AI-DRIVEN INTRUSION DETECTION SYSTEM USING AUTOENCODERS AND LSTM FOR ENHANCED NETWORK SECURITY

Abstract

The increasing complexity of cyber threats has necessitated the development of advanced Intrusion Detection
Systems (IDS) capable of detecting both known and novel network attacks. Traditional rule-based IDS methods
often struggle with the rapid evolution of attack strategies. This paper proposes an AI-driven IDS using
Autoencoders for feature extraction and Long Short-Term Memory (LSTM) networks for classification. The
Autoencoders perform unsupervised anomaly detection by identifying deviations from normal network behavior,
while LSTM networks capture the temporal patterns of network traffic. The integration of these techniques,
enhanced by cloud computing, allows for efficient real-time processing of large-scale network data. Experimental
results show that the proposed system improves detection accuracy and scalability, offering a robust solution to
evolving network security challenges. However, further refinement is required to address issues of overfitting and
improve generalization.