Traffic Classification For Botnet Detection Using Deep Learning

In this modern world, we are naturally dependent on technology, which is constantly evolving. At the same time, our fear of data fraud is increasing every day. The ever-increasing number of attacks on our servers indicates that people have learned to adapt to new technologies. In addition, the number of botnet attacks is not insignificant, and botnets threaten computer networks. This can impact security systems, including malware distribution, phishing, spamming, and click fraud. Due to their harmful effects, botnets must be identified as soon as possible. Due to the dynamic character of botnets, however, detection has proven difficult. In this proposed method, deep learning is used to analyze and predict botnet data, and model performance metrics such as accuracy, precision, Recall, and F1 score are compared to prior work. In the proposed procedure, patterns are also discussed by analyzing botnet data. Various scenarios and sensors separate the NCC and NCC-2 datasets. The NCC dataset is separated into 13 scenarios, while the NCC-2 dataset is separated into 3 sensors. The dataset is preprocessed using the clean method, the null value handling method, the imbalance method, and the feature selection method to extract the essential characteristics. The dataset is then trained using LSTM, GRU, and Bi-LSTM models, and 99.93% accuracy is achieved.