DAEMON: Dataset-Agnostic Explainable Malware Classification Using Multi-Stage Feature Mining

Numerous metamorphic and polymorphic malicious variants are generated automatically on a daily basis by mutation engines that transform the code of a malicious program while retaining its functionality, in order to evade signature-based detection. These automatic processes have greatly increased the number of malware variants, deeming their fully-manual analysis impossible. Malware classification is the task of determining to which family a new malicious variant belongs. Variants of the same malware family show similar behavioral patterns. Thus, classifying newly discovered variants helps assess the risks they pose and determine which of them should undergo manual analysis by a security expert. This motivated intense research in recent years of how to devise high-accuracy automatic tools for malware classification. In this paper, we present DAEMON - a novel dataset-agnostic and even platform-agnostic malware classifier. We've optimized DAEMON using a large-scale dataset of x86 binaries, belonging to a mix of several malware families targeting computers running Windows. We then applied it, without any algorithmic change, features re-engineering or parameter tuning, to two other large-scale datasets of malicious Android applications of numerous malware families. DAEMON obtained top-notch classification results on all datasets, making it the first provably dataset-agnostic malware classifier to date. An important byproduct of the type of features used by DAEMON and the manner in which they are mined is that its classification results are explainable.