Verification of cryptographic protocols : A bound on the number of agents

Nowadays, cryptography and security protocols are used everywhere, and public confidence in credit cards or electronic voting relies on their trustworthiness. As flaws can stay unremarked during several years, the scientific community has undertaken the automation of their research through formal methods, or to prove the protocols secure. Two families of models have been introduced : computational models are based on real properties of cryptographic primitives whereas symbolic models, which I used, assume perfect cryptography but allow more automation. In each model, security properties should hold for an arbitrary number of agents involved in an unbounded number of sessions. This together with the unbounded size of the messages, makes the verification problem undecidable. Works verification of security properties in the symbolic model search for decidable classes of protocols and properties, or try to improve automated research of flaws.