FM 2016: Formal Methods
MetadataShow full item record
David, C., Kesseli, P., Kroening, D., & Lewis, M. (2016). Danger Invariants. FM 2016: Formal Methods, 182-198. https://doi.org/10.1007/978-3-319-48989-6_12
Static analysers search for overapproximating proofs of safety commonly known as safety invariants. Conversely, static bug finders (e.g. Bounded Model Checking) give evidence for the failure of an assertion in the form of a counterexample trace. As opposed to safety invariants, the size of a counterexample is dependent on the depth of the bug, i.e., the length of the execution trace prior to the error state, which also determines the computational effort required to find them. We propose a way of expressing danger proofs that is independent of the depth of bugs. Essentially, such danger proofs constitute a compact representation of a counterexample trace, which we call a danger invariant. Danger invariants summarise sets of traces that are guaranteed to be able to reach an error state. Our conjecture is that such danger proofs will enable the design of bug finding analyses for which the computational effort is independent of the depth of bugs, and thus find deep bugs more efficiently. As an exemplar of an analysis that uses danger invariants, we design a bug finding technique based on a synthesis engine. We implemented this technique and compute danger invariants for intricate programs taken from SV-COMP 2016.
External DOI: https://doi.org/10.1007/978-3-319-48989-6_12
This record's URL: https://www.repository.cam.ac.uk/handle/1810/294334
All rights reserved