NSCTI: A Hybrid Neuro-Symbolic Framework for AI-Driven Predictive Cyber Threat Intelligence

NSCTI: A Hybrid Neuro-Symbolic Framework for AI-Driven Predictive Cyber Threat Intelligence

연구 분야: Safety

학회: 2025 4th International Conference on Computational Modelling, Simulation and Optimization (ICCMSO)

With the rapid evolution of cyber threats, traditional Artificial Intelligence (AI)-driven security models often fail to provide real-time, interpretable, and adaptive threat intelligence. This paper proposes Neuro-Symbolic Cyber Threat Intelligence (NSCTI), a novel hybrid framework that integrates Deep NeuroSymbolic Learning (DNSL) with Graph-Based Threat Reasoning (GBTR) to enhance predictive cybersecurity analytics. The NSCTI framework comprises three core components: Hybrid Deep Learning-Based Threat Detection (HDL-TD), which leverages Graph Neural Networks (GNNs), Long Short-Term Memory (LSTM), and Hidden Markov Models (HMMs) for dynamic attack pattern recognition; Neuro-Symbolic Adversarial Defense (NSAD), which employs reinforcement learning-driven adversarial resilience mechanisms to mitigate evolving cyber threats; and Trust-Aware Federated Cyber Intelligence (TFCI), which utilizes federated learning (FL) and blockchain-based threat sharing to ensure secure, decentralized CTI. Experimental evaluations on benchmark datasets (CICIDS2017, UNSW-NB15, and N-BaIoT) demonstrate that NSCTI achieves 99.7% attack detection accuracy, a 0.5 % false positive rate (FPR), and a 35% reduction in computational overhead compared to existing cybersecurity frameworks. Security analysis confirms NSCTI's robustness against Man-in-the-Middle (MITM), replay attacks, and adversarial perturbations, making it a scalable and proactive cyber defense solution. Future research will explore quantumenhanced neuro-symbolic learning and self-adaptive reinforcement learning to further strengthen autonomous cybersecurity in large-scale IoT networks.