Cyber Threats Detection Based on Artificial Neural Networks Using Event Profiles

The rapid proliferation of interconnected systems has elevated the sophistication of cyber threats far beyond what signature-based defenses can address. This paper introduces a novel framework that employs Artificial Neural Networks (ANNs) trained on structured Event Profiles — rich, multi-attribute representations that aggregate raw network events into semantically meaningful units encompassing temporal signatures, entity context, behavioral vectors, and risk metadata. Unlike raw event streams, Event Profiles encode higher-order relationships across correlated activities, providing the ANN with contextually enriched input that dramatically improves discrimination between benign and malicious behavior. Evaluated on the CICIDS-2017 and NSL-KDD benchmarks, the framework achieves a mean detection accuracy of 97.8% with a false-positive rate of 2.6%, outperforming all evaluated baselines including LSTM-based approaches. End-to-end alert latency is reduced to 12 ms through an event-profile micro-service pipeline that integrates natively with enterprise SIEM platforms.