Enhanced Real-Time Detection of Cyber Threats Through Adaptive Machine Learning in Network Traffic Analysis
A. Yovan Felix, Marshal Dionee L, Marimuthu Pandian Ma, M Bhuvanesh, J. Albert Mayan, Benito. Y · 2025
The technological advancements in security systems have made it easier for them to detect urgent cyber threats before implementing the protective strategies. Traditional Network Intrusion Detection Systems (IDS) are based on claims that have been employed over the years, namely static rule-based methods that depend upon past system rules and are unable to cope with the new threats. Through traffic correlation analysis, real time network traffic analysis is proposed using a machine learning based system that could effectively detect current cyber-attacks using LSTM and RNN models. Using the proposed system, an amazing 95.4% accuracy with the fewest possible false positives (92.1%) and low detection latency of 120 ms is achieved, significantly better than traditional IDS methods, e.g. 88.3% accuracy and detection latency of 500 ms. The model has packet size, flow duration, source/destination IP addresses, transmission protocols as input features. Thus, our system shows higher detection accuracy, faster response time, and more comprehensive pattern of attack recognition than standard IDS. Therefore, the research focuses on the prospects of adaptive machine learning schemes for creating advanced defense solutions for real-time network security infrastructure.