学术文献库

Many software analysis techniques attempt to determine whether bugs are reachable, but for security purpose this is only part of the story as it does not indicate whether the bugs found could be easily triggered by an attacker. The recently introduced notion of robust reachability aims at filling this gap by distinguishing the input controlled by the attacker from those that are not. Yet, this qualitative notion may be too strong in practice, leaving apart bugs which are mostly but not fully replicable. We aim here at proposing a quantitative version of robust reachability, more flexible and still amenable to automation. We propose quantitative robustness, a metric expressing how easily an attacker can trigger a bug while taking into account that he can only influence part of the program input, together with a dedicated quantitative symbolic execution technique (QRSE). Interestingly, QRSE relies on a variant of model counting (namely, functional E-MAJSAT) unseen so far in formal verification, but which has been studied in AI domains such as Bayesian network, knowledge representation and probabilistic planning. Yet, the existing solving methods from these fields turn out to be unsatisfactory for formal verification purpose, leading us to propose a novel parametric method. These results have been implemented and evaluated over two security-relevant case studies, allowing to demonstrate the feasibility and relevance of our ideas.