# Detecting Multiple Seller Collusive Shill Bidding

## TL;DR

This paper extends a reputation-based system to detect collusive shill bidding involving multiple sellers, adjusting bidder scores to account for collusion and demonstrating effectiveness through simulated auction data.

## Contribution

It introduces an algorithm to identify seller collusion and modifies the Shill Score to mitigate collusion effects, enhancing detection accuracy.

## Key findings

- Algorithm successfully detects seller collusion in simulated data
- Adjusted Shill Scores reduce false positives caused by collusion
- Experimental results validate the effectiveness of the proposed method

## Abstract

Shill bidding occurs when fake bids are introduced into an auction on the seller's behalf in order to artificially inflate the final price. This is typically achieved by the seller having friends bid in her auctions, or the seller controls multiple fake bidder accounts that are used for the sole purpose of shill bidding. We previously proposed a reputation system referred to as the Shill Score that indicates how likely a bidder is to be engaging in price inflating behaviour with regard to a specific seller's auctions. A potential bidder can observe the other bidders' Shill Scores, and if they are high, the bidder can elect not to participate as there is some evidence that shill bidding occurs in the seller's auctions. However, if a seller is in collusion with other sellers, or controls multiple seller accounts, she can spread the risk between the various sellers and can reduce suspicion on the shill bidder. Collusive seller behaviour impacts one of the characteristics of shill bidding the Shill Score is examining, therefore collusive behaviour can reduce a bidder's Shill Score. This paper extends the Shill Score to detect shill bidding where multiple sellers are working in collusion with each other. We propose an algorithm that provides evidence of whether groups of sellers are colluding. Based on how tight the association is between the sellers and the level of apparent shill bidding is occurring in the auctions, each participating bidder's Shill Score is adjusted appropriately to remove any advantages from seller collusion. Performance has been tested using simulated auction data and experimental results are presented.