Self-configuring intrusion detection using adaptive goal target optimization based deep Bi-LSTM in blockchain networking systems

Intrusion detection systems (IDS) have a significant positive impact on the detection and mitigation of potential security breaches and attacks. Traditional IDS environments examine the information that is available for malicious detection, which clearly shows that human analysis and attempts lead to system failure. In this research, the Adaptive Goal Target Optimization based Deep Bi-LSTM (AGLSTM) is proposed for self-configuring intrusion detection in blockchain networking systems. Data security and network security are the necessity for the heterogeneous source data communicating in the blockchain networks, which is ensured through the developed model. The heterogeneous data is pre-processed to remove missing values and duplication, and the data balancing issues are handled using an optimized SMOTE framework. The balanced data is communicated with the AGLSTM classifier for self-configuring intrusion detection, where the proposed adaptive goal-target optimization is developed using the conventional hybridization of rider and cheetah optimization using their faster convergence phenomenon. The accuracy, sensitivity, and specificity of the proposed AGLSTM classifier are measured as 99.93%, 99.92%, and 99.94% for the UNSW-NB 15 dataset using Training Percentage (TP) analysis and 99.63%, 99.73%, and 99.51% for k-fold analysis. Similarly, with the BoT-IoT dataset, the AGLSTM classifier achieved 99.70% accuracy for TP analysis and 99.70% for k-fold analysis, outperforming other approaches.