Design and experimental validation of an artificial intelligence driven network security situation awareness system

Network security faces escalating challenges from sophisticated cyberattacks, requiring proactive and intelligent defense systems. Traditional intrusion detection techniques often fail to provide real-time awareness. This research presents an artificial intelligence (AI)-driven framework to enhance situation awareness (SA) and strengthen adaptive responses against emerging security threats. A linear support vector machine mutated with marine predator algorithm optimization (LSVM-MPAO) is proposed for threat classification. Datasets comprising 2,500 labeled traffic flows, logs, and malicious activity signatures were employed, specifically benchmarked with AI-driven network security. Collected raw data is preprocessed using robust scaling with outlier handling for duplicates, missing values, and noise, followed by Z-score normalization to ensure consistent feature scaling. To reduce dimensionality while maintaining discriminative characteristics, principal component analysis (PCA) is utilized. Reinforcement learning (RL) models’ successive defensive methods to support adaptive decision-making. LSVM classifies network traffic using an optimal hyperplane, while the MPA optimizes its parameters, enhancing accuracy, convergence, and robustness for detecting diverse and complex network cyber threats effectively. The suggested model was implemented in Python, which shows the results of a mean square error (MSE) of 0.0003, the mean absolute percentage error (MAPE) of 0.03, the true detection rate (TDR) of 93.84%, the false detection rate (FDR) of 3.62%, and the root mean square error (RMSE) of 0.0172 in forecasting network security threats compared to the baseline systems. The AI-driven network security SA system demonstrates effective anomaly detection and adaptive decision support. Additionally, it improves cybersecurity (CS) with intelligent, proactive, and scalable defense solutions.