In this paper, we have argued for the use of tamper-resistant hardware to build secure auction markets. This is a conceptual design - performance will be a key issue in any actual implementation. It is not only our architecture which impacts on performance, but because trusted hardware typically lags behind off-the-shelf commercial processors. We believe our architecture can be highly successful using existing trusted hardware platforms.
A particularly intriguing area for future research is the actual question of how we distribute work in the case of multiple simultaneous auctions (potentially with overlapping groups of bidders and overlapping groups of sellers.) Such a solution is complicated by the requirement that prices be allocated fairly. (While third party arbitrage could help ensure this, we suspect that it is possible to realize this directly in the mechanism design.)
Regardless of the details of implementation, trusted hardware can certainly provide a proven fair approach to realizing auctions. The advantages extend across a wide-range of auction types - from small, ad hoc auctions to large ongoing auction markets. Given the degree of fraud possible with existing auction architectures, there is a clear case for using trusted hardware to ensure fairness and security.