1. 1 Interactive Proofs proof systems interactive proofs and their power Arthur-Merlin games

2. 2 Black Board Topics (non-powerpointed) IP for PERMANENT Average case hardness of PERMANENT Checker for Permanent Direct proof of IP-PSPACE via PSPACE TM. Epsilon links (modus ponens view) of IP proofs.

3. 3 Proof systems L = { (A, 1 k ) : A is a true mathematical assertion with a proof of length k} What is a “proof”? complexity insight: meaningless unless can be efficiently verified

4. 4 Proof systems given language L, goal is to prove x  L proof system for L is a verification algorithm V –completeness: x  L  9 proof, V accepts (x, proof) “true assertions have proofs” –soundness: x  L  8 proof*, V rejects (x, proof*) “false assertions have no proofs” –efficiency: 8 x, proof: V(x, proof) runs in polynomial time in |x|

5. 5 Classical Proofs previous definition: “classical” proof system recall: L  NP iff expressible as L = { x | 9 y, |y| < |x| k, (x, y)  R } and R  P. NP is the set of languages with classical proof systems (R is the verifier)

6. 6 Interactive Proofs Two new ingredients: –randomness: verifier tosses coins, errs with some small probability –interaction: rather than “reading” proof, verifier interacts with computationally unbounded prover NP proof systems lie in this framework: prover sends proof, verifier does not use randomness

7. 7 Interactive Proofs interactive proof system for L is an interactive protocol (P, V) ProverVerifier...... common input: x accept/ reject # rounds = poly(|x|)

8. 8 Interactive Proofs interactive proof system for L is an interactive protocol (P, V) –completeness: x  L  Pr[V accepts in (P, V)(x)]  2/3 –soundness: x  L  8 P* Pr[V accepts in (P*, V)(x)]  1/3 –efficiency: V is p.p.t. machine repetition: can reduce error to any ε

9. 9 Interactive Proofs IP = {L : L has an interactive proof system} Observations/questions: –philosophically interesting: captures more broadly what it means to be convinced a statement is true –clearly NP  IP. Potentially larger. How much larger? –if larger, randomness is essential (why?)

10. 10 Graph Isomorphism graphs G 0 = (V, E 0 ) and G 1 = (V, E 1 ) are isomorphic (G 0  G 1 ) if exists a permutation π:V  V for which (x, y)  E 0  (π(x), π(y))  E 1 1 2 3 4 1 2 4 3 1 4 3 2

11. 11 Graph Isomorphism GI = {(G 0, G 1 ) : G 0  G 1 } –in NP –not known to be in P, or NP-complete GNI = complement of GI –not known to be in NP Theorem (GMW): GNI  IP –indication IP may be more powerful than NP

12. 12 GNI in IP interactive proof system for GNI: ProverVerifier input: (G 0, G 1 ) flip coin c  {0,1}; pick random π H = π(G c ) if H  G 0 r = 0, else r = 1 r accept iff r = c

13. 13 GNI in IP completeness: –if G 0 not isomorphic to G 1 then H is isomorphic to exactly one of (G 0, G 1 ) –prover will choose correct r soundness: –if G 0  G 1 then prover sees same distribution on H for c = 0, c = 1 –no information on c  any prover P* can succeed with probability at most 1/2

14. 14 The power of IP We showed GNI 2 IP GNI  IP suggests IP more powerful than NP, since we don’t know how to show GNI in NP GNI in coNP Theorem (LFKN): coNP  IP

15. 15 The power of IP Proof idea: input: φ(x 1, x 2, …, x n ) –prover: “I claim φ has k satisfying assignments” –true iff φ(0, x 2, …, x n ) has k 0 satisfying assignments φ(1, x 2, …, x n ) has k 1 satisfying assignments k = k 0 + k 1 –prover sends k 0, k 1 –verifier sends random c  {0,1} –prover recursively proves “φ’ = φ(c, x 2, …, x n ) has k c satisfying assignments” –at end, verifier can check for itself.

16. 16 The power of IP Analysis of proof idea: –Completeness: φ(x 1, x 2, …, x n ) has k satisfying assignments  accept with prob. 1 –Soundness: φ(x 1, x 2, …, x n ) does not have k satisfying assigns.  accept prob.  1 – 2 -n –Why? It is possible that k is only off by one; verifier only catches prover if coin flips c are successive bits of this assignment

17. 17 The power of IP Solution to problem (ideas): –replace {0,1} n with (F q ) n –verifier substitutes random field element at each step –vast majority of field elements catch cheating prover (rather than just 1) Theorem: L = { (φ, k): CNF φ has exactly k satisfying assignments} is in IP

18. 18 The power of IP First step: arithmetization –transform φ(x 1, … x n ) into polynomial p φ (x 1, x 2, … x n ) of degree d over a field F q ; q prime > 2 n –recursively: x i  x i   φ  (1 - p φ ) φ  φ’  (p φ )(p φ’ ) φ  φ’  1 - (1 - p φ )(1 - p φ’ ) –for all x  {0,1} n we have p φ (x) = φ(x) –degree d  |φ| –can compute p φ (x) in poly time from φ and x

19. 19 The power of IP Prover wishes to prove: k = Σ x 1 = 0, 1 Σ x 2 = 0,1 … Σ x n = 0, 1 p φ (x 1, x 2, …, x n ) Define: k z = Σ x 2 = 0,1 … Σ x n = 0, 1 p φ (z, x 2, …, x n ) prover sends: k z for all z  F q verifier: –checks that k 0 + k 1 = k –sends random z  F q continue with proof that k z = Σ x 2 = 0,1 … Σ x n = 0, 1 p φ (z, x 2, …, x n ) at end: verifier checks for itself

20. 20 The power of IP Prover wishes to prove: k = Σ x 1 = 0, 1 Σ x 2 = 0,1 … Σ x n = 0, 1 p φ (x 1, x 2, …, x n ) Define: k z = Σ x 2 = 0,1 … Σ x n = 0, 1 p φ (z, x 2, …, x n ) a problem: can’t send k z for all z  F q solution: send the polynomial: p 1 (X)= Σ x 2 = 0,1 … Σ x n = 0, 1 p φ (X, x 2, …, x n ) –recall degree d  |φ|

21. 21 The actual protocol ProverVerifier input: ( φ, k ) p 1 (0)+p 1 (1)=k? pick random z 1 in F q p 1 (x) = Σ x 2, …, x n  {0,1} p φ ( x, x 2, …, x n ) p 1 (x) z1z1 p 2 (x) = Σ x 3, …, x n  {0,1} p φ (z 1, x, x 3, …, x n ) p 2 (x) z2z2 p 2 (0)+p 2 (1)=p 1 (z 1 )? pick random z 2 in F q p 3 (x) = Σ x 4, …, x n  {0,1} p φ (z 1, z 2, x, x 4 …, x n ) p 3 (0)+p 3 (1)=p 2 (z 2 )? pick random z 3 in F q p 3 (x) p n (x) p n (0)+p n (1)=p n-1 (z n-1 )? pick random z n in F q. p n (z n ) = p φ (z 1, z 2, …, z n )?

22. 22 Analysis of protocol Completeness: –if (φ, k)  L then honest prover on previous slide will always cause verifier to accept

23. 23 Analysis of protocol Soundness: –let p i (x) be the correct polynomials –let p i *(x) be the polynomials sent by (cheating) prover –(φ, k)  L  p 1 (0) + p 1 (1) ≠ k –either p 1 *(0) + p 1 *(1) ≠ k (and V rejects) –or p 1 * ≠ p 1  Pr z 1 [p 1 *(z 1 ) = p 1 (z 1 )]  d/q  |φ|/2 n –assume (p i+1 (0)+p i+1 (1)= ) p i (z i ) ≠ p i *(z i ) –either p i+1 *(0) + p i+1 *(1) ≠ p i *(z i ) (and V rejects) –or p i+1 * ≠ p i+1  Pr z i+1 [p i+1 *(z i+1 ) = p i+1 (z i+1 )]  |φ|/2 n

24. 24 Analysis of protocol Soundness (continued): –if verifier does not reject, there must be some i for which: p i * ≠ p i and yet p i *(z i ) = p i (z i ) –for each i, probability is  |φ|/2 n –union bound: probability that there exists an i for which the bad event occurs is  n|φ|/2 n  poly(n)/2 n << 1/3

25. 25 IP = PSPACE Theorem: (Shamir) IP = PSPACE –Note: IP  PSPACE enumerate all possible interactions, explicitly calculate acceptance probability interaction extremely powerful ! An implication : you can interact with master player of Generalized Geography and determine if she can win from the current configuration even if you do not have the power to compute optimal moves!

26. 26 IP = PSPACE need to prove PSPACE  IP –use same type of protocol as for coNP –some modifications needed

27. 27 IP = PSPACE protocol for QSAT –arithmetization step produces arithmetic expression p φ : ( 9 x i ) φ  Σ x i = 0, 1 p φ ( 8 x i ) φ  ∏ x i = 0, 1 p φ –start with QSAT formula in special form (“simple”) no occurrence of x i separated by more than one “ 8 ” from point of quantification

28. 28 IP = PSPACE –quantified Boolean expression φ is true if and only if p φ > 0 –Problem: ∏’s may cause p φ > 2 2 |φ| –Solution: evaluate mod 2 n  q  2 3n –prover sends “good” q in first round “good” q is one for which p φ mod q > 0 –Claim: good q exists # primes in range is at least 2 n

29. 29 The QSAT protocol ProverVerifier input: φ p 1 (0)+p 1 (1) = k? or p 1 (0)p 1 (1) = k? pick random z 1 in F q p 1 (x): remove outer Σ or ∏ from p φ k, q, p 1 (x) z1z1 p 2 (x): remove outer Σ or ∏ from p φ [x 1  z 1 ] p 2 (x) z2z2 p 2 (0)+p 2 (1)=p 1 (z 1 )? or p 2 (0)p 2 (1) = p 1 (z 1 )? pick random z 2 in F q p 3 (x): remove outer Σ or ∏ from p φ [x 1  z 1, x 2  z 2 ] p 3 (x) p n (x) p n (0)+p n (1)=p n-1 (z n-1 )? or p n (0)p n (1) = p n-1 (z n-1 )? pick random z n in F q p n (z n ) = p φ [x 1  z 1,…, x n  z n ].

30. 30 Analysis of the QSAT protocol Completeness: –if φ  QSAT then honest prover on previous slide will always cause verifier to accept

31. 31 Analysis of the QSAT protocol Soundness: –let p i (x) be the correct polynomials –let p i *(x) be the polynomials sent by (cheating) prover –φ  QSAT  0 = p 1 (0) +/x p 1 (1) ≠ k –either p 1 *(0) +/x p 1 *(1) ≠ k (and V rejects) –or p 1 * ≠ p 1  Pr z 1 [p 1 *(z 1 ) = p 1 (z 1 )]  2|φ|/2 n –assume (p i+1 (0) +/x p i+1 (1)=) p i (z i ) ≠ p i *(z i ) –either p i+1 *(0) +/x p i+1 *(1) ≠ p i *(z i ) (and V rejects) –or p i+1 * ≠ p i+1  Pr z i+1 [p i+1 *(z i+1 ) = p i+1 (z i+1 )]  2|φ|/2 n φ is “simple”

32. 32 Analysis of protocol Soundness (continued): –if verifier does not reject, there must be some i for which: p i * ≠ p i and yet p i *(z i ) = p i (z i ) –for each i, probability is  2|φ|/2 n –union bound: probability that there exists an i for which the bad event occurs is  2n|φ|/2 n  poly(n)/2 n << 1/3 Conclude: QSAT is in IP

33. 33 Example Papadimitriou – pp. 475-480 φ = 8 x 9 y(x  y)  8 z((x  z)  (y  z))  9 w(z  (y  w)) p φ = ∏ x=0,1 Σ y=0,1 [(x + y) * ∏ z=0,1 [(xz + y(1-z)) + Σ w=0,1 (z + y(1-w))]] (p φ = 96 but V doesn’t know that yet !)

34. 34 Example p φ = ∏ x=0,1 Σ y=0,1 [(x + y) * ∏ z=0,1 [(xz + y(1-z)) + Σ w=0,1 (z + y(1-w))]] Round 1: (prover claims p φ > 0) –prover sends q = 13; claims p φ = 96 mod 13 = 5; sends k = 5 –prover removes outermost “∏”; sends p 1 (x) = 2x 2 + 8x + 6 –verifier checks: p 1 (0)p 1 (1) = (6)(16) = 96  5 (mod 13) –verifier picks randomly: z 1 = 9

35. 35 Example φ = 8 x 9 y(x  y)  8 z((x  z)  (y  z))  9 w(z  (y  w)) p φ = ∏ x=0,1 Σ y=0,1 [(x + y) * ∏ z=0,1 [(xz + y(1-z)) + Σ w=0,1 (z + y(1-w))]] p φ [x  9] = Σ y=0,1 [(9 + y) * ∏ z=0,1 [(9z + y(1-z)) + Σ w=0,1 (z + y(1-w))]]

36. 36 Example p 1 (9) = Σ y=0,1 [(9 + y) * ∏ z=0,1 [(9z + y(1-z)) + Σ w=0,1 (z + y(1-w))]] Round 2: (prover claims this = 6) –prover removes outermost “Σ”; sends p 2 (y) = 2y 3 + y 2 + 3y –verifier checks: p 2 (0) + p 2 (1) = 0 + 6 = 6  6 (mod 13) –verifier picks randomly: z 2 = 3

37. 37 Example φ = 8 x 9 y(x  y)  8 z((x  z)  (y  z))  9 w(z  (y  w)) p φ = ∏ x=0,1 Σ y=0,1 [(x + y) * ∏ z=0,1 [(xz + y(1-z)) + Σ w=0,1 (z + y(1-w))]] p φ [x  9, y  3] = [(9 + 3) * ∏ z=0,1 [(9z + 3(1-z)) + Σ w=0,1 (z + 3(1-w))]]

38. 38 Example p 2 (3) = [(9 + 3) * ∏ z=0,1 [(9z + 3(1-z)) + Σ w=0,1 (z + 3(1-w))]] Round 3: (prover claims this = 7) –everyone agrees expression = 12*(…) –prover removes outermost “∏”; sends p 3 (z) = 8z + 6 –verifier checks: p 3 (0) * p 3 (1) = (6)(14) = 84; 12*84  7 (mod 13) –verifier picks randomly: z 3 = 7

39. 39 Example φ = 8 x 9 y(x  y)  8 z((x  z)  (y  z))  9 w(z  (y  w)) p φ = ∏ x=0,1 Σ y=0,1 [(x + y) * ∏ z=0,1 [(xz + y(1-z)) + Σ w=0,1 (z + y(1-w))]] p φ [x  9, y  3, z  7] = 12 * [(9*7 + 3(1-7)) + Σ w=0,1 (7 + 3(1-w))]

40. 40 Example 12*p 3 (7) = 12 * [(9*7 + 3(1-7)) + Σ w=0,1 (7 + 3(1-w))] Round 4: (prover claims = 12*10) –everyone agrees expression = 12*[6+(…)] –prover removes outermost “Σ”; sends p 4 (w) = 10w + 10 –verifier checks: p 4 (0)+p 4 (1) = 10 + 20 = 30; 12*[6+30]  12*10 (mod 13) –verifier picks randomly: z 4 = 2 –Final check: 12*[(9*7+3(1-7))+(7+3(1-2))] = 12*[6+p 4 (2)] = 12*[6+30]

41. 41 Arthur-Merlin Games IP permits verifier to keep coin-flips private –necessary feature? –GNI protocol breaks without it Arthur-Merlin game: interactive protocol in which coin-flips are public –Arthur (verifier) may as well just send results of coin-flips and ask Merlin (prover) to perform any computation Arthur would have done

42. 42 Arthur-Merlin Games Clearly Arthur-Merlin  IP –“private coins are at least as powerful as public coins” Proof that IP = PSPACE actually shows PSPACE  Arthur-Merlin  IP  PSPACE –“public coins are at least as powerful as private coins” !

43. 43 Arthur-Merlin Games Delimiting # of rounds: –AM[k] = Arthur-Merlin game with k rounds, Arthur (verifier) goes first –MA[k] = Arthur-Merlin game with k rounds, Merlin (prover) goes first Theorem: AM[k] (MA[k]) equals AM[k] (MA[k]) with perfect completeness. – i.e., x  L implies accept with probability 1 –proof on problem set

44. 44 Arthur-Merlin Games Theorem: for all constant k  2 AM[k] = AM[2]. Proof: –we show MA[2]  AM[2] –implies can move all of Arthur’s messages to beginning of interaction: AMAMAM…AM = AAMMAM…AM … = AAA…AMMM…M

45. 45 Arthur-Merlin Games Proof (continued): –given L  MA[2] x  L  9 m Pr r [(x, m, r)  R] = 1  Pr r [ 9 m (x, m, r)  R] = 1 x  L  8 m Pr r [(x, m, r)  R]  ε  Pr r [ 9 m (x, m, r)  R]  2 |m| ε –by repeating t times with independent random strings r, can make error ε < 2 -t –set t = m+1 to get 2 |m| ε < ½. order reversed

46. 46 MA and AM Two important classes: –MA = MA[2] –AM = AM[2] definitions without reference to interaction: –L  MA iff 9 poly-time language R x  L  9 m Pr r [(x, m, r)  R] = 1 x  L  8 m Pr r [(x, m, r)  R]  ½ –L  AM iff 9 poly-time language R x  L  Pr r [ 9 m (x, m, r)  R] = 1 x  L  Pr r [ 9 m (x, m, r)  R]  ½

47. 47 MA and AM L  AM iff 9 poly-time language R x  L  Pr r [ 9 m (x, m, r)  R] = 1 x  L  Pr r [ 9 m (x, m, r)  R]  ½ Relation to other complexity classes: –both contain NP (can elect to not use randomness) –both contained in ∏ 2. L  ∏ 2 iff 9 R  P for which: x  L  Pr r [ 9 m (x, m, r)  R] = 1 x  L  Pr r [ 9 m (x, m, r)  R] < 1 –so clear that AM  ∏ 2 –know that MA  AM

48. 48 MA and AM P NP coNP Σ2Σ2 Π2Π2 AMcoAM MAcoMA

49. 49 MA and AM Theorem: coNP  AM  PH = AM. Proof: –suffices to show Σ 2  AM (and use AM  Π 2 ) –L  Σ 2 iff 9 poly-time language R x  L  9 y 8 z (x, y, z)  R x  L  8 y 9 z (x, y, z)  R –Merlin sends y –1 AM exchange decides coNP query: 8 z (x, y, z)  R ? –3 rounds; in AM

50. 50 MA and AM We know Arthur-Merlin = IP. –“public coins = private coins” Theorem (GS): IP[k]  AM[O(k)] –stronger result –implies for all constant k  2, IP[k] = AM[O(k)] = AM[2] So, GNI  IP[2] = AM

51. 51 Back to Graph Isomorphism The payoff: –not known if GI is NP-complete. –previous Theorems: if GI is NP-complete then PH = AM –unlikely! –Proof: GI NP-complete  GNI coNP- complete  coNP  AM  PH = AM

52. Derandomization revisited L  MA iff 9 poly-time language R x  L  9 m Pr r [(x, m, r)  R] = 1 x  L  8 m Pr r [(x, m, r)  R]  ½ Recall PRGs: –for all circuits C of size at most s : |Pr y [C(y) = 1] – Pr z [C(G(z)) = 1]| ≤ ε 52 NP AM MA seedoutput string G t bitsu bits

53. Using PRGs for MA L  MA iff 9 poly-time language R x  L  9 m Pr r [(x, m, r)  R] = 1 x  L  8 m Pr r [(x, m, r)  R]  ½ produce poly-size circuit C such that C(x, m, r) = 1, (x,m,r) 2 R for each x, m can hardwire to get C x,m 9 m Pr y [C x,m (y) = 1] = 1 (“yes”) 8 m Pr y [C x,m (y) = 1] ≤ 1/2 (“no”) 53

54. Using PRGs for MA can compute Pr z [C x,m (G(z)) = 1] exactly –evaluate C x (G(z)) on every seed z  {0,1} t –running time (O(|C x,m |)+(time for G))2 t x  L  9 m [ Pr z [C x,m (G(z)) = 1] = 1 ] x  L  8 m [ Pr z [C x,m (G(z)) = 1]  ½ + ² ] – L 2 NP if PRG with t = O(log n), ² < 1/4 Theorem: E requires exponential size circuits  MA = NP. 54 poly(n) poly-time

55. 55 MA and AM P NP coNP Σ2Σ2 Π2Π2 AMcoAM MAcoMA (under a hardness assumption)

56. 56 MA and AM P NP coNP Σ2Σ2 Π2Π2 AMcoAM MA = = coMA (under a hardness assumption) What about AM, coAM?