DGA Botnet Detection Using Supervised Learning Methods
Hieu Mac, Duc Tran, Van Tong, Linh Giang Nguyen, Hai Anh Tran · 2017
Modern botnets are based on Domain Generation Algorithms (DGAs) to build a resilient communication between bots and Command and Control (C&C) server. The basic aim is to avoid blacklisting and evade the Intrusion Protection Systems (IPS). Given the prevalence of this mechanism, numerous solutions have been developed in the literature. In particular, supervised learning has received an increased interest as it is able to operate on the raw domains and is amenable to real-time applications. Hidden Markov Model, C4.5 decision tree, Extreme Learning Machine, Long Short-Term Memory networks have become the state of the art in DGA botnet detection. There also exist several advanced supervised learning methods, namely Support Vector Machine (SVM), Recurrent SVM, CNN+LSTM and Bidirectional LSTM, which have not been suitably appropriated in such domain. This paper presents a first attempt to thoroughly investigate all the above methods, evaluate them on the real-world collected DGA dataset involving 38 classes with 168,900 samples, and should provide a valuable reference point for future research in this field.