Breaking bidder collusion in large-scale spectrum auctions.

ABSTRACTDynamic spectrum auction is an effective solution to provide spectrum on-demand to many small wireless networks. As the number of participants grows, bidder collusion becomes a serious threat. In this paper, we study bidder collusion in large-scale spectrum auctions, investigating its impact on auction outcomes. We found that the nature of the complex interference constraints among bidders provides a fertile breeding ground for colluders, causing significant damage in auction efficiency and revenue. In particular, collusion group of small size plays a dominant role since it is easy to form and hard to be detected. We propose Athena, a new collusion-resistant auction framework for large-scale dynamic spectrum auction. Athena implements a soft collusion resistance, allowing the auctioneer to exploit the tradeoff between the level of collusion resistance and the cost of achieving such level of resistance. Unlike existing solutions, Athena enables spectrum reuse across bidders, achieves soft collusion resistance against any form of collusive bidding strategy, maintains provable revenue guarantee, and does so with polynomial-time complexity. To provide a comprehensive evaluation, we first analytically prove Athena's collusion resistance and revenue guarantee (under any bids), and then experimentally verify our analytical conclusions using empirical bid distributions.