Logo des Repositoriums

Auflistung nach Schlagwort "Vulnerability Detection"

1 - 2 von 2
  • Konferenzbeitrag
    DifFuzz: Differential Fuzzing for Side-Channel Analysis
    (Software Engineering 2020, 2020) Nilizadeh, Shirin; Noller, Yannic; Noller, Yannic
    This summary is based on our research results on ``DifFuzz: Differential Fuzzing for Side-Channel Analysis'' which was published in the proceedings of the 41st International Conference on Software Engineering. Side-channel analysis aims to investigate the risk that a potential attacker can infer any secret information through observations of the system, such as the execution time or the memory consumption. Side-channel vulnerabilities therefore represent security risks that can cause serious damage and need to be identified and repaired. DifFuzz applies differential fuzzing to identify inputs that trigger such vulnerabilities. Our fuzzing approach analyzes multiple program executions, which vary in their secret information, and uses resource-guided heuristics to identify inputs that maximize the observable cost difference between these executions. Our evaluation shows that such a dynamic analysis approach can find the same side-channel vulnerabilities as state-of-the-art static analysis techniques, and even more vulnerabilities since it does not rely on models for its analysis. Additionally, the advantage of DifFuzz compared to other techniques is not only that it can generate inputs that show a vulnerability, but that the resulting cost difference can also be used to estimate the severity of an identified vulnerability. This enables the comparing of repaired versions of an application.
  • Textdokument
    Ongoing Automated Data Set Generation for Vulnerability Prediction from Github Data
    (GI SICHERHEIT 2022, 2022) Hinrichs, Torge
    This paper describes the development of a continuous github repository analysis pipeline with the focus on creating a data set for vulnerability prediction in source code. Currently, used data sets consist only of source code functions or methods without additional meta information. This paper assumes that the surrounding code of vulnerable functions can be beneficial to the detection rate. In order to test this assumption, large data sets are needed that can be created using the proposed pipeline. Although the pipeline requires some improvements, in a first test run 1.5 million repositories could be analyzed and evaluated. The resulting data set will be published in the future.