Beyond M-score for detection of misusability by malicious insiders

As data is very important for any enterprise, prevention of misuse of data in the context of data mining assumes much importance of late. There are many threats to the privacy of data from insiders and outsiders. In fact, prevention of misuse of data is very challenging problem to be addressed. Towards this end Harel et al. proposed a measure named “M-score” which uses sensitivity weights in order to estimate misusablity. It can tell the risk level when data is exposed for mining. However, the sensitivity related aspects of this measure and distinguishing factors can be further improved. In this paper, we improve M-score to support multiple distinguishing factors and many combinations of sensitivity levels so as to make it more useful and accurate for estimating misusability of data before outsourcing it to third party for data mining. The prototype built by us demonstrates the usefulness of the proposed improvements over M-score. The empirical results revealed that our work has significant improvement in measuring misusability of data.