Visible to the public Biblio

Filters: Author is Liu, Xuan  [Clear All Filters]
2020-10-14
Song, Yufei, Yu, Zongchao, Liu, Xuan, Tian, Jianwei, CHEN, Mu.  2019.  Isolation Forest based Detection for False Data Attacks in Power Systems. 2019 IEEE Innovative Smart Grid Technologies - Asia (ISGT Asia). :4170—4174.
Power systems become a primary target of cyber attacks because of the vulnerability of the integrated communication networks. An attacker is able to manipulate the integrity of real-time data by maliciously modifying the readings of meters transmitted to the control center. Moreover, it is demonstrated that such attack can escape the bad data detection in state estimation if the topology and network information of the entire power grid is known to the attacker. In this paper, we propose an isolation forest (IF) based detection algorithm as a countermeasure against false data attack (FDA). This method requires no tedious pre-training procedure to obtain the labels of outliers. In addition, comparing with other algorithms, the IF based detection method can find the outliers quickly. The performance of the proposed detection method is verified using the simulation results on the IEEE 118-bus system.
Ou, Yifan, Deng, Bin, Liu, Xuan, Zhou, Ke.  2019.  Local Outlier Factor Based False Data Detection in Power Systems. 2019 IEEE Sustainable Power and Energy Conference (iSPEC). :2003—2007.
The rapid developments of smart grids provide multiple benefits to the delivery of electric power, but at the same time makes the power grids under the threat of cyber attackers. The transmitted data could be deliberately modified without triggering the alarm of bad data detection procedure. In order to ensure the stable operation of the power systems, it is extremely significant to develop effective abnormal detection algorithms against injected false data. In this paper, we introduce the density-based LOF algorithm to detect the false data and dummy data. The simulation results show that the traditional density-clustering based LOF algorithm can effectively identify FDA, but the detection performance on DDA is not satisfactory. Therefore, we propose the improved LOF algorithm to detect DDA by setting reasonable density threshold.