Presentation is loading. Please wait.

Presentation is loading. Please wait.

Analysis of optimistic multi-party contract signing Rohit Chadha 1,2, Steve Kremer 3,4, Andre Scedrov 1 1 University of Pennsylvania 2 University of Sussex.

Similar presentations


Presentation on theme: "Analysis of optimistic multi-party contract signing Rohit Chadha 1,2, Steve Kremer 3,4, Andre Scedrov 1 1 University of Pennsylvania 2 University of Sussex."— Presentation transcript:

1 Analysis of optimistic multi-party contract signing Rohit Chadha 1,2, Steve Kremer 3,4, Andre Scedrov 1 1 University of Pennsylvania 2 University of Sussex 3 Université Libre de Bruxelles 4 Birmingham University

2 Digital Contract signing Use digital signatures to sign a contract over a network –Contract text is already agreed upon Special instance of fair exchange protocols Important issue for secure electronic commerce Naive 2-party example : A  B : Sig A (contract) B  A : Sig B (contract)

3 Digital Contract signing Use digital signatures to sign a contract over a network Special instance of fair exchange protocols Important issue for secure electronic commerce Naive 2-party example : A  B : Sig A (contract) B  A :  Bob may be malicious and not send his signature Asymmetry : someone must be the first to send his signature

4 Properties of Contract Signing Fairness –If A can get B’s signature, then B can get A’s signature, and vice-versa Timeliness –Avoids that a participant gets stuck Advantage –A participant has an advantage if it has a strategy to complete the exchange and it has a strategy to abort the exchange Abuse-freeness (provable advantage) –Avoids that a participant can prove to an external party that it has the power to choose the outcome of the protocol

5 Evolution of contract signing Randomized protocols Trusted Party intervenes Use trusted party as a delivery authority May cause a bottleneck … Trusted Party intervenes only in case of problem (optimistic approach) More complex, and more error-prone … In 1980, Even & Yacobi showed that no fair deterministic contract signing protocol exists without the participation of a trusted party.

6 Formal methods & contract signing [Shmatikov, Mitchell, 2000] –Model-checker Murphi –invariant checking [Chadha, Kanovich, Scedrov, 2001] –Specification in MSR –inductive proofs [Kremer, Raskin, 2002] –Model-checker Mocha –ATL (temporal logic with game semantics) [Chadha, Mitchell, Scedrov, Shmatikov 2003] –general results (protocol independent) on advantage  Only 2-party contract signing protocols have been studied

7 Topologies Unlike for 2-party protocols, the different instances of fair exchange protocols differ significantly in the multi-party case 1-to-many non-repudiation and certified e-mail ring topology barter full graph contract signing 1 n 3 2... 1 n 3 2 1 n 32 Contract signing requires the most complicated protocols

8 Multi-party contract signing n participants want to sign a contract Properties for a honest participant must hold against any coalition of dishonest participants, i.e., against up to n-1 dishonest participants Every participant must receive the signature of all other participants (topology is a full graph)

9 Multi-party protocols Astonishingly few so far [Asokan, Baum-Waidner, Schunter, Waidner, T.R. 1998] Optimistic synchronous multi-party contract signing [Baum-Waidner, Waidner, T.R. 1998 & ICALP 2000] Optimistic asynchronous multi-party contract signing [Garay, MacKenzie, DISC 1999] Optimistic asynchronous multi-party contract signing [Baum-Waidner, Waidner, ICALP 2001] Optimistic asynchronous multi-party contract signing with reduced number of rounds

10 Protocol model All participants are players 2 versions of each player described using guarded commands –honest : follow the protocol –dishonest : may send messages out of order and continue the main protocol after contacting the trusted party Messages are immediately available for reading Only structural flaws are considered –no modelling of the cryptographic primitives Mocha cannot handle parametric specifications –Small C++ programs for the GM protocol and the BW protocol, that generate the Mocha specification for a given number of participants

11 The model-checker Mocha Guarded commands describing the protocol ATL formula Mocha ATSModel-Checking YESNO C++ program

12 The BW protocol [Baum, Waidner, ICALP 2000] Rather simple protocol, with symmetric behaviour of each participant T can overturn aborts We used Mocha to verify fairness for n=2,…,5, but no flaw was found The basic protocol does not aim to provide abuse-freeness Non-standard definition of contract –a special protocol for verifying the validity of a contract is defined

13 GM protocol [Garay, MacKenzie, DISC 1999] Recursive description of the protocol The protocol is divided into n levels –In each protocol level specific promises are used –Promises are implemented using private contract signatures (convertible designated verifier signatures) The i-level protocol is triggered when P i receives 1-level promises from P i+1 through P n In i-level protocol participants P i through P 1 exchange i-level promises –They agree on the contract with promises (not signatures) P i through P 1 close higher level protocols After the n-level protocol actual signatures are exchanged

14 GM main protocol for P i P i P i-1...P 1 Distribute 1-level promises (i-1) level protocol Collect (i-1) level promises Exchange i-level promises

15 GM main prot. (4 participants) P4P3P2P1 otherwise stop otherwise stop 1-level promise

16 GM main prot. (4 participants) P4P3P2P1 otherwise abort 1-level promise 2-level promise otherwise recover otherwise recover otherwise recover

17 GM main prot. (4 participants) P4P3P2P1 3-level promise otherwise recover otherwise recover 3-level promise otherwise recover 3-level promise otherwise recover otherwise recover 3-level promise

18 GM main prot. (4 participants) P4P3P2P1 otherwise recover otherwise recover 4-level promise otherwise recover otherwise recover 4-level promise otherwise recover 4-level promise otherwise recover

19 GM main prot. (4 participants) P4P3P2P1 otherwise recover otherwise recover Signature otherwise recover otherwise recover otherwise recover Signature otherwise recover Signature

20 GM abort and resolve for P i To abort, P i sends to T S Pi (m,P i,(P 1,...,P n ), abort) To resolve, Pi sends to T S Pi ({PCS Pj (m,k j ), P i, T} (j  {1... n}\{i}),S Pi (m,1) where – if j>i, k j is the maximum level of a promise received from P j on m –if j { "@context": "http://schema.org", "@type": "ImageObject", "contentUrl": "http://images.slideplayer.com/13/3715031/slides/slide_20.jpg", "name": "GM abort and resolve for P i To abort, P i sends to T S Pi (m,P i,(P 1,...,P n ), abort) To resolve, Pi sends to T S Pi ({PCS Pj (m,k j ), P i, T} (j  {1...", "description": "n}\{i}),S Pi (m,1) where – if j>i, k j is the maximum level of a promise received from P j on m –if j

21 GM protocol for T Each participant may contact T only once T replies with a resolved contract or an abort token T may overturn an abort, but never a resolve T maintains the following information for each contract to decide when to overturn an abort –validated: a boolean indicating whether the contract has been validated or not –S: the set of indices of parties that have aborted –F: set of indices of parties which helps T to decide when to overturn an abort

22 An attack on abuse-freeness Note that P 1 cannot abort Abort responses include the participants that have aborted If P 1 receives an abort from T it must have send a resolve request Use T as an oracle : –When T receives a resolve request T verifies all promises and, by answering to P 1, provides evidence that all participants have started the protocol

23 An attack on abuse-freeness (2) Consider the protocol instance where n=3 Using Mocha, we show that abuse- freeness does not hold for a honest P 3 P 1 and P 2 have a strategy to reach a state where –P 1 has an abort reply and –P 1 and P 2 have a strategy to obtain P3’s signature –honest P 3 does not have a strategy to obtain P1’s and P 2 ’s signature

24 An attack on abuse-freeness (3) At the beginning P 2 aborts P 1 tries to resolve, but gets an abort reply from T, which it can show to Charlie At that point P 1 and P 2 can choose the outcome –stop the protocol : P 3 is not able to overturn the abort –complete the protocol in an optimistic way Easy fix: make abort replies to different participants indistinguishable

25 An attack on fairness The first attack was discovered when noticing an error in the proof Consider the protocol instance where n=4 Using Mocha, we show that fairness does not hold for a honest P 2 There exists a path such that –P 1, P 3 and P 4 have P 2 ’s signature –there exists a path such that P 2 does not obtain all other signatures Similar attacks can be shown against P 1 and P 3 Using Mocha we did not discover any attack on fairness for n=3

26 An attack on fairness (2) P 1, P 3 and P 4 collude against P 2 P 3 aborts at the beginning – T adds P 3 to S P 1 resolves, but T responds with an abort –T adds P 1 to S and P 2 to F P 2 tries to recover, but as P 2 is in F, T responds with an abort P 4 resolves and T overturns the abort

27 An attack on fairness (3) More generally the attack scenarios are as follows –dishonest P k1 aborts but continues the protocol –dishonest P k2 tries to recover but does not succeed as a side-effect it adds one or several participants to the set F –honest P k3 tries to recover but does not succeed –dishonest P k4 recovers and overturns the abort

28 Another attack on fairness P 4 and P 1 dishonest Dishonest P 4 aborts Dishonest P 3 tries to recover but does not succeed –as a side-effect adds to P 1 the set F Honest P 1 has to recover, but fails Honest P 2 has to recover and overturns

29 Correcting the GM protocol Major revisions required –Getting the decision to overturn abort correct –Recovery protocol and T’s protocol changed Central idea in the revision –Abort overturned if and only if T infers that each signer that contacted it in the past has been dishonest –Idea borrowed from Baum-Waidner protocol Mocha did not discover any attacks for both 3 and 4 signers

30 Revised GM protocol Recovery messages modified so that T can infer the highest-level promises that an honest signer would have sent when the recovery was launched –Leads to a smaller number of possible resolve requests T maintains the list S of signers who have contacted T in the past and received an abort For each signer P i in S,T maintains two integer variables –h i : highest level promise that an honest P i would have sent to any higher indexed signers before contacting T –l i : highest level promise that an honest P i would have sent to any lower indexed signers before contacting T

31 Protocol for T If T ever replies with a signed contract, then T responds with the contract for any further request If the first request to T is a resolve request, then T sends back a signed contract If the first request is an abort request, then T aborts the contract. T may overturn this abort decision in future. T maintains the abort if it cannot deduce that a signer in the list S is dishonest T overturns the abort if it can deduce that all the signers in S have behaved dishonestly T deduces that a signer P i in S is dishonest when contacted by P j if –j>i and P i presents to T a k-level promise from P i such that k>h i, or –j l i

32 Conclusion First formal analysis of multi-party contract signing protocols Using the model-checker Mocha and the logic ATL instances of two protocols have been verified Two new attacks have been discovered in the GM protocol –Abuse-freeness can be broken using side information given by T: easy to fix –Fairness can be broken when n > 3: requires major changes to be fixed

33 Future work Extend strand space formalism to model fair exchange protocols –derive Mocha specifications directly from strands –correctness proofs when no attack is found Extend the analysis to a more complete model –Dolev-Yao-like intruder –Parametric verification Study different topologies, e.g. ring topologies in fair exchange Extend general results on advantage, presented in [Chadha, Mitchell, Scedrov, Shmatikov 2003] to multi- party protocols

34 GM main prot. (4 participants) P4P3P2P1 1-level promise 3-level promise 1-level promise 2-level promise 3-level promise 4-level promise Signature

35 GM main protocol for P i (detailed) PiPi 1-level promise from j (n  j < i) Otherwise, stop 1-level promise to j (i < j  1) agreement of P i...P 1 i-1-level promise to j (i < j  1) i-level promise to j (i < j  1) i-level promise from j (i < j  1) i-level promise to i+1 Otherwise, abort Otherwise, resolve i-level promise from i+1 Otherwise, resolve i+1-level promise to j (i < j  1) i+1-level promise from j (i < j  1) Otherwise, resolve i+1-level promise to j (i < j  i+2) i+2-level promise from j (i+2  j < i) Otherwise, resolve... Otherwise, resolve n+1-level promise from j (n  j < i) and signatures n+1-level promise to j ( j  i) and signatures


Download ppt "Analysis of optimistic multi-party contract signing Rohit Chadha 1,2, Steve Kremer 3,4, Andre Scedrov 1 1 University of Pennsylvania 2 University of Sussex."

Similar presentations


Ads by Google