Hybrid Network Attack Detection Using Snort and Deep Learning

The proliferation of sophisticated cyber attacks necessitates advanced intrusion detection systems that combine the reliability of signature-based methods with the adaptability of deep learning. This paper presents a novel hybrid framework integrating Snort's signature-based detection with deep learning architectures (ANN, CNN, LSTM, RNN) through our proposed ACLR (Adaptive Cross-Layer Recognition) mechanism. The framework dynamically allocates computational resources and detection strategies based on traffic characteristics, achieving 99.2% detection accuracy with 0.8% false positive rate on the CIC-IDS2017 and CSE-CIC-IDS2018 datasets. Our approach reduces preprocessing overhead by 47% through intelligent feature selection and maintains sub-10ms latency for traffic rates up to 40 Gbps. The adaptive weight allocation mechanism enables 93.7% detection rate for zero-day attacks while preserving computational efficiency for known threats. Experimental results demonstrate 4.3% improvement over state-of-the-art methods with 31.8% lower communication overhead compared to existing hybrid approaches.