DeepSign: Deep Learning for Automatic Malware Signature Generation and Classification

TL;DR

This paper introduces a deep learning approach using deep belief networks to automatically generate malware signatures, achieving high accuracy in classifying new malware variants regardless of behavior type.

Contribution

The paper presents a novel deep belief network method for malware signature generation that outperforms traditional signature-based detection in identifying new malware variants.

Findings

Achieved 98.6% classification accuracy on malware variants

Signatures are invariant and effective across different malware behaviors

Method is agnostic to input types from sandbox environments

Abstract

This paper presents a novel deep learning based method for automatic malware signature generation and classification. The method uses a deep belief network (DBN), implemented with a deep stack of denoising autoencoders, generating an invariant compact representation of the malware behavior. While conventional signature and token based methods for malware detection do not detect a majority of new variants for existing malware, the results presented in this paper show that signatures generated by the DBN allow for an accurate classification of new malware variants. Using a dataset containing hundreds of variants for several major malware families, our method achieves 98.6% classification accuracy using the signatures generated by the DBN. The presented method is completely agnostic to the type of malware behavior that is logged (e.g., API calls and their parameters, registry entries, websites and ports accessed, etc.), and can use any raw input from a sandbox to successfully train the deep neural network which is used to generate malware signatures.