News
SyzAnalyzer: A tool for analysing difficult-to-reproduce errors in the Linux kernel
Abstract
In today's world, the Linux operating system holds a significant position, being utilized in a wide range of applications, from high-performance server systems to consumer devices. The Linux kernel is actively being developed, and as a result the size of its code base has exceeded forty million lines. This, in turn, can lead to a significant number of bugs and vulnerabilities, which poses a critical risk to the security and stability of Linux-based systems. One of the popular methods of error detection is fuzzing testing, which is based on the generation of programs that perform a large number of pseudorandom system calls. The most popular fuzzing testing tool for operating systems is Syzkaller. When a kernel crash is detected, the fuzzer tries to generate so-called reproducer programs that allow reproducing the detected error. However, a significant number of the bugs found are challenging to reproduce. Syzkaller currently lacks the tools to automate the process of verifying the reproducibility of errors and identifying the commit that triggered them. This significantly reduces the likelihood that errors will be investigated and resolved. In this paper, we present SyzAnalyzer, which is designed to automate the process of analyzing difficult-to-reproduce errors. The tool allows reducing the effort required to localize such errors and has already allowed to debug several difficult-to-reproduce errors.
Keywords
Edition
Proceedings of the Institute for System Programming, vol. 38, issue 6, part 2, 2026, pp. 7-20
ISSN 2220-6426 (Online), ISSN 2079-8156 (Print).
DOI: 10.15514/ISPRAS-2026-38(6)-22
For citation
Full text of the paper in pdf (in Russian)
Back to the contents of the volume