Network Intrusion Detection using Auto-encoder Neural Networks and MLP

With the increasing reliance on networked systems across various domains, ensuring the security of these networks has become paramount. Intrusion Detection System plays a crucial role in safeguarding networks by identifying and mitigating malicious activities. Traditional IDS approaches often struggle with the complexity and variability of modern network attacks. In this study, we propose a novel approach leveraging Auto-encoders in conjunction with Multi - Layer perceptron for network intrusion detection. Three distinct algorithms were implemented and evaluated: Auto-encoders, Auto-encoders combined with Multilayer Perceptron, and Convolution Neural Networks. The experiments demonstrated promising results, showcasing the effectiveness of each algorithm in detecting network intrusions. This multi-algorithmic approach not only enhances detection accuracy but also broadens the scope of potential threats identified. The findings underscore the significance of employing a diverse set of machine learning techniques in network security which is effective for anomaly detection with less labeled data, finding hidden patterns in complex network traffic, and potentially working well with imbalanced data.