access to F , and is nearly uniformly distributed.

  13. Secret Key Capacity n F K = K (X , ) 2 2 2 x 2 n F n K = K (X , ) K = K (X , ) F x x 1 1 3 3 3 1 1 3 x m n F K = K (X , ) m m m • Achievable SK-rate: The (entropy) rate of such a SK, achievable with suitable communication (with the number of rounds possibly depending on n ). • SK-capacity C SK = largest achievable SK-rate.

  14. Some Recent Related Work • Maurer 1990, 1991, 1993, 1994, · · · • Ahlswede-Csisz´ ar 1993, 1994, 1998, · · · • Bennett, Brassard, Cr´ epeau, Maurer 1995. • Csisz´ ar 1996. • Maurer - Wolf 1997, 2003, · · · • Venkatesan - Anantharam 1995, 1997, 1998, 2000, · · · • Csisz´ ar - Narayan 2000. • Renner-Wolf 2003. . . . . . .

  15. The Connection

  16. Special Case: Two Users n n X X 2 1 ~H(X |X ) 1 2 x x 1 2 ~H(X |X ) 2 1 Observation C SK = I ( X 1 ∧ X 2 ) [Maurer 1993, Ahlswede - Csisz´ ar 1993] = H ( X 1 , X 2 ) − [ H ( X 1 | X 2 ) + H ( X 2 | X 1 )] = Total rate of shared CR − Smallest achievable CO-rate ( R min ) .

  17. The Main Result • SK-capacity [I. Csisz´ ar - P. N., ’02]: C SK = H ( X 1 , . . . , X m ) − Smallest achievable CO-rate, R min , i.e., smallest rate of communication which enables each terminal to reconstruct all the m components of the multiple source. • A single-letter characterization of R min , thus, leads to the same for C SK . Remark : The source coding problem of determining the smallest achievable CO-rate R min does not involve any secrecy constraints .

  18. Secret Key Capacity Theorem [I. Csisz´ ar - P. N., ’02]: The SK-capacity C SK for a set of terminals { 1 , . . . , m } equals C SK = H ( X 1 , . . . , X m ) − R min , and can be achieved with noninteractive communication. Proof : Converse: From Main Lemma. Idea of achievability proof: If L represents ε -CR for the set of terminals, achievable with communication F for some block length n , then 1 n H ( L | F ) is an achievable SK-rate if ε is small. With L ∼ = ( X n 1 , . . . , X n m ) , we have 1 = H ( X 1 , . . . , X m ) − 1 nH ( L | F ) ∼ nH ( F ) . Remark: The SK-capacity is not increased by randomization at the terminals. Case : m = 2; C SK = I ( X 1 ∧ X 2 ).

  19. Example x 2 x x 1 3 x m [I. Csisz´ ar - P. N.,’03]: • X 1 , · · · , X m − 1 are { 0 , 1 } -valued, mutually independent, ( 1 2 , 1 2 ) rvs, and X mt = X 1 t + · · · + X ( m − 1) t mod 2 , t ≥ 1 . • Total rate of shared CR= H ( X 1 , . . . , X m ) = H ( X 1 , . . . , X m − 1 ) = m − 1 bits. • R min = . . . = m ( m − 2) bits m − 1 • C SK = ( m − 1) − m ( m − 2) 1 = m − 1 bit. m − 1

  20. Example – Scheme for Achievability • Claim : 1 bit of perfect SK (i.e., with ε = 0) is achievable with observation length n = m − 1. • Scheme with noninteractive communication: - Let n = m − 1. - For i = 1 , · · · , m − 1, X i transmits F i = f i ( X n i ) = block X n i excluding X ii . - X m transmits F m = f m ( X n m ) = ( X m 1 + X m 2 mod 2 , X m 1 + X m 3 mod 2 , · · · , X m 1 + X mn mod 2) . • X 1 , · · · , X m all recover ( X n 1 , · · · , X n m ). (Omniscience) • In particular, X 11 is independent of F = ( F 1 , · · · , F m ). 1 1 • X 11 is an achievable perfect SK , so C SK ≥ m − 1 H ( X 11 ) = m − 1 bit.

  21. Eavesdropper with Wiretapped Side Information (X ,...,X ) User 2 21 2n (X ,...,X ) (X ,...,X ) 31 3n 11 1n User 1 User 3 Wiretapper User m (Z ,...,Z ) 1 n (X ,...,X ) m1 mn • The secrecy requirement now becomes 1 nI ( K ∧ F , Z n ) ≤ ε. • General problem of determining the “Wiretap Secret Key” capacity, C WSK , remains unsolved.

  22. Wiretapping of Noisy User Sources The eavesdropper can wiretap noisy versions of some or all of the components of the underlying multiple source. Formally, m � Pr { Z 1 = z 1 , . . . , Z m = z m | X 1 = x 1 , . . . , X m = x m } = Pr { Z i = z i | X i = x i } . i =1 Theorem [I. Csisz´ ar - P. N., ’03]: The WSK-capacity for a set of terminals { 1 , . . . , m } equals C WSK = H ( X 1 , . . . , X m , Z 1 , . . . , Z m ) − “Revealed” entropy H ( Z 1 , . . . , Z m ) − Smallest achievable CO-rate for user terminals when they additionally know ( Z 1 , . . . , Z m ) = H ( X 1 , . . . , X m | Z 1 , . . . , Z m ) − R min ( Z 1 , . . . , Z m ) , provided that randomization is permitted at the user terminals. Case : m = 2; C WSK = I ( X 1 ∧ X 2 | Z 1 , Z 2 ).

  23. A Few Variants

Recommend


More recommend