Rumor containment in peer-to-peer message sharing online social networks

Rumors in online social networks (OSNs) create social chaos, financial losses, and endanger property, which makes rumor containment an important issue. We consider an OSN in which the users communicate via private peer-to-peer messages. We consider the proposed peer-to-peer linear threshold (PLT) and peer-to-peer independent cascade-variant (PICV) models for information diffusion in OSNs, which are variants of the classic IC and LT models, respectively. To combat the rumor spread in the OSN with peer-to-peer message sharing, we employ blocking and positive information diffusion strategies. While in blocking strategy, few users of the OSN called the blocked seed nodes are blocked from spreading the rumor, in positive information diffusion strategy, correct information is introduced into few users of the OSN called positive seed nodes. The positive seed nodes further spread the correct information to other users with time. For a given time-period called the rumor-relevance interval, we determine average number of rumor-influenced nodes for the random, the max-degree, the greedy, the proximity heuristic, and the proposed proximity-weight-degree (PWD)-based containment seed node selection schemes for both blocking and positive information diffusion strategies for PLT and PICV models. We compare the effect of the rumor-relevance interval duration and number of seed nodes on the average number of rumor-influenced nodes for different seed selection algorithms. Our experimental results show that proximity-weight-degree-based seed selection algorithm performs on par with the high-complexity greedy scheme.