Enhanced Signature Based Intrusion Detection System Using Hyper Scalar Feature Engineering With LSTM Gated Recurrent Neural Network

In the fast-changing world of cybersecurity, cyber threats have been challenging the traditional defence mechanisms in the signature-based Intrusion Detection Systems (IDS). Although these systems are effective for detecting known threats and cannot handle advanced, unknown and evasion-based attacks. The proposed work presents an enhanced signature-based IDS framework to bridge the gap of conventional approaches toward detecting advanced persistent threats and provide timely responses to security incidents. The proposed methodology uses hyper-scaler feature engineering with a Long Short-Term Memory Gated Recurrent Neural Network (LSTM-GRNN) improves the efficacy and accuracy in intrusion detection. The approach pre-processes to start with the min-max normalization by ensuring uniform scaling of feature values. A new technique named Intrusion Behavior Feature Pattern Impact Rate (IBFPIR) is proposed to determine the relevance of feature patterns that are more related to intrusion behavior in malicious activities. For optimization of feature selection, a new advanced optimization approach such as Simplified Whale Optimization Algorithm (SWOA) is used for information gain while minimizing redundancy and reducing the dimensionality along with superior model performance. Finally, the LSTM-GRNN architecture is applied to classify intrusion behaviors based on the refined features. The long-term dependencies in time-series data captured by the LSTM combined with gated recurrent units is used to learn patterns during intrusion detection. The proposed system gives a better performance interms of accuracy (97%), precision (98%), recall (97%), F1 score (98%), with reduced false positive rate (FPR of 4%) and false negative rate (FNR of 5%) compared with existing models. The proposed work gives a significant development in intrusion detection systems in safeguarding sensitive data against cyber threats.