While botnets have been extensively studied, bot malware is constantly advancing and seeking to exploit new attack vectors and circumvent existing measures. Existing intrusion detection systems are unlikely to be effective countering advanced techniques deployed in recent botnets. This chapter proposes a deep learning-based botnet traffic analyser called Botnet Traffic Shark (BoTShark). BoTShark uses only network transactions and is independent of deep packet inspection technique; thus, avoiding inherent limitations such as the inability to deal with encrypted payloads. This also allows us to identify correlations between original features and extract new features in every layer of an Autoencoder or a Convolutional Neural Networks (CNNs) in a cascading manner. Moreover, we utilise a Softmax classifier as the predictor to detect malicious traffics efficiently.
CITATION STYLE
Homayoun, S., Ahmadzadeh, M., Hashemi, S., Dehghantanha, A., & Khayami, R. (2018). BoTShark: A deep learning approach for botnet traffic detection. In Advances in Information Security (Vol. 70, pp. 137–153). Springer New York LLC. https://doi.org/10.1007/978-3-319-73951-9_7
Mendeley helps you to discover research relevant for your work.