Empirical Quantification of Spurious Correlations in Malware Detection
This work addresses the issue of unreliable malware detection models for security practitioners, but it is incremental as it builds on prior observations without major breakthroughs.
The paper tackled the problem of spurious correlations in deep learning-based malware detection, specifically quantifying how much models rely on compiler-generated empty spaces rather than compiled code, and introduced a ranking of two models to assess their suitability for production.
End-to-end deep learning exhibits unmatched performance for detecting malware, but such an achievement is reached by exploiting spurious correlations -- features with high relevance at inference time, but known to be useless through domain knowledge. While previous work highlighted that deep networks mainly focus on metadata, none investigated the phenomenon further, without quantifying their impact on the decision. In this work, we deepen our understanding of how spurious correlation affects deep learning for malware detection by highlighting how much models rely on empty spaces left by the compiler, which diminishes the relevance of the compiled code. Through our seminal analysis on a small-scale balanced dataset, we introduce a ranking of two end-to-end models to better understand which is more suitable to be put in production.