Proving observational equivalence with ProVerif

Size: px

Start display at page:

Download "Proving observational equivalence with ProVerif"

Homer Waters
6 years ago
Views:

1 Proving observational equivalence with ProVerif Bruno Blanchet INRIA Paris-Rocquencourt based on joint work with Martín Abadi and Cédric Fournet and with Vincent Cheval June 2015 Bruno Blanchet (INRIA) Observational equivalence with ProVerif June / 40

2 Introduction Analysis of cryptographic protocols: Powerful automatic tools for proving properties on behaviors (traces) of protocols (secrecy of keys, correspondences). For instance, ProVerif was initially designed to prove trace properties. Many important properties can be formalized as process equivalences, not as properties on behaviors: secrecy of a boolean x in P(x): P(true) P(false) the process P implements an ideal specification Q: P Q Equivalences are usually proved by difficult, long manual proofs. Already much research on this topic, using in particular sophisticated bisimulation techniques (e.g., Boreale et al). Some recent tools for a bounded number of sessions (APTE, Akiss) Bruno Blanchet (INRIA) Observational equivalence with ProVerif June / 40

3 Selected work relying on equivalences January 25, 1998 SRC Research Report 149 A Calculus for Cryptographic Protocols The Spi Calculus Martín Abadi and Andrew D. Gordon Bruno Blanchet (INRIA) Observational equivalence with ProVerif June / 40

4 Selected work relying on equivalences Secrecy by Typing in Security Protocols Martín Abadi Systems Research Center Compaq December 8, 1998 Abstract We develop principles and rules for achieving secrecy properties in security protocols. Our approach is based on traditional classification techniques, and extends those techniques to handle concurrent processes that use shared-key cryptography. The rules have the form of typing rules for a basic concurrent language with cryptographic primitives, the spi calculus. They guarantee that, if a protocol typechecks, then it does not leak its secret inputs. Bruno Blanchet (INRIA) Observational equivalence with ProVerif June / 40

5 Selected work relying on equivalences ÅÓ Ð Î ÐÙ Æ Û Æ Ñ Ò Ë ÙÖ ÓÑÑÙÒ Ø ÓÒ Å ÖØ Ò ÐÐ Ä Ê Ö ÄÙ ÒØ Ì ÒÓÐÓ Ö ÓÙÖÒ Ø Å ÖÓ Ó Ø Ê Ö ØÖ Ø Ï ØÙ Ý Ø ÒØ Ö Ø ÓÒ Ó Ø Ò Û ÓÒ ØÖÙØ Û Ø Ö ÙØ ÓÑÑÓÒ ÓÖÑ Ó Ö Ø¹ÓÖ Öµ ÓÑÑÙÒ Ø ÓÒº Ì Ò¹ Ø Ö Ø ÓÒ ÖÙ Ð Ò ÙÖ ØÝ ÔÖÓØÓÓÐ Û Ö Ø Ñ Ò ÑÓØ Ú Ø Ò Ü ÑÔÐ ÓÖ ÓÙÖ ÛÓÖ Ø Ð Ó ÔÔ Ö Ò ÓØ Ö ÔÖÓ Ö ÑÑ Ò ¹Ð Ò Ù ÓÒØ ÜØ º ËÔ ÐÐÝ Û ÒØÖÓ Ù ÑÔÐ Ò Ö Ð ÜØ Ò ÓÒ Ó Ø Ô ÐÙÐÙ Û Ø Ú ÐÙ Ô ¹ Ò ÔÖ Ñ Ø Ú ÙÒØ ÓÒ Ò ÕÙ Ø ÓÒ ÑÓÒ Ø ÖÑ º Ï Ú ÐÓÔ Ñ ÒØ Ò ÔÖÓÓ Ø Ò ÕÙ ÓÖ Ø ÜØ Ò Ð Ò Ù Ò ÔÔÐÝ Ø Ñ Ò Ö ÓÒ Ò ÓÙØ ÓÑ ÙÖ ØÝ ÔÖÓØÓÓÐ º Ì ÆÙÐØ Ö Ó Ø Ò ÖÙÑÚ ÒØ Ø ÖÓÙ ÓÒ¹Ø ¹ Ý ÜØ Ò ÓÒ º Ì ÜØ Ò ÓÒ Ö Ò ÖÓÑ ÕÙ ÔÙÒØ ÓÖ Ø Ò ÜØ Ü ÑÔÐ Ð Ø³ ÔÖ Ø Ò Ø Ø Û Ú Ø ØÝÔ Ó ÒØ Ö µ ØÓ Ø Ð ÓÖ ÓÙ Ú ÐÓÔÑ ÒØ Ó Ò Û ÐÙÐ Ù Ø Ô ÐÙÐÙ Ò Ø Ú Ö ÒØ º Ò Ö ÐÐÝ Ø ÜØ Ò¹ ÓÒ Ö Ò Ù ÐÓ Ö ØÓ Ö Ð Ø ÔÖÓ Ö ÑÑ Ò Ð Ò Ù ÓÖ ÑÓ Ð Ò Ð Ò Ù Ø Ø ÒÓØ ÐÛ Ý Ø Ò º ÐØ ÓÙ Ñ ÒÝ Ó Ø Ö ÙÐØ Ò ÐÙÐ Ö Ó Ò ÔÓÓÖÐÝ ÙÒ Ö ØÓÓ ÓØ Ö Ö ÖÓ Ù Ø Ò ÙÒ ÓÖÑ ÒÓÙ ØÓ Ú Ö Ø ÓÖÝ Ò Ú Ö ØÝ Ó ÔÔÐ Ø ÓÒ º ÁÒ Ô ÖØ ÙÐ Ö ÑÔÙÖ ÜØ Ò ÓÒ Ó Ø Ð Ñ ÐÙÐÙ Û Ø ÙÒØ ÓÒ ÝÑ ÓÐ Ò Û Ø ÕÙ Ø ÓÒ ÑÓÒ Ø ÖÑ ÐØ ÖÙÐ µ Ú Ò Ú ÐÓÔ Ý Ø Ñ Ø ÐÐÝ Û Ø ÓÒ Ö¹ ½ ÓÖ ÑÔÙÖ ØÝ Ð Ù º Ë Ñ Ð ÖÐÝ ÑÔÙÖ Ú Ö ÓÒ Ó Ë Ò ËÈ Û Ø Ú ÐÙ ¹Ô Ò Ö ÒÓØ ÐÛ Ý Ô ÙØ Ó Ø Ò Ò Ø Ò ÈÙÖ ØÝ Ó Ø Ò ÓÑ ÓÖ ÓÒÚ Ò Ò Ò Ú Ò ÓÖ Ø ÙÐÒ Ò Ø Ð Ñ ÐÙÐÙ Ø Ô ÐÙÐÙ Ò ÓÒÚ Ò ÒØ ½ º ÁÒ Ø Ô Ô Ö Û ÒØÖÓ Ù ØÙ Ý Ò Ù Ò Ò ÐÓ¹ BrunoÓØ Ö Blanchet ÓÙÒ Ø ÓÒ Ð (INRIA) ÔÖÓ Ö ÑÑ Ò Ð Ò Ù º Observational ÓÖ Ü ÑÔÐ equivalence ÓÙ with ÙÒ ÓÖÑ ProVerif ÜØ Ò ÓÒ Ó Ø Ô ÐÙÐÙ Û June Û ÐÐ 2015 Ø 3 / 40

6 Equivalences as properties of behaviors (1) Goal: extend ProVerif to the proof of process equivalences. We focus on equivalences between processes that differ only by the terms they contain, e.g., P(true) P(false). Many interesting equivalences fall into this category. We introduce biprocesses to represent pairs of processes that differ only by the terms they contain. P(true) and P(false) are variants of a biprocess P(diff[true, false]). The variants give a different interpretation to diff[true, false], true for the first variant, false for the second one. Bruno Blanchet (INRIA) Observational equivalence with ProVerif June / 40

7 Equivalences as properties of behaviors (2) We introduce a new operational semantics for biprocesses: A biprocess reduces when both variants reduce in the same way and after reduction, they still differ only by terms (so can be written using diff). We establish P(true) P(false) by reasoning on behaviors of P(diff[true, false]): If, for all reachable configurations, both variants reduce in the same way, then we have equivalence. Bruno Blanchet (INRIA) Observational equivalence with ProVerif June / 40

8 Overview of the verification method Protocol: biprocess Pi calculus + cryptography Signature: Rewrite rules + equations Automatic translator Horn clauses Resolution with selection bad is not derivable Equivalence is true bad is derivable Equivalence may be false Bruno Blanchet (INRIA) Observational equivalence with ProVerif June / 40

9 The process calculus Dialect of the applied pi calculus [Abadi, Fournet, POPL 01]. M,N ::= terms x,y,z variable a,b,c,k,s name f(m 1,...,M n ) constructor application D ::= term evaluations M term fail special failure value h(d 1,...,D n ) function evaluation P, Q, R ::= processes M(x).P input M N.P output let x = D in P else Q term evaluation 0 P Q!P (νa)p if M then P else Q encoded as a term evaluation. Bruno Blanchet (INRIA) Observational equivalence with ProVerif June / 40

10 Representation of cryptographic primitives Two possible representations: When success/failure is visible: destructors with rewrite rules constructor sencrypt destructor sdecrypt(sencrypt(x, y), y) x otherwise sdecrypt(u,v) fail The else branch of the term evaluation is executed when D fails, that is, D evaluates to fail. When success/failure is not visible: equations sdecrypt(sencrypt(x,y),y) = x sencrypt(sdecrypt(x,y),y) = x Bruno Blanchet (INRIA) Observational equivalence with ProVerif June / 40

11 Representation of cryptographic primitives Two possible representations: When success/failure is visible: destructors with rewrite rules constructor sencrypt destructor sdecrypt(sencrypt(x, y), y) x otherwise sdecrypt(u,v) fail The else branch of the term evaluation is executed when D fails, that is, D evaluates to fail. When success/failure is not visible: equations sdecrypt(sencrypt(x,y),y) = x sencrypt(sdecrypt(x,y),y) = x Bruno Blanchet (INRIA) Observational equivalence with ProVerif June / 40

12 Semantics D M when the term evaluation D evaluates to M. D fail when the term evaluation D fails. Uses rewrite rules of destructors and equations. transforms processes so that reduction rules can be applied. Main reduction rules: N M.Q N (x).p Q P{M/x} if Σ N = N (Red I/O) let x = D in P else Q P{M/x} (Red Fun 1) if D M let x = D in P else Q Q (Red Fun 2) if D fail Bruno Blanchet (INRIA) Observational equivalence with ProVerif June / 40

13 Observational equivalences and biprocesses Two processes P and Q are observationally equivalent (P Q) when the adversary cannot distinguish them. A biprocess P is a process with diff. fst(p) = the process obtained by replacing diff[m,m ] with M. snd(p) = the process obtained by replacing diff[m,m ] with M. P satisfies observational equivalence when fst(p) snd(p). Bruno Blanchet (INRIA) Observational equivalence with ProVerif June / 40

14 Semantics of biprocesses A biprocess reduces when both variants of the process reduce in the same way. N M.Q N (x).p Q P{M/x} if Σ fst(n) = fst(n ) and Σ snd(n) = snd(n ) (Red I/O) let x = D in P else Q P{diff[M 1,M 2 ]/x} (Red Fun 1) if fst(d) M 1 and snd(d) M 2 let x = D in P else Q Q (Red Fun 2) if fst(d) fail and snd(d) fail P Q fst(p) snd(p) fst(q) snd(q) Bruno Blanchet (INRIA) Observational equivalence with ProVerif June / 40

15 Proof of observational equivalence using biprocesses Let P 0 be a closed biprocess. If for all configurations P reachable from P 0 (in the presence of an adversary), both variants of P reduce in the same way, then P 0 satisfies observational equivalence. Bruno Blanchet (INRIA) Observational equivalence with ProVerif June / 40

16 Proof of observational equivalence using biprocesses Let P 0 be a closed biprocess. If for all configurations P reachable from P 0 (in the presence of an adversary), both variants of P reduce in the same way, then P 0 satisfies observational equivalence. An adversary is represented by a plain evaluation context (evaluation context without diff), so: If, for all plain evaluation contexts C and reductions C[P 0 ] P, both variants of P reduce in the same way, then P 0 satisfies observational equivalence. Bruno Blanchet (INRIA) Observational equivalence with ProVerif June / 40

17 Formalizing reduce in the same way The biprocess P is uniform when fst(p) Q 1 implies P Q for some biprocess Q with fst(q) Q 1, and symmetrically for snd(p) Q 2. P Q P Q fst(p) Q 1 fst(q) snd(p) snd(q) Q 2 If, for all plain evaluation contexts C and reductions C[P 0 ] P, the biprocess P is uniform, then P 0 satisfies observational equivalence. Bruno Blanchet (INRIA) Observational equivalence with ProVerif June / 40

18 Result [Blanchet, Abadi, Fournet, JLAP, 2008] Let P 0 be a closed biprocess. Suppose that, for all plain evaluation contexts C, all evaluation contexts C, and all reductions C[P 0 ] P, 1 the (Red I/O) rules apply in the same way on both variants. if P C [N M.Q N (x).r], then Σ fst(n) = fst(n ) if and only if Σ snd(n) = snd(n ), 2 the (Red Fun) rules apply in the same way on both variants. if P C [let x = D in Q else R], then fst(d) fail if and only if snd(d) fail. Then P 0 satisfies observational equivalence. Bruno Blanchet (INRIA) Observational equivalence with ProVerif June / 40

19 Application The previous result reduces the proof of observational equivalence to the proof of trace properties for biprocesses. We extend the Horn clause approach of ProVerif from processes to biprocesses, to prove these properties. Basically, each argument of the predicates is doubled, with one argument for each side. Bruno Blanchet (INRIA) Observational equivalence with ProVerif June / 40

20 Example: probabilistic encryption Probabilistic public-key encryption is modeled by an equation: dec(enc(x,pk(s),a),s) = x Without knowledge of the decryption key, ciphertexts appear to be unrelated to the plaintexts. Ciphertexts are indistinguishable from fresh names: satisfies equivalence. (νs)(c pk(s)!c (x).(νa)c diff[enc(x,pk(s),a),a] ) This equivalence can be proved using the previous result, and verified automatically by ProVerif. Bruno Blanchet (INRIA) Observational equivalence with ProVerif June / 40

21 Example: private authentication Simplified version of the private authentication protocol [Abadi, Fournet, TCS, 2004]. {N A,pk A } pkb {x,y} pkb y = pka? Bruno Blanchet (INRIA) Observational equivalence with ProVerif June / 40

22 Example: private authentication Simplified version of the private authentication protocol [Abadi, Fournet, TCS, 2004]. {N A,pk A } pkb {x,y} pkb y = pka? {x,n B,pk B } y yes Bruno Blanchet (INRIA) Observational equivalence with ProVerif June / 40

23 Example: private authentication Simplified version of the private authentication protocol [Abadi, Fournet, TCS, 2004]. {N A,pk A } pkb {x,y} pkb y = pka? {x,n B,pk B } y yes {N B } pkb no Bruno Blanchet (INRIA) Observational equivalence with ProVerif June / 40

24 Proving anonymity {x,y} pkb y = pka? {x,n B,pk B } y yes {N B } pkb no {x,y} pkb y = pka? {x,n B,pk B } y yes {N B } pkb no Caesar should not be able to know who Obelix wants to talk to. Bruno Blanchet (INRIA) Observational equivalence with ProVerif June / 40

25 False attack {N A,pk A } pkb {x,y} pkb y = pka? {x,n B,pk B } y yes {N B } pkb no {N A,pk A } pkb {x,y} pkb y = pka? {x,n B,pk B } y yes {N B } pkb no Bruno Blanchet (INRIA) Observational equivalence with ProVerif June / 40

26 False attack {N A,pk A } pkb {x,y} pkb y = pka? {x,n B,pk B } y yes {N A,pk A } pkb {x,y} pkb y = pka? {x,n B,pk B } y yes {N B } pkb no Bruno Blanchet (INRIA) Observational equivalence with ProVerif June / 40

27 False attack {N A,pk A } pkb {x,y} pkb y = pka? {x,n B,pk B } y yes {N A,pk A } pkb {x,y} pkb y = pka? {N B } pkb no The test takes a different branch: the proof of equivalence fails. Bruno Blanchet (INRIA) Observational equivalence with ProVerif June / 40

28 Demo File private auth falseattack.pv Bruno Blanchet (INRIA) Observational equivalence with ProVerif June / 40

29 Solution: encode tests inside terms [Cheval, Blanchet, POST 13] The if-then-else destructor: ifthenelse(x,x,u,v) u otherwise ifthenelse(x,y,u,v) v Bruno Blanchet (INRIA) Observational equivalence with ProVerif June / 40

30 Solution: encode tests inside terms [Cheval, Blanchet, POST 13] The if-then-else destructor: ifthenelse(x,x,u,v) u otherwise ifthenelse(x,y,u,v) v {x,y} pkb y = pka? {x,n B,pk B } y yes {N B } pkb no Bruno Blanchet (INRIA) Observational equivalence with ProVerif June / 40

31 Solution: encode tests inside terms [Cheval, Blanchet, POST 13] The if-then-else destructor: ifthenelse(x,x,u,v) u otherwise ifthenelse(x,y,u,v) v {x,y} pkb M M = ifthenelse(y,pk A,{x,N B,pk B } y,{n B } pkb ) Bruno Blanchet (INRIA) Observational equivalence with ProVerif June / 40

32 Demo File private auth direct.pv Bruno Blanchet (INRIA) Observational equivalence with ProVerif June / 40

33 Merging two processes In order to prove that P and Q are observationally equivalent using this approach, we need to merge P and Q into a biprocess R: fst(r) P snd(r) Q Trivial when P and Q have exactly the same structure. There is a general merging: R = if diff[true,false] then P else Q does not allow ProVerif to prove observational equivalence. Do something more clever! Bruno Blanchet (INRIA) Observational equivalence with ProVerif June / 40

34 A related problem: simplifying biprocesses In the private authentication example, we moved a test from the process level to the term level. Doing the same on the process R = if diff[true,false] then P else Q could allow ProVerif to prove observational equivalence. Bruno Blanchet (INRIA) Observational equivalence with ProVerif June / 40

35 A related problem: simplifying biprocesses In the private authentication example, we moved a test from the process level to the term level. Doing the same on the process R = if diff[true,false] then P else Q could allow ProVerif to prove observational equivalence....but moving this test implies merging the structures of P and Q! Bruno Blanchet (INRIA) Observational equivalence with ProVerif June / 40

36 Merging algorithm Merge branches of biprocesses: Automatically transforms the private authencation example Allows merging P and Q by applying it to if diff[true,false] then P else Q Two steps: 1 Normalize the process (Try to reorganize processes that send the same messages so that they have the same syntactic form.) 2 Merge branches on the normalized process Bruno Blanchet (INRIA) Observational equivalence with ProVerif June / 40

37 Normalization algorithm 1 Make sure term evaluations always succeed New destructors: catchfail(x) x otherwise catchfail(fail) c fail not c fail (c fail ) false otherwise not c fail (x) true let x = D in P else Q let x = catchfail(d) in if not c fail (x) then P else Q (We allow destructors outside let; they can be encoded using an additional let.) 2 Remove unused restrictions and lets;!0 0 Bruno Blanchet (INRIA) Observational equivalence with ProVerif June / 40

38 Normalization algorithm 3 Move new, let, if before outputs M N.(νa)P (νa)m N.P M N.let x = D in P let x = D in M N.P M N.if M then P else Q if M then M N.P else M N.Q (Avoid capture of names and variables.) 4 Move new, let, if before parallel composition ((νa).p) Q (νa).(p Q) (let x = D in P) Q let x = D in (P Q) (if M then P else P ) Q if M then P Q else P Q 5 Group replicated processes inside a parallel composition!p P!P P!(P P ) Bruno Blanchet (INRIA) Observational equivalence with ProVerif June / 40

39 Normalization algorithm 6 Move new, let before if if M then (νa)p else Q (νa)if M then P else Q if M then let x = D in P else Q let x = D in if M then P else Q 7 Move new before let let x = D in (νa)p (νa)let x = D in 8 Move if (with its new/let) before replication!s ν s let if M then P else Q s ν s let if M then!s ν s let P else!s ν s let Q The equational theory is closed under one-to-one renaming, so the test M will always yield the same result independently of the considered fresh names. 9 Remove double replication!s let!s ν s let Q!s νs let s let Q Bruno Blanchet (INRIA) Observational equivalence with ProVerif June / 40

40 Grammar of normalized processes P ::= s ν s let T s ν consists of a sequence of (νa) s let consists of a sequence of let x = D in such that D always succeeds. T ::= Q if M then T 1 else T 2 decision tree test Q ::= R 1... R n S parallel composition (n 0) R ::= M(x).P M N.Q communication input output S ::= optional replicated process!s ν s let Q replicated process 0 no replicated process Bruno Blanchet (INRIA) Observational equivalence with ProVerif June / 40

41 Some explanation When two processes Q have the same structure, we will be able to merge them. The structure ignores the precise channels and messages of inputs and outputs and considers parallel composition up to associativity and commutativity. It seems difficult to go further with general syntactic rules. Bruno Blanchet (INRIA) Observational equivalence with ProVerif June / 40

42 Merging Merge leaves of decision trees then then C else then C else then C else C 1 else then C 2 else then C 3 else then Q P C 4 else Bruno Blanchet (INRIA) Observational equivalence with ProVerif June / 40

43 Merging Merge leaves of decision trees then then C then C else then else C else C 1 else then C 2 else then C 3 else then C 4 else If necessary, reorganize the decision tree so that the merged leaves are the two branches of an if. Q P Bruno Blanchet (INRIA) Observational equivalence with ProVerif June / 40

44 Merging inputs and outputs (R) New destructor ite(true,u,v) u otherwise ite(x,u,v) v Outputs then C else M N.Q M N.Q ite(c,m,m ) ite(c,n,n ). C then else Q Q Inputs then C else M(x).Q M (x ).Q ite(c,m,m )(x ). C then else Q{ x / x } Q { x / x } Bruno Blanchet (INRIA) Observational equivalence with ProVerif June / 40

45 Merging parallel compositions (Q) then C else R 1... R n S C R 1... R n S then else R 1 then... C R i else 1 R n then C R i else n S S i 1,...i n is a permutation of 1,...,n. Bruno Blanchet (INRIA) Observational equivalence with ProVerif June / 40

46 Merging replications (S) Nil Replication (naive version) then C else then C else!s ν s let Q!s νs let Q!s ν s νs let s let C then else Q Q!s ν1 s let1!s ν2 s let2 Q can actually be merged with!s νs let Q by merging Q and Q, because!s νs let Q!!s νs let Q. Separate the processes in s ν s let Q and s νs let Q into independent groups. Possibly add a replication in front of some of these groups. Bruno Blanchet (INRIA) Observational equivalence with ProVerif June / 40

47 Merging general processes (P) then C else s ν s let T s ν s νs let s s νs let let if C then T else T T Bruno Blanchet (INRIA) Observational equivalence with ProVerif June / 40

48 Properties of this algorithm Soundness: these transformations preserve observational equivalence Incomplete. Open question: partial completeness? All channels public Other conditions? Bruno Blanchet (INRIA) Observational equivalence with ProVerif June / 40

49 Demo File private auth biprocess.pv File private auth processes.pv Bruno Blanchet (INRIA) Observational equivalence with ProVerif June / 40

50 Conclusion ProVerif is the only tool that can prove process equivalences for an unbounded number of sessions. Restricted but useful class of equivalences: First restricted to processes that differ only by the terms they contain. Extended by automatic merging of processes. Tool available at Bruno Blanchet (INRIA) Observational equivalence with ProVerif June / 40

51 Back to the applied pi calculus ÅÓ Ð Î ÐÙ Æ Û Æ Ñ Ò Ë ÙÖ ÓÑÑÙÒ Ø ÓÒ Å ÖØ Ò ÐÐ Ä Ê Ö ÄÙ ÒØ Ì ÒÓÐÓ Ö ÓÙÖÒ Ø Å ÖÓ Ó Ø Ê Ö ØÖ Ø Ï ØÙ Ý Ø ÒØ Ö Ø ÓÒ Ó Ø Ò Û ÓÒ ØÖÙØ Û Ø Ö ÙØ ÓÑÑÓÒ ÓÖÑ Ó Ö Ø¹ÓÖ Öµ ÓÑÑÙÒ Ø ÓÒº Ì Ò¹ Ø Ö Ø ÓÒ ÖÙ Ð Ò ÙÖ ØÝ ÔÖÓØÓÓÐ Û Ö Ø Ñ Ò ÑÓØ Ú Ø Ò Ü ÑÔÐ ÓÖ ÓÙÖ ÛÓÖ Ø Ð Ó ÔÔ Ö Ò ÓØ Ö ÔÖÓ Ö ÑÑ Ò ¹Ð Ò Ù ÓÒØ ÜØ º ËÔ ÐÐÝ Û ÒØÖÓ Ù ÑÔÐ Ò Ö Ð ÜØ Ò ÓÒ Ó Ø Ô ÐÙÐÙ Û Ø Ú ÐÙ Ô ¹ Ò ÔÖ Ñ Ø Ú ÙÒØ ÓÒ Ò ÕÙ Ø ÓÒ ÑÓÒ Ø ÖÑ º Ï Ú ÐÓÔ Ñ ÒØ Ò ÔÖÓÓ Ø Ò ÕÙ ÓÖ Ø ÜØ Ò Ð Ò Ù Ò ÔÔÐÝ Ø Ñ Ò Ö ÓÒ Ò ÓÙØ ÓÑ ÙÖ ØÝ ÔÖÓØÓÓÐ º Ì ÆÙÐØ Ö Ó Ø Ò ÖÙÑÚ ÒØ Ø ÖÓÙ ÓÒ¹Ø ¹ Ý ÜØ Ò ÓÒ º Ì ÜØ Ò ÓÒ Ö Ò ÖÓÑ ÕÙ ÔÙÒØ ÓÖ Ø Ò ÜØ Ü ÑÔÐ Ð Ø³ ÔÖ Ø Ò Ø Ø Û Ú Ø ØÝÔ Ó ÒØ Ö µ ØÓ Ø Ð ÓÖ ÓÙ Ú ÐÓÔÑ ÒØ Ó Ò Û ÐÙÐ Ù Ø Ô ÐÙÐÙ Ò Ø Ú Ö ÒØ º Ò Ö ÐÐÝ Ø ÜØ Ò¹ ÓÒ Ö Ò Ù ÐÓ Ö ØÓ Ö Ð Ø ÔÖÓ Ö ÑÑ Ò Ð Ò Ù ÓÖ ÑÓ Ð Ò Ð Ò Ù Ø Ø ÒÓØ ÐÛ Ý Ø Ò º ÐØ ÓÙ Ñ ÒÝ Ó Ø Ö ÙÐØ Ò ÐÙÐ Ö Ó Ò ÔÓÓÖÐÝ ÙÒ Ö ØÓÓ ÓØ Ö Ö ÖÓ Ù Ø Ò ÙÒ ÓÖÑ ÒÓÙ ØÓ Ú Ö Ø ÓÖÝ Ò Ú Ö ØÝ Ó ÔÔÐ Ø ÓÒ º ÁÒ Ô ÖØ ÙÐ Ö ÑÔÙÖ ÜØ Ò ÓÒ Ó Ø Ð Ñ ÐÙÐÙ Û Ø ÙÒØ ÓÒ ÝÑ ÓÐ Ò Û Ø ÕÙ Ø ÓÒ ÑÓÒ Ø ÖÑ ÐØ ÖÙÐ µ Ú Ò Ú ÐÓÔ Ý Ø Ñ Ø ÐÐÝ Û Ø ÓÒ Ö¹ ½ ÓÖ ÑÔÙÖ ØÝ Ð Ù º Ë Ñ Ð ÖÐÝ ÑÔÙÖ Ú Ö ÓÒ Ó Ë Ò ËÈ Û Ø Ú ÐÙ ¹Ô Ò Ö ÒÓØ ÐÛ Ý Ô ÙØ Ó Ø Ò Ò Ø Ò ÈÙÖ ØÝ Ó Ø Ò ÓÑ ÓÖ ÓÒÚ Ò Ò Ò Ú Ò ÓÖ Ø ÙÐÒ Ò Ø Ð Ñ ÐÙÐÙ Ø Ô ÐÙÐÙ Ò ÓÒÚ Ò ÒØ ½ º ÁÒ Ø Ô Ô Ö Û ÒØÖÓ Ù ØÙ Ý Ò Ù Ò Ò ÐÓ¹ BrunoÓØ Ö Blanchet ÓÙÒ Ø ÓÒ Ð (INRIA) ÔÖÓ Ö ÑÑ Ò Ð Ò Ù º Observational ÓÖ Ü ÑÔÐ equivalence ÓÙ with ÙÒ ÓÖÑ ProVerif ÜØ Ò ÓÒ Ó Ø Ô ÐÙÐÙ Û June Û 2015 ÐÐ Ø 39 / 40

52 Advertisement With Martín Abadi and Cédric Fournet, we recently revisited their applied pi calculus paper [POPL 01]: Minor changes to the language to make it closer to ProVerif Detailed proofs of all results Revised examples New example on indifferentiability Bruno Blanchet (INRIA) Observational equivalence with ProVerif June / 40

Proving More Observational Equivalences with ProVerif

Proving More Observational Equivalences with ProVerif Vincent Cheval 1 and Bruno Blanchet 2 1 LSV, ENS Cachan & CNRS & INRIA Saclay Île-de-France, France cheval@lsv.ens-cachan.fr 2 INRIA Paris-Rocquencourt,