Security Threat Detection and Defense Theory of Intelligent Operation and Maintenance System for Rail Transit Based on Adversarial Machine Learning

Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).

The intelligent operation and maintenance system (IOMS) for rail transit is a safety control system based on IoT, big data, and automated control. Its open interconnectivity and multi‐source heterogeneous data nature expose rail transit to covert adversarial attacks such as data tampering, model poisoning, and decision interference, which traditional security models cannot effectively address. This study proposes the PGD‐CLA‐ADS model, an adversarial machine learning‐based security threat detection and defense framework. The model generates simulated threat scenarios via enhanced PGD attack samples, employs a CNN‐LSTM‐attention multi‐dimensional feature extraction network to capture attack characteristics, and designs an adaptive defense system for real‐time threat response. Experimental results demonstrate that the model achieves 95.8% threat detection accuracy on public datasets with a false positive rate of 2.1% and an attack response latency within 0.3 s.