Deep Defense: A Hybrid LSTM-ANN Architecture for Intelligent Intrusion Detection System

The rapidly evolving cyber threat landscape requires innovative and adaptive intrusion detection solutions. Traditional signature-based intrusion detection systems, despite their high accuracy, are inherently inflexible and susceptible to evasion tactics, making them ineffective against novel and sophisticated attacks. In contrast, machine learning-powered anomaly-based intrusion detection systems demonstrate improved detection capabilities, but their advantages are often outweighed by substantial drawbacks, including significant computational resource requirements when the dataset and its features for behavior analysis increase. Also, there are high power consumption and latency issues in communication that compromise network performance. This study proposes a hybrid ANN-LSTM approach for performing deep packet inspection and efficiently detecting malicious network packets. By analyzing binary container data, our model reduces time and space complexity, thereby mitigating the computational resources and power consumption challenges but also increasing the efficiency of intrusion detection by 3%, making it more intelligent than various established IDS.