• Dipali Ajagar
• H.R.Kulkarni
• Rajashri Gaikwad

The escalating sophistication and volume of cyber threats, particularly zero-day exploits and polymorphic malware, necessitates a fundamental shift from reactive, signature-based defenses to proactive, AI-driven security frameworks. Traditional Machine Learning (ML) models are inadequate due to their reliance on manual feature engineering and their inability to process complex, multi-modal data streams. This study addresses the critical fragmentation gap in current research by proposing a novel Hybrid Deep Learning framework integrating Convolutional Neural Networks and Long Short-Term Memory architectures.The framework utilizes a two-stream parallel architecture: the LSTM stream captures long-range temporal dependencies in sequential network traffic (NSL-KDD), while the CNN stream performs spatial feature extraction on malware code images (Maling). The features are combined in a Fusion Layer for unified classification. Rigorous evaluation demonstrates the model's superior performance and practical utility: the framework achieved 98.8% accuracy in Network Anomaly Detection and 97.5% accuracy in Malware Classification, culminating in a 98.1% Unified F1-Score. Crucially, the model maintained an exceptionally low False Positive Rate (FPR) of 1.2%, confirming its operational viability by minimizing alert fatigue in security environments. This low rate is primarily attributed to the feature of redundancy provided by the hybrid fusion layer. This research validates the indispensable role of unified DL architectures in building robust, proactive defense components. Future work is directed toward integrating Deep Reinforcement Learning (DRL) for autonomous adaptation against Adversarial AI attacks and deploying Explainable AI (XAI) to ensure model trust, human oversight (Human-in-the-Loop control), and adherence to ethical and legal compliance.

Artificial Intelligence, Deep Learning, Hybrid CNN-LSTM, Cyber Threat Detection, Network Anomaly Detection, Malware Classification, Explainable AI, False Positive Rate.

"The Review of Artificial Intelligence in Modern Cyber Threat Detection: An Analysis of Proactive Security Frameworks", International Journal of Emerging Technologies and Innovative Research (www.jetir.org), ISSN:2349-5162, Vol.13, Issue 1, page no.329-340, January-2026, Available :http://www.jetir.org/papers/JETIRHG06033.pdf

"The Review of Artificial Intelligence in Modern Cyber Threat Detection: An Analysis of Proactive Security Frameworks", International Journal of Emerging Technologies and Innovative Research (www.jetir.org | UGC and issn Approved), ISSN:2349-5162, Vol.13, Issue 1, page no. pp329-340, January-2026, Available at : http://www.jetir.org/papers/JETIRHG06033.pdf