Functionally Equivalent Variants using Information Synchronization (FEVIS)
The FEVIS project, part of DARPA’s CFAR program, is developing a means to generate functionally equivalent variants using information synchronization. Information synchronization combines program analysis, automated detection of redundant functions and code sequences, and behavioral specifications to generate functionally equivalent variants. Researchers from the University of Minnesota Department of Computer Science and Engineering are key members of the team. Their role is to analyze the software system to identify the places where it makes sense to focus on generating the functional equivalences.
Initially, the technology will assist a human user in understanding the redundancies and other choice points for generating variants. The work will then be extended to automatically generate as many variants as are requested, by systematically choosing among the possible variants at each choice point indicated by the user. The resulting set of variants will be provided to other CFAR performers that are monitoring the software for divergences in behavior among the variants that may indicate an attack.
Distribution Statement “A” (Approved for Public Release, Distribution Unlimited)