A Bidirectional GRU-Integrated Seq2Seq–ConvLSTM Framework for Improved Network Intrusion Detection

Ganga Bhavani Billa; Navyasri Vipparthi

doi:10.62647/

Authors

Ganga Bhavani Billa Associate Professor, Department of CSE, Bonam Venkata Chalamayya Engineering College, India Author
Navyasri Vipparthi PG Scholar, Department of CSE, Bonam Venkata Chalamayya Engineering College, India Author

DOI:

https://doi.org/10.62647/

Keywords:

Network Intrusion Detection, Hybrid Deep Learning, Seq2Seq, ConvLSTM, Bidirectional Layer, GRU, Temporal-Spatial Features, Real-Time Detection, Flask Deployment, Cybersecurity, Deep Sequential Models, Feature Optimization, Anomaly Detection, Intrusion Prediction, Neural Networks.

Abstract

In order to improve computational efficiency and train temporal-spatial features more effectively, this study expands the hybrid Seq2Seq-ConvLSTM intrusion detection model by adding GRU and Bidirectional layers. The GRU layer expedites training and prediction without compromising accuracy, while the Bidirectional layer improves identification of complicated sequential attack patterns by capturing forward and backward temporal relationships. A web interface built on Flask allows users to submit test datasets and view classification results instantaneously, enabling real-time intrusion detection. This is how the upgraded model is implemented. Results from experiments reveal that the suggested extended hybrid strategy performs better than the baseline Seq2Seq-ConvLSTM and Random Forest models in contemporary network security settings, in terms of accuracy, latency, and resilience.