Improving Security Testing with Usage-Based Fuzz Testing

Abstract

Along with the increasing importance of software systems for our daily life, attacks on these systems may have a critical impact. Since the number of attacks and their effects increases the more systems are connected, the secure operation of IT systems becomes a fundamental property. In the future, this importance will increase, due to the rise of systems that are directly connected to our environment, e.g., cyber-physical systems and the Internet of Things. Therefore, it is inevitable to find and fix security-relevant weaknesses as fast as possible. However, established automated security testing techniques such as fuzzing require significant computational effort. In this paper, we propose an approach to combine security testing with usage-based testing in order to increase the efficiency of security testing. The main idea behind our approach is to utilize that little tested parts of a system have a higher probability of containing security-relevant weaknesses than well tested parts. Since the execution of a system by users can also be to some degree being seen as testing, our approach plans to focus the fuzzing efforts such that little used functionality and/or input data are generated. This way, fuzzing is targeted on weakness-prone areas which in turn should improve the efficiency of the security testing.