Combining Invariant Violation with Execution Path Classification for Detecting Multiple Types of Logical Errors and Race Conditions.

Context: Modern automated source code analysis techniques can be very successful in detecting a priori de- fined defect patterns and security vulnerabilities. Yet, they cannot detect flaws that manifest due to erroneous translation of the software’s functional requirements into the source code. The automated detection of logical errors that are attributed to a faulty implementation of applications’ functionality, is a relatively uncharted territory. In previous research, we proposed a combination of automated analyses for logical error detection. In this paper, we develop a novel business-logic oriented method able to filter mathematical depictions of software logic in order to augment logical error detection, eliminate previous limitations in analysis and provide a formal tested logical error detection classification without subjective discrepancies. As a proof of concept, our method has been implemented in a prototype tool called PLATO that can detect various types of logical errors. Potential logical errors are thus detected that are ranked using a fuzzy logic system with two scales characterizing their impact: (i) a Severity scale, based on the execution paths’ characteristics and Information Gain, (ii) a Reliability scale, based on the measured program’s Computational Density. The method’s effectiveness is shown using diverse experiments. Albeit not without restrictions, the proposed automated analysis seems able to detect a wide variety of logical errors, while at the same time limiting the false positives.