Shill bidding is the most severe and persistent type of auction fraud where bidders place artificial bids to inflate the final price of an online auction. Attempting to detect shill bidding is quite a challenging task as users can easily register in an auction system by providing a false identity. Most existing shill detection techniques wait until an auction has finished before taking action. However, this situation means that innocent bidders will have already been cheated before the shill bidder has been detected. Therefore, there is a pressing need to introduce effective mechanisms for detecting shill bidding while an auction is in progress in order to take immediate actions to prevent innocent bidders from becoming victims. In this paper, we propose a mechanism for detecting shill bidders in real-time. The algorithm builds a case against suspect bidders by examining their behaviour during the current auction and also uses evidence from their past behaviour across multiple auctions. The algorithm is then able to take appropriate actions towards the suspected shill bidders accordingly. Experimental results using simulated and commercial auction data show that our proposed algorithm can potentially highlight shill bidding attempts during online auctions with 99.4% detection accuracy on average.