Reputation attacks detection for effective trust assessment among cloud services

Abstract

Consumers' feedback is a good source to help assess overall trustworthiness of cloud services. However, it is not unusual that a trust management system experiences malicious behaviors from its users (i.e., collusion or Sybil attacks). In this paper, we propose techniques for the detection of reputation attacks to allow consumers to effectively identify trustworthy cloud services. We introduce a credibility model that not only identifies misleading trust feedbacks from collusion attacks but also detects Sybil attacks, either strategic (in a long period of time) or occasional (in a short period of time). We have collected a large collection of consumer's trust feedbacks given on real-world cloud services (over 10, 000 records) to evaluate and demonstrate the applicability of our approach and show the capability of detecting such malicious behaviors.