An Intelligent Fuzzy Graph-Based Framework for Network Threat Detection, Cyber Risk Assessment, and Adaptive Mitigation
The rapid growth of digital communication technologies, cloud computing, Internet of Things (IoT) devices, and interconnected network infrastructures has significantly increased the complexity and frequency of cyber threats, making accurate threat detection and intelligent cyber risk assessment essential for ensuring network security. Conventional machine learning-based intrusion detection techniques often experience limitations in handling uncertain network behaviours, evolving attack patterns, and complex relationships among communicating entities, resulting in reduced detection accuracy and increased false alarm rates. To address these challenges, this study proposes an Intelligent Fuzzy Graph-Based Framework that integrates fuzzy reasoning with graph-based network modelling to enhance network threat detection, cyber risk assessment, and adaptive mitigation. Initially, network traffic was collected and pre-processed through feature extraction and normalization to remove redundancy and improve data quality. Subsequently, the processed network entities are represented as a weighted fuzzy graph, where nodes denote network devices and edges represent communication relationships with fuzzy membership values reflecting uncertainty in network interactions. A fuzzy inference mechanism was then employed to calculate dynamic cyber risk scores by analysing multiple security indicators, while graph-based structural analysis captures communication dependencies and attack propagation patterns for improved threat identification. Based on the computed risk scores, the framework performs intelligent threat classification and prioritises adaptive mitigation strategies for high-risk network entities. The proposed framework was evaluated using benchmark cybersecurity datasets and compared with conventional models, including Support Vector Machine (SVM), Random Forest, XGBoost, and Graph Neural Network (GNN). Experimental results demonstrate superior performance, achieving an accuracy of 99.28%, precision of 99.12%, recall of 99.05%, F1-score of 99.10%, Matthews Correlation Coefficient (MCC) of 98.92%, and ROC-AUC of 99.36%. The framework substantially reduces false positive and false negative predictions while maintaining consistent detection capability across multiple cyber-attack categories. These findings confirm that the integration of fuzzy reasoning and graph-based analysis provides a robust, scalable, and reliable solution for intelligent cybersecurity systems, enabling accurate threat detection, continuous cyber risk assessment, and adaptive mitigation in dynamic and large-scale network environments.