Enhancing Network Security: A Modular Neural Network Approach to Detect Suspicious Patterns

The increasing complexity and sophistication of cyber threats pose significant challenges to network security. Detecting suspicious patterns in network traffic-such as unusual activity, unexpected communications, and atypical data patterns—is critical for identifying potential cyber-attacks and unauthorized network use. In this study, we propose a modular neural network approach, integrating Recurrent Neural Networks (RNN) and Nonlinear AutoRegressive eXogenous (NARX) models, to effectively detect these anomalies. The modular structure divides the problem into manageable subproblems, allowing for enhanced accuracy and flexibility. The RNN’s ability to process sequential data, combined with the NARX model’s capacity to capture both internal and external dependencies, makes this hybrid approach particularly effective in analyzing complex network traffic patterns. We trained and tested the proposed model using both typical and atypical network traffic data, achieving a mean squared error close to zero during the training phase. This research contributes to the development of more resilient and adaptive cybersecurity solutions, capable of safeguarding networks against evolving threats.

• Sotir Sotirov
• Vladimir Poulkov
• Agata Manolova

2024 27th International Symposium on Wireless Personal Multimedia Communications (WPMC)

https://ieeexplore.ieee.org/abstract/document/10863532