formal computational unlinkability proofs of rfid
play

Formal Computational Unlinkability Proofs of RFID Protocols Hubert - PowerPoint PPT Presentation

Formal Computational Unlinkability Proofs of RFID Protocols Hubert Comon, Adrien Koutsos January 29, 2018 Hubert Comon, Adrien Koutsos Formal Proofs of RFID Protocols January 29, 2018 1 / 21 Motivations Security protocols Distributed


  1. Formal Computational Unlinkability Proofs of RFID Protocols Hubert Comon, Adrien Koutsos January 29, 2018 Hubert Comon, Adrien Koutsos Formal Proofs of RFID Protocols January 29, 2018 1 / 21

  2. Motivations Security protocols Distributed programs which aim at providing some security properties. The KCL + RFID protocol $ ← R : n R $ T A : n T ← 1 : R − → T A : n R 2 : T A − → R : � A ⊕ H ( n T , k A ) , n T ⊕ H ( n R , k A ) � Hubert Comon, Adrien Koutsos Formal Proofs of RFID Protocols January 29, 2018 2 / 21

  3. Security Properties Security protocols are short: few lines of specification. Security properties are complex: the attacker controls the network. ⇒ Need to use formal methods. The problem Given a protocol P and a class of attackers C , show that: ∀A ∈ C ( P | A ) satisfies φ sec Hubert Comon, Adrien Koutsos Formal Proofs of RFID Protocols January 29, 2018 3 / 21

  4. Attacker Models Models Dolev Yao Computational Messages representation: Abstract terms Bitstrings Explicitly specified Polynomial Time Adversaries capabilities: through a TRS Probabilistic TMs Advantages and drawbacks Dolev Yao Computational Good proof automation Few proof automation Not a realistic model Strong security guarantees But with implicit hypothesis Hubert Comon, Adrien Koutsos Formal Proofs of RFID Protocols January 29, 2018 4 / 21

  5. The Complete Symbolic Attacker Model The Complete Symbolic Attacker Model [Bana,Comon 2012] A first-order logic. Axioms specifying what the adversary cannot do. Security of a protocol expressed as a goal formula. Advantages All hypotheses appear explicitly in the axioms. Possible proof automation. Security implies computational security. Two logics Reachability properties: [Scerri 2016] We focus on the indistinguishability logic. Hubert Comon, Adrien Koutsos Formal Proofs of RFID Protocols January 29, 2018 5 / 21

  6. Motivations 1 The Complete Symbolic Attacker Model 2 Syntax Computational semantics Axioms 3 Structural Axioms Pseudo Random Function Case Studies: Security of Two RFID Protocols 4 Conclusion 5 Hubert Comon, Adrien Koutsos Formal Proofs of RFID Protocols January 29, 2018 6 / 21

  7. Syntax Term algebra Control flow function symbols: if _ then _ else _ , EQ ( _ ; _ ) , true , false Protocol function symbols: {� _ , _ � , π 1 ( _ ) , π 2 ( _ ) , H ( _ , _ ) , _ ⊕ _ } Adversarial function symbols G . A set of names N . A set of variables X . Formulas φ ::= � u ∼ � v | φ ∧ φ | ¬ φ | ⊥ | ∀ x .φ where � v are sequences of terms u , � Hubert Comon, Adrien Koutsos Formal Proofs of RFID Protocols January 29, 2018 7 / 21

  8. Example The KCL + protocol 1 : R − → T A : n R 2 : T A − → R : � A ⊕ H ( n T , k A ) , n T ⊕ H ( n R , k A ) � Example Terms: m A = � A ⊕ H ( n T , k A ) , n T ⊕ H ( g ( n R ) , k A ) � Formula: n R , m A ∼ n R , m B Hubert Comon, Adrien Koutsos Formal Proofs of RFID Protocols January 29, 2018 8 / 21

  9. Computational Semantics of Terms Computational model M c : term interpretation f / n ∈ Σ ∪ G interpreted as a polynomial time Turing Machine. n ∈ N interpreted as a random sampling { if _ then _ else _ , EQ ( _ ; _ ) , true , false } interpretations are the expected ones. Computational model M c : predicate interpretation ∼ interpreted as computational indistinguishability. Example For every computational model M c we have: M c | = A ⊕ n 1 ∼ B ⊕ n 2 Hubert Comon, Adrien Koutsos Formal Proofs of RFID Protocols January 29, 2018 9 / 21

  10. Proof Technique Goal u ∼ � Ground formula � v expressing the security of the protocol. The formula is automatically obtained by folding the executions of the protocol [Bana,Comon 14]. Axioms A : what the adversary cannot do Computationally valid structural axioms. Implementation and cryptographic axioms. Soundness Theorem [Bana,Comon 14] If A ∧ � u �∼ � v is unsatisfiable then the protocol is computationally secure. (under some cryptographic/implementation assumptions) Hubert Comon, Adrien Koutsos Formal Proofs of RFID Protocols January 29, 2018 10 / 21

  11. Motivations 1 The Complete Symbolic Attacker Model 2 Syntax Computational semantics Axioms 3 Structural Axioms Pseudo Random Function Case Studies: Security of Two RFID Protocols 4 Conclusion 5 Hubert Comon, Adrien Koutsos Formal Proofs of RFID Protocols January 29, 2018 11 / 21

  12. Structural Axioms : Examples Relation axioms x ∼ y Sym x ∼ y y ∼ z Trans Refl x ∼ x y ∼ x x ∼ z ∼ is not a congruence! Counter-Example: n ∼ n and n ∼ n ′ , but n , n �∼ n , n ′ . Function Application If you cannot distinguish the arguments, you cannot distinguish the images. x 1 , . . . , x n ∼ y 1 , . . . , y n FunApp f ( x 1 , . . . , x n ) ∼ f ( y 1 , . . . , y n ) Hubert Comon, Adrien Koutsos Formal Proofs of RFID Protocols January 29, 2018 12 / 21

  13. Pseudo Random Function Definition H is a Pseudo Random Function if for every PPTM adversary A : | Pr ( k : A O H ( · , k ) ( 1 η ) = 1 ) − Pr ( g : A O g ( · ) ( 1 η ) = 1 ) | is negligible in η , where: k is drawn uniformly in { 0 , 1 } η . g is drawn uniformly in the set of all functions from { 0 , 1 } ∗ to { 0 , 1 } η . Hubert Comon, Adrien Koutsos Formal Proofs of RFID Protocols January 29, 2018 13 / 21

  14. Translation in the Logic Axiom for one hash H ( s , k ) ∼ n Where k does not appear in s . Bad axiom for two hashes If s and t are syntactically distinct, H ( s , k ) , H ( t , k ) ∼ H ( s , k ) , n Counter-Example: s = g ( A ) , t = g ( B ) and we interpret the attacker function g as a constant function. Hubert Comon, Adrien Koutsos Formal Proofs of RFID Protocols January 29, 2018 14 / 21

  15. Translation in the Logic The PRF 2 Axioms H ( s , k ) , if EQ ( t ; s ) then 0 else H ( t , k ) ∼ H ( s , k ) , if EQ ( t ; s ) then 0 else n where: H and k only occur in ( s , t ) as H ( s , k ) . n does not occur in ( s , t ) . Theorem : Soundness The ( PRF n ) n ∈ N axioms are valid in every computational model M c such that the interpretation of H satisfies the PRF assumption. Hubert Comon, Adrien Koutsos Formal Proofs of RFID Protocols January 29, 2018 15 / 21

  16. Motivations 1 The Complete Symbolic Attacker Model 2 Syntax Computational semantics Axioms 3 Structural Axioms Pseudo Random Function Case Studies: Security of Two RFID Protocols 4 Conclusion 5 Hubert Comon, Adrien Koutsos Formal Proofs of RFID Protocols January 29, 2018 16 / 21

  17. Security Property KCL + Protocol: Unlinkability for 2 rounds (A , A vs. A , B) φ sec n R , m 1 , n ′ R , m A 2 ∼ n R , m 1 , n ′ R , m B ≡ 2 2 where m 1 , m A 2 are the terms: m 1 = � A ⊕ H ( n T , k A ) , n T ⊕ H ( g ( n R ) , k A ) � m X 2 = � X ⊕ H ( n ′ T , k X ) , n ′ T ⊕ H ( g ′ ( n R , m 1 , n ′ R ) , k X ) � Unlinkability for n Rounds. A formula φ sec expressing unlinkability for n rounds of a protocol can n be automatically computed from the specification. If A ∧ ¬ φ sec is unsatisfiable then the protocol satisfies Strong n Privacy [Juels,Weis 2009] for n rounds. Hubert Comon, Adrien Koutsos Formal Proofs of RFID Protocols January 29, 2018 17 / 21

  18. Case Studies Theorem: Unlinkability of KCL + Assuming PRF for the keyed hash function, the KCL + protocol verifies Strong Privacy for two agents and any number of rounds. Theorem: Unlinkability of LAK + Assuming PRF for the keyed hash function, the LAK + protocol verifies Strong Privacy for two agents and two rounds. Hubert Comon, Adrien Koutsos Formal Proofs of RFID Protocols January 29, 2018 18 / 21

  19. Motivations 1 The Complete Symbolic Attacker Model 2 Syntax Computational semantics Axioms 3 Structural Axioms Pseudo Random Function Case Studies: Security of Two RFID Protocols 4 Conclusion 5 Hubert Comon, Adrien Koutsos Formal Proofs of RFID Protocols January 29, 2018 19 / 21

  20. Conclusion Contributions Designed and proved axioms for PRF, CR, XOR and PRNG. Formally expressed Strong Privacy [Juels,Weis 2009] in our model. Proved Strong Privacy of KCL + for an arbitrary number of rounds. Proved Strong Privacy LAK + protocol for two rounds. Showed attacks against KCL + and LAK + for weaker assumptions. Future Work More examples, with more primitives (RFID or not). Automation through decidability of (a fragment of) the logic. Interactive/automatic prover. Hubert Comon, Adrien Koutsos Formal Proofs of RFID Protocols January 29, 2018 20 / 21

  21. Thanks for your attention Hubert Comon, Adrien Koutsos Formal Proofs of RFID Protocols January 29, 2018 21 / 21

Recommend


More recommend