EXPLAINABLE AGENTIC AI FOR REAL-TIME CYBERATTACK DETECTION AND AUTONOMOUS RESPONSE IN PAKISTAN’S CRITICAL DIGITAL INFRASTRUCTURE
Keywords:
Agentic artificial intelligence, Explainable AI, Cyberattack detection, Autonomous cyber defence, Critical infrastructure, Human oversight.Abstract
The increasing digitalization of Pakistan’s critical infrastructure has intensified exposure to sophisticated cyber threats, highlighting the limitations of conventional intrusion detection and response systems. This study proposed an explainable agentic artificial intelligence (AI) framework for real-time cyberattack detection and autonomous response. The framework integrated machine-learning-based threat detection, explainable AI techniques (SHAP and LIME), coordinated intelligent agents, and human-supervised response mechanisms. A quantitative experimental design was proposed using publicly available cybersecurity datasets, including CICIDS2017, UNSW-NB15, and ToN-IoT, within a controlled cyber-range environment. Detection performance was evaluated through accuracy, precision, recall, F1-score, and false-positive rate, while operational effectiveness was assessed through detection latency, response time, threat containment, and service continuity. The results indicated potential improvements in detection performance, response efficiency, and operational reliability compared with conventional intrusion detection approaches. The study contributes to socio-technical systems theory by emphasizing the integration of intelligent automation, explainability, and human oversight in cybersecurity. The proposed framework offers a potential foundation for strengthening cyber resilience across Pakistan’s critical digital infrastructure.