AI-Based Cyber Threats Detection System

Cyber threats such as phishing, malware, brute-force login attempts, denial-of-service activity, and suspicious networkbehaviour challenge organisations that depend on connected systems. Signature-based tools can identify known patterns butmay struggle with unfamiliar or rapidly changing attacks. This paper proposes an AI-Based Cyber Threats Detection Systemthat analyses network-flow and system-event features to flag potentially malicious activity. The design combines datapreprocessing, supervised machine-learning classification, anomaly detection, risk scoring, and an alert dashboard. Theproposed pipeline covers authorised log collection, data validation, feature preparation, model training, evaluation, andanalyst review. A reproducible evaluation plan uses precision, recall, F1-score, false-positive rate, detection latency, andconfusion-matrix analysis. No experimental performance results are claimed because the system has not been benchmarkedin a controlled deployment. The design emphasises privacy, explainability, human review, and responsible use of securitydatasets. It can serve as a foundation for a student prototype and can later be extended with streaming ingestion and threatintelligence enrichment.