Markovian Processes for Quantitative Information Leakage

Fabrizio Biondi 1
1 ESTASYS - Efficient STAtistical methods in SYstems of systems
Inria Rennes – Bretagne Atlantique , IRISA-D4 - LANGAGE ET GÉNIE LOGICIEL
Résumé : Quantification of information leakage is a successful approach for evaluating the security of a system. It models the system to be analyzed as a channel with the secret as the input and an output as observable by the attacker as the output, and applies information theory to quantify the amount of information transmitted through such channel, thus effectively quantifying how many bits of the secret can be inferred by the attacker by analyzing the system’s output. Channels are usually encoded as matrices of conditional probabilities, known as channel matrices. Such matrices grow exponentially in the size of the secret and observables, are cumbersome to compute and store, encode both the behavior of the system and assumptions about the attacker, and assume an input-output behavior of the system. For these reasons we propose to model the system-attacker scenario with Markovian models. We show that such models are more compact and treatable than channel matrices. Also, they clearly separate the behavior of the system from the assumptions about the attacker, and can represent even non-terminating behavior in a finite model. We provide techniques and algorithms to model and analyze both deterministic and randomized processes with Markovian models and to compute their information leakage for a very general model of attacker. We present the QUAIL tool that automates such analysis and is able to compute the information leakage of an imperative WHILE language. Finally, we show how to use QUAIL to analyze some interesting cases of secret-dependent protocols.
Type de document :
Thèse
Cryptography and Security [cs.CR]. IT University of Copenhagen, 2014. English
Liste complète des métadonnées

Littérature citée [158 références]  Voir  Masquer  Télécharger

https://hal.inria.fr/tel-01242619
Contributeur : Fabrizio Biondi <>
Soumis le : vendredi 18 décembre 2015 - 10:31:39
Dernière modification le : mercredi 11 avril 2018 - 02:00:44
Document(s) archivé(s) le : samedi 29 avril 2017 - 12:44:45

Identifiants

  • HAL Id : tel-01242619, version 1

Citation

Fabrizio Biondi. Markovian Processes for Quantitative Information Leakage. Cryptography and Security [cs.CR]. IT University of Copenhagen, 2014. English. 〈tel-01242619〉

Partager

Métriques

Consultations de la notice

291

Téléchargements de fichiers

1099