- •Abstract
- •Resumé
- •Contents
- •Remerciements
- •Personal Bibliography
- •Introduction
- •The Need for Dedicated Cryptographic Primitives for RFID Tags
- •Privacy Issues in RFID Systems
- •Our Privacy Model
- •Preliminaries
- •Notations
- •Probabilities and Negligible Functions
- •Classical Cryptography
- •Message Authentication Codes
- •Cryptographic Hash Functions
- •Universal Hash Functions
- •Pseudo-Random Functions
- •The Random Oracle Model
- •Proof Techniques
- •Hard Problems
- •The LPN Problem and the HB Family
- •The LPN Problem
- •Extensions of the LPN Problem
- •Security Models for the HB Family
- •The HB Protocol
- •The GRS Attack
- •Attempts to Thwart the GRS Attack
- •Description
- •Proposed Parameter Sets
- •Asymptotic Complexity Analysis
- •Optimizing the Attack
- •Thwarting the Attack: the Case of Vectors without False Rejections
- •Perspectives
- •SQUASH
- •Description
- •Handling Window Truncation
- •Handling the Truncation of the Combinaison of Many Integers
- •Generalization
- •Conclusion
- •Privacy Failures in RFID Protocols
- •ProbIP and the SAT Problem
- •Violation of Anonymous Privacy
- •Future Development
- •MARP
- •Description
- •Auth2
- •Description
- •YA-TRAP+
- •O-TRAP
- •A Backward and Forward Untraceable Protocol
- •Tracing O-FRAP
- •Violating the Forward Privacy of O-FRAP
- •Conclusion
- •Privacy Models for RFID
- •The ADO Model
- •Description
- •RFID System
- •Correctness
- •Privacy
- •From Narrow Privacy to Privacy
- •Narrow-Strong and Forward Privacy Using Public-Key Encryption
- •Achieving Strong Privacy
- •Our Proposal: Incorporate the Blinder into the Adversary
- •Sampling Algorithms and the ISH Hypothesis
- •Plaintext-Awareness
- •Instances of Plaintext-Aware Encryption Schemes
- •From PA+ to PA++ Plaintext-Awareness
- •Privacy
- •Security Proof
- •Correctness
- •Security
- •The Case of Mutual Authentication
- •RFID System with Mutual Authentication
- •Correctness
- •Privacy
- •Correctness and Security for the Reader
- •Security for the Tags
- •Strong Privacy with Mutual Authentication
- •Strong Privacy
- •Conclusion
- •The Security of RFID Primitives
- •Our Contributions
- •Further Work
- •Our Contributions
- •Further Work
- •Final Notes
- •List of Figures
- •List of Tables
- •List of Definitions
- •Bibliography
- •Curriculum Vitæ
|
|
Destructive. It refers to the class of adversaries for who tampering with a tag results in its destruction. In a more formal sense, a Destructive adversary is not allowed to issue any query with vtag a er requesting C (vtag).
Forward. A er a Forward adversary corrupts a tag, she is only allowed to corrupt other tags or terminate.
Weak. is class captures the set of adversaries who cannot corrupt any tag.
Orthogonal to this classi cation, we also consider the case in which adversaries do not have access to the R oracle. Such adversaries are referred to as Narrow and for every class listed abo e, we consider a Narrow counterpart.
Narrow-Strong. is class denotes the set of Strong adversaries who do not access R .
Narrow-Destructive. isincludesallDestructiveadversarieswhocannotaccessR .
Narrow-Forward. is is equivalent to Narrow [ Forward.
Narrow-Weak. is is the class of the weakest adversaries who can neither corrupt tags nor access R .
Regarding the relation between those adversarial classes, it is clear that for every non-narrow classP wehaveNarrow P P . ItalsoholdsthatWeak Forward Destructive Strong and Narrow-Weak Narrow-Forward Narrow-Destructive Narrow-Strong.
. . Matching Con ersation
Before de ning the necessary properties of an RFID scheme, we formalize the event that a tag and a reader have an undisturbed protocol instance. is notion will prove to be useful to de ne correctness and security for RFID systems.
De nition . (Matching Conversation)
We say that a protocol instance had a matching con ersation with the tag ID if they exchanged well interleaved and faithfully (but maybe with some time delay) messages until is completed.
. . Correctness
Basically, correctness formalizes the fact that whenever the reader and a tag ID participate in an undisturbed protocol session, the reader authenticates the tag, that is, it ends up with Output = ID, except with a small negligible probability. e di erence between our de nition and vaudenay’s de nition of correctness is that we take into account all possible actions that may have happened in the past for the system. at is, we require that a legitimate tag remains successful in authenticating itself and an illegitimate one gets rejected regardless of the past events that occurred in the RFID system. e de nition we propose is in fact close to the de nition of adaptive completeness from the ZK-Privacy model [DLYZ ].
.
|
|
|
De nition . (Correctness of an RFID Scheme) |
|
Let A be a Strong adversary interacting with the RFID system in which she creates n RFID tags |
|
and produces no output. We also assume without loss of generality that A ees all tags before |
|
terminating. |
|
An RFID scheme is said to be correct if for every such A and every e cient sampling algorithm |
|
Samp on the set of the system’s tags, we have |
Pr |
2 b = 1 ^ Output( ) = T (vtag) |
||
|
6 |
^ |
? |
|
4 |
|
_ |
|
6 |
b = 0 |
Output( ) = |
|
( |
pk; sk) |
SetupReader(1k) |
|||||
|
Execute |
A |
(pk) |
|
|
|||
|
(vtag; b) |
|
|
(Samp) |
||||
|
|
|
|
D T |
|
|||
|
|
|
|
|
||||
|
|
|
|
|
|
|
|
|
|
|
( ; ) |
E ( |
|
) |
|||
|
|
vtag |
||||||
|
|
|
|
|
|
|
|
7 |
|
7 |
= 1 negl(k) |
3 |
|
5 |
|
We also propose a weaker notion of correctness in which only tags that have not completed more than t consecutive unsuccessful instances get authenticated by the reader. at is, we propose the following de nition.
De nition . (Weak Correctness for Simple RFID Schemes)
A simple RFID system is said to be weakly-correct if
ere exists an e ciently computable predicate ′ such that if a tag ID and the reader have a matching con ersation in a session and the tag ID has previously completed t successive sessions without the reader authenticating it, we have
jPr[ ′(ID; t) ! 1] Pr[Output( ) = ID]j = negl(k)
For every Strong adversary A that produces no output but ees all the tags before terminating, and every e cient sampling algorithm Samp on the set of the system’s tags, we have
2
(b = 1 ^ ′(T (vtag); t)
6 ^Output( ) = T (vtag))
Pr 6
4 _
(b = 0 ^ Output( ) =?)
(pk; sk)
(vtag; b)
( ; )
SetupReader(1k) 3
A(pk) 7
7 = 1 negl(k)
D T (Samp) 5
E (vtag)
For simple schemes, this de nition of weak-correctness means that the output of ′ is computationally indistinguishable from the output of with matching sessions and known tag identi ers. Clearly, this de nition is less restrictive than Vaudenay’s who mandated perfect indistinguishability between the two predicates, i.e., the original de nition states thar the two predicates have to be equivalent. Since the de nition of correctness leaves a negligible probability that the reader authenticates another tag in place of the one it is running the instance with, re ecting this probability in the ′ predicate is reasonable.
. ’
|
|
. . Security
Security is the equivalent of soundness in the eJW model. It formalizes the fact that no adversary should be able to make the reader accept a protocol session in which the adversary has been actively involved in the sense that sge did not only relay messages. In summary, an RFID scheme is said to be secure if no Strong adversary is able to make a reader protocol instance recognize an uncorrupted tag ID except with negligible probability and that is even if the adversary corrupts all the other tags, unless and the tag have a matching conversation.
De nition . (Security of an RFID System)
We say that an RFID scheme is secure if for every Strong adversary, the probability that the reader ends with a tape Output = ID for a session that has no matching con ersation with the tag ID is negligible in the RFID scheme’s security parameter.
Simple RFID schemes enjoy an interesting property: their security reduces to an adversary playing with a system consisting of a single tag IDt and having access to an oracle implementing(sk; ; ; ) to which she can submit triplets of the form (ID; KID; ) with the restriction ID ≠ IDt. In the following, we give the formal proof that this simpli cation holds for our class of simple and weakly-correct RFID systems.
De nition . (Security of Simple and Weakly-Correct RFID Systems)
For simple RFID schemes that are weakly-correct, we consider the following simpli ed security game for adversaries who are given access to an oracle O who checks the predicate (sk; ; ; ).
: (sk; pk)
: C T 1(ID)
: vtag D T (ID) : L
: Run AO interacting with L , S R , and S T . AO is not allowed to specify ID in its queries to O .
: b R ( )
: Output if has no matching con ersation with ID and R ( ) = 1.
e scheme is said to be simply secure if the winning probability of any adversary playing the simple security experiment is negligible in the security parameter.
Lemma .
For simple and weakly-correct RFID schemes, simple security implies security.
Proof. We use the game proof methodology to reduce an adversary against the security of the scheme to an adversary playing the simple security game. We denote by Si the event that A wins the experiment described by game i.
Game . |
is denotes the original security game played by a xed Strong adversary A. |
We |
|
let S0 be the event that A succeeds. Recall that A has access to all interfaces. |
We |
.
|
|
|
|
|
|||
|
assume, w.l.o.g., that A stops as soon as it wins the security game, i.e., one protocol |
||||||
|
session identi |
|
es a tag ID without the two having a matching conversation. |
||||
|
Game . We relax A’s winning condition by declaring that is su cient that one instance |
||||||
|
with transcript satis es on an input (ID; KID) for which ID had no matching |
||||||
|
conversation with . We further stop the adversary as soon as it wins the game |
||||||
|
under this condition. Note that the adversary wins the original security game if this |
||||||
|
tag has been selected from the set E . |
erefore, we nd that |
|
||||
|
Pr[S1] Pr[S0] |
|
|
|
|||
|
Game . We add a new condition for A to win by requiring it to correctly guess the target |
||||||
|
tag ID when created and the target instance when launched. If S3 is the event that |
||||||
|
the adversary wins this game and n, t are the number of tags created and sessions |
||||||
|
launched respectively, we have |
|
|
|
|||
|
|
1 |
|
|
|
|
|
|
Pr[S2] |
|
Pr[S1] |
|
|
|
|
|
nt |
|
|
|
|||
|
Game . In this game, we simulate all A’s drawings. |
at is we construct an algorithm A1 |
|||||
|
such that, each time a tag is created, A1 draws it, and subsequently simulates all A’s |
||||||
|
D T and F queries. Clearly, the views of A in both games are perfectly |
||||||
|
indistinguishable so the winning probability remains una ected. In other words, |
||||||
|
Pr[S3] |
Pr[S2] = 0 |
|
|
|
||
|
Game . We now simulate the creation of all tags except the target one. |
at is, we process |
|||||
|
all C T queries with a parameter di erent from ID in the following way. A |
||||||
|
calls SetupTagpk( ) togeneratethetagstateandthekeyforthedatabase. Ifthequery |
||||||
|
concerns a legitimate tag, A inserts the entry into a list of legitimate tags Tags1. |
||||||
|
Since Ahasknowledgeofallstatesofthetags, shecansimulateallS T queries |
||||||
|
related to any tag, except ID that is forwarded to the S T interface (Recall that |
||||||
|
A draws tags herself so she knows the real ID of every tag). |
e simulation is thus |
|||||
|
perfect, i.e., |
|
|
|
|
|
|
|
We also need to show that Output, and thus R , can be simulated. To deter- |
||||||
|
mine the outcome of a protocol session, A tests queries O on every entry except |
||||||
|
(ID; KID) to determine which entry satis |
es . As for (ID; KID), A assumes that |
|||||
|
would answer 0 if the instance does not have matching conversation with that tag. |
||||||
|
Otherwise, it assumes it to be 1. erefore, when the predicate tested with (ID; KID) |
||||||
|
would have yielded 0, A perfectly simulates Output (the rest of the protocol mes- |
||||||
|
sages do not depend on KID if ID has not been identi ed). If the predicate would |
||||||
|
have answered 1 with (ID; KID) and without matching conversation, it should al- |
||||||
|
ready have been the target session and this is addressed with another selection in |
||||||
|
Game . So, simulation is perfect and we |
nd that |
|
Pr[S4] = Pr[S3]
. ’