Due to the increasing number of VoIP users, spam call has a potential to become a more severe problem than email spam because of the real time processing requirements of voice communication. We propose a trust-based mechanism that uses duration of calls between users to distinguish legitimate callers and spammers. Our trust computation algorithm is motivated by the simple observation that a legitimate user has long call duration and bidirectional communication while a spammer has short call duration and unidirectional communication. Trust value is automatically assigned to each user and adjusted by human calling behavior. It allows avoiding biasing problems that occur when one good user incorrectly rates another good user as a bad user. We also propose the trust inference mechanism in order to calculate a trust value for an unknown caller to a callee.
We also introduce a method to subscribe to an advertisement service for some users while preserving spam prevention for other users. From the perspective of a VoIP service provider, spam callers are also a type of customer and sometimes they are even valuable for increasing revenue. For example, some organizations may purchase an advertising service for broadcasting their information to targeted customers. Targeted customers who prefer to subscribe to this content will classify this call as a legitimate call while other users will classify this call as a spam call. If a service provider fails to design ways to manage spam calls that allow for such services, they can cause a loss of revenue when a system blocks all suspect traffic.
Since there is no VoIP corpus available for testing detection mechanism, we verify the detection accuracy of our proposal by simulating a realistic calling model based on VoIP usage statistics from a telecommunication company. The preliminary simulation results show that our approach is effective to assign trust for users in a VoIP network. Finally, we present a detailed sensitivity and specificity that confirm the high accuracy rate of the detection system.