Transactions on Hybrid Neural Architectures & Threat Coordination

Dr. S. Jumlesha 1
Professor, Department of AIDSAnnamacharya Institute of Technologyand Sciences, Tirupati – 517520, A.P.
ahmedsadhiq@gmail.com
N. Nikhil 4
Department of AIDSAnnamacharya Institute of Technology
and Sciences, Tirupati – 517520, A.P.nandanavanamnikhil@gmail.com
SK. Naseer Hussain 2
Department of AIDSAnnamacharya Institute of Technology
and Sciences, Tirupati – 517520, A.P.naseerhussain7739@gmail.com
C. Pranitha 5
Department of AIDSAnnamacharya Institute of
Technology and Sciences, Tirupati –517520, A.P.pranithachiyyavaram@gmail.com

T. Sarath Kumar Reddy 3
Department of AIDSAnnamacharya Institute of
Technology and Sciences,Tirupati – 517520, A.P.
sarathreddy9377@gmail.com

Abstract— The modern cyber threats are becoming increasingly complex and cunning, combining high volume statistical signals (e.g., DDoS-like traffic surges) with complex contextual meaning (e.g.,sophisticated exploitation techniques) to evade detection by conventional detection systems. The traditional Intrusion Detection System (IDS), based on a single detection mechanism, cannot detect all these complex intrusion scenarios, resulting in false positives and false negatives. To overcome this weakness of traditional Intrusion Detection Systems, we propose Sentinel-GAN, a Modular Neural Orchestration Framework for Autonomous Threat Synthesis and Proactive Intelligence. In our paper, we propose a Hybrid Ensemble Architecture by mathematically combining a Generative Adversarial Network (GAN)with a DistilBERT model. Using our proposed Domain Specific Selection Strategy, continuous variables such as packet sizes and durations are fed into the GAN model to detect statistical anomalies, while categorical variables such as ports and protocols are converted into natural language and fed into the DistilBERT model to detect semantic malicious intent. The two independent