Cause Effect Path Based Coverage Criteria for Testing Buffer Overflow Vulnerability

Buffer overflow vulnerabilities are well known, as a number of attacks have exploited the vulnerability over decades. Despite the fact that there has been considerable effort towards prevention and detection of buffer overflow vulnerabilities using static and dynamic analyses and testing, there are periodic instances of exploitation of buffer over-flow vulnerability. More attention has been paid to compiled languages than scripting languages with respect to the prevention and detection of the buffer overflow vulnerability, however, the issue is not entirely resolved even for the former case. This paper attempts to characterize code coverage criteria for buffer overflow vulnerability in a novel way based on cause-effect chains that lead to the exploitation of the vulnerability in sequential programs. Weak and strong notions of coverage criteria as the basis for testing buffer overflow vulnerability are proposed. Feedback mechanisms by a programming language environment to developers and testers about uncovered cause effect paths containing potential buffer overflow vulnerabilities are introduced. A regression test selection algorithm for buffer overflow vulnerability is discussed along with various cases that arise. An advantage of the approach for practitioners is that a systematic method can be employed to detect buffer overflow vulnerabilities early in the software development life cycle, for example, in unit testing phase.