What is it about?
In this paper, we draw inspiration from recent advances in cybersecurity and propose a novel approach using the concept of defensive distillation to construct a new architecture for source code authorship attribution, enhancing its robustness against adversarial attacks (reducing misclassification rates). Through empirical analysis, we demonstrate the effectiveness of our approach, defensive distillation, combined with varied feature selection, in reducing misclassifications on perturbed source code files from Google Code Jam and GitHub databases while maintaining a 95% accuracy on legitimate source code files.
Featured Image
Why is it important?
Although it poses a privacy threat to open-source programmers, SCAA has proven immensely valuable in developing forensic-based applications, including ghostwriting detection, copyright dispute resolution, identifying authors of malicious applications using source code, and other code analysis applications. Recent advancements in SCAA techniques have demonstrated exceptional performance on diverse datasets. Specific challenges have emerged, such as gradient-based attacks and universal perturbations that can adversarially modify source code, significantly reducing the accuracy of state-of-the-art classification techniques based on deep neural networks to as low as 10%.
Read the Original
This page is a summary of: AuthAttLyzer: A Robust defensive distillation-based Authorship Attribution framework, December 2022, ACM (Association for Computing Machinery),
DOI: 10.1145/3586102.3586109.
You can read the full text:
Contributors
The following have contributed to this page