A deep recursively learning LSTM model to improve cyber security Botnet attack intrusion detection

The occurrence of botnets over the network is crucial as it shows advent effect on various applications like finance, cyber-security, and healthcare application. Botnets are refined and more dangerous in their functionality over the network model. Most of the prevailing models and flow- and rule-based models feel challenging to predict the bot functionalities in a preventive manner. Therefore, the modeling of efficient and automated botnet detection approaches is highly essential. This research concentrates on modeling a novel botnet detection approach based on the recursively analyzing the flow of features of the network nodes spatially and temporally where the attack samples are intra-dependent time-series data. The hierarchical structural design of the network helps to integrate various levels of feature information and learns the spatial and temporal information automatically among the adjacent network connection. This process is carried out by the proposed architectural model known as Recursively Learning Long Short-Term Memory over spatial and temporal ((RLTM)-T-2-st). Thus, the bot activities are detected by recursively analyzing the limited number of nodes. The (RLTM)-T-2-st model is modeled to improve the efficiency of the network by eliminating unnecessary activities. The proposed model is validated using the online accessible CTU-13 dataset and benchmarked against the prevailing classification approaches for botnet detection. The simulation is done in a MATLAB environment, and the outcomes work efficiently and evaluated with prevailing models to project the significance of the (RLTM)-T-2-st model.