Abstract:

Influenced by the transformation of power system to an integrated energy one and the evolution of network attack technology, power information security and protection is getting mounting challenges. Network intrusion detection system （NIDS）can identify weaknesses in the power system. Pursuing detection accuracy, but neglecting the missing data in manual feature extraction is the problem in the course of improving the traditional algorithm. Thus, an intrusion detection method based on the Gated Recurrent Unit （GRU） and Convolutional Neural Network （CNN） is proposed. The method uses GRU to extract the original time series features, takes CNN to obtain multi-dimensional features, and realizes the mapping of abnormal traffic with Softmax classifier. The method has been practiced in the training of two experimental models, KDD99 data set and False Data Injection Attack （FDIAs）. The results show that the method performs better in classification and detection accuracy than the traditional one, which verifies the effectiveness and practicability of the method.

Key words: integrated energy, intrusion detection, cyber security, deep learning, GRU, CNN, cloud computing,big data,IoT,mobile internet,AI and blockchain