A Game Theoretical Analysis On The Effectiveness Of Sybil Attacks In Social Networks

Recently，online social network has been a research hotspot. Theexisting works on thwarting sybil attacks mainly focus on sybil detectionwhile pay little attention to evaluating the impact of sybil attacks on normalusers as well as the whole network. In this work, we begin with anempirical study towards Dianping.com, a Chinese top review site. Byperforming sybil detection as well as regression modelling, it demonstratesthat there is a strong correlation between the sybil reviews and normalreviews. However, the dependency pattern varies with different qualitytypes of restaurant.We present a game-theoretical analysis which models the interactionof sybil and normal user as the sybil game.The results of the simulation demonstrate that the effectiveness of thesybil attacks are not only affected by the quantity of sybil reviews, but alsovaries with the different market shares of restaurants (low quality, mediumquality and high quality). Another discovery by our simulations is that thecustomer’s utility will maintain unchanged when the quantity of sybilreviews reach a certain level, which indicates that the sybil reviews havelost customer’s trust in the long term although it works at the beginning justas the proverb says ’’lies have short legs’’. As a result, according to therelationship between the impacting factor in the simulation, some suggestshave been given for the defence of the Sybil attack.