Probable Innocence in the Presence of Independent Knowledge

Trust in Crowds: Probabilistic Behaviour in Anonymity Protocols

“…Several formal definitions and frameworks have been proposed for reasoning about secure information flow analysis (e.g., [7,8,19]), side-channel analysis (e.g., [13]) and anonymity. Our work follows a recent trend in the analysis of anonymity protocols directed to the application of informationtheoretic notions (e.g., [1,2,[4][5][6]9,12,15,17,18]), whereby the work closer to the present one are those by Reiter and Ruben [16], Halpen and O'Neill [10], Chatzikokolakis and Palamidessi [3], and a recent paper Hamadou et al [12].…”

“…Finally we recall that the concept of probable innocence was recently generalised in [12]. Instead of just comparing the probability of being innocent with the probability of being guilty, loc.…”

“…In [12] the authors first generalise the concepts of probable innocence and relate it to Smith's concept of protocol vulnerability [19]. Instead of just comparing the probability of being innocent with the probability of being guilty, they compare such probabilities against a parameter α.…”

Quantifying leakage in the presence of unreliable sources of information

“…The min-entropy approach remains however the canonical framework, not only for its simplicity, but also because the worst-case min-leakage (aka min-capacity) has been proved to be an upper bound to the g-leakage [25]. In [28] the authors extended the vulnerability model of [24] in the context of the Crowds protocol for anonymous message posting to encompass the frequent situation where attackers have extra knowledge. They pointed out that in Crowds the adversary indeed has extra information (viz., the target servers) and assumed that she knows the correlation between that and the secret (viz., the users' preferences for servers).…”

“…Indeed, [28] assumes that the adversary's extra information is accurate, and such an assumption is generally not warranted. Inaccuracy can indeed arise, e.g.…”