2021/10/22 by Marek Galovic, Galovic, Marek, Branislav Bošanský +3 · 1 citation
Computer Science · #Advanced Malware Detection Techniques #Adversarial Robustness in Machine Learning #Cryptography and Security (cs.CR) #FOS: Computer and information sciences #Machine Learning (cs.LG) #Network Security and Intrusion Detection
paper · pdf · doi:10.48550/arxiv.2110.11987
openalex publication_date 2021/10/22 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
In malware behavioral analysis, the list of accessed and created files very\noften indicates whether the examined file is malicious or benign. However,\nmalware authors are trying to avoid detection by generating random filenames\nand/or modifying used filenames with new versions of the malware. These changes\nrepresent real-world adversarial examples. The goal of this work is to generate\nrealistic adversarial examples and improve the classifier's robustness against\nthese attacks. Our approach learns latent representations of input strings in\nan unsupervised fashion and uses gradient-based adversarial attack methods in\nthe latent domain to generate adversarial examples in the input domain. We use\nthese examples to improve the classifier's robustness by training on the\ngenerated adversarial set of strings. Compared to classifiers trained only on\nperturbed latent vectors, our approach produces classifiers that are\nsignificantly more robust without a large trade-off in standard accuracy.\n