정적메모리 모니터링을 통한 딥러닝 기반 원자력 발전소내 사이버 공격 탐지

초록

The Device that include modules used within a nuclear power plant is regulated to employ static memory rather than dynamic memory, as per the U.S. NRC (Nuclear Regulatory Commission) regulatory requirements. Therefore, this research proposes a deep learning-based attack detection system to identify normal and abnormal states in accordance with changes in static memory usage. The proposed system visualizes time-series data of memory usage using Markov Transition Field (MTF) and employs LSTM Auto-Encoder (AE) for attack detection. It further proposes a model to classify attack types by visualizing the reconstructed data and analyzes its performance. The analysis results demonstrate that the proposed system effectively detects and classifies attacks in static memory systems.

키워드