Argus: Automated Software Bug Discovery and Assessment

Software bugs are a significant risk to mission success and human safety. Software testing and analysis, which is used to combat bugs, is difficult, and thus time-consuming and expensive. It is particularly difficult to find latent bugsbugs that do not have obvious, observable effects on the system. This results in undiscovered and unfixed bugs in the system. New approaches to automated software testing offer the opportunity to catch bugs with less time and cost than manual approaches. To achieve this, ATC-NY will develop Argus, an automated software testing tool that finds latent errors in a program by analyzing large amounts of testing output. Argus uses big-data machine learning techniques to autonomously analyze records of program behavior to find and prioritize anomalous behavior that may indicate an undiscovered software bug.