Alsharif, FawzyBasaran, Emir CanAgca, Ahmet EmirTopcuoglu, Emirhan2026-01-312026-01-3120259798331554378https://doi.org/10.1109/FiCloud66139.2025.00072https://hdl.handle.net/20.500.12662/1057312th International Conference on Future Internet of Things and Cloud, FiCloud 2025 -- 2025-08-11 through 2025-08-13 -- Istanbul -- 214221This paper presents an AI-driven cybersecurity system tailored for 5G-enabled mini computers running the Pardus operating system. The system integrates real-time network monitoring using Suricata with dynamic threat intelligence from the Open Threat Exchange (OTX), allowing for the automated blocking of globally recognized threats. Detection is performed through a two-stage Random Forest model trained on the 5G-NIDD dataset, where 24 features were mapped from Suricata's traffic output. The binary classification model achieved 99.78% accuracy in distinguishing benign from malicious traffic, while the multiclass model identified specific attack types-such as HTTPFlood, SYNScan, and UDPFlood-with 97.48% accuracy. Designed with the computational constraints of edge environments in mind, the system incorporates lightweight, high-precision AI models and rule-based response mechanisms capable of executing threat-specific countermeasures in real time. Evaluation on a 5G-enabled mini computer demonstrates that the proposed approach delivers reliable detection performance with low latency and minimal resource usage, making it a practical solution for industrial IoT, smart infrastructure, and mobile edge deployments where autonomous and adaptive cyber defense is critical. © 2025 IEEE.eninfo:eu-repo/semantics/closedAccess5G technologyAIbased cybersecurityautonomous defense mechanismmini computersPardus OSthreat detection and classificationConference Object10.1109/FiCloud66139.2025.000722-s2.0-105021992179484N/A477