Researchers win best paper for smart meter security research at QEST 2015


Kim Gudeman, Information Trust Institute

Graduate student Varun Badrinath Krishna, Professor William H Sanders, and Information Trust Institute Research Scientist Gabriel Weaver have received the best paper award for their work on smart meter security at the 12th International Conference on Quantitative Evaluation of Systems. The conference was Sept. 1-3 in Madrid.

Increasingly, electric utilities are deploying smart meters to improve power delivery service. However, these meters – which use Internet-based communications networks to report data -- may be vulnerable to cyber attacks, whether the motive be pilfering  electricity or destabilizing the power market. The paper, “PCA-Based Method for Detecting Integrity Attacks on Advanced Metering Infrastructure,” proposes combining two unsupervised learning methods in a unique way to help verify intrusion attempts.

QEST Program Committee Chair Javier Campos and Varun Badrinath Krishna at QEST.
QEST Program Committee Chair Javier Campos and Varun Badrinath Krishna at QEST.

It is believed to be the first time the two methods -- called Principal Component Analysis and Density-based Spatial Clustering of Applications with Noise – have been combined and successfully applied in computer security and smart grid research. 

"This is my first research paper as the first author," said Krishna, who also conducts research at the Information Trust Institute. "When the conference Program Committee chairs announced that we had won the Best Paper Award, I really froze in disbelief. When members of the Program Committee came to congratulate me, I told them that it must just be beginner's luck. But they said it was a really good paper and stood out on its merit, despite tough competition from many great researchers."

Cyber intruders attack by compromising the Advanced Metering Infrastructure (AMI), a network of smart meters that measure electricity consumption. The authors propose to identify intrusion by leveraging a unique aspect of electricity consumption: its repetitive nature. For example, the consumption of a university lecture hall at 3 p.m. on a Tuesday is likely very similar to the level at 3 p.m. on previous Tuesdays.

In identifying intrusions, the model outperforms the “Average Detector” proposed in related work. Future work will focus on improving false positive results. For example, it is difficult to discern a malicious intrusion (someone stealing electricity) from a legitimate anomaly (someone traveling and not consuming much electricity at home).

The researchers conducted their work using a dataset of real smart-meter readings obtained from Ireland’s Commission for Energy Regulation. The paper was based on research funded by the U.S. Department of Energy.

Along with Krishna, Weaver and Sanders, head of the Department of Electrical and Computer Engineering at Illinois, recently submitted a project proposal that was selected for funding by the Siebel Energy Institute. Read more about the project here.