Privacy Protection with Customized Java Sandbox Architecture

概要

論文の詳細を見る
Multi-party cooperative problem solving techniques require multiple inputs from multiple participants. The participants may want to keep their inputs secret from one another. However in traditional agent based systems, the inputs are exchanged directly with one another and can be disclosed to their users resulting in high privacy loss. We use a customized Java Sandbox architecture for an agent server in which the agents exchange their private inputs, compute the result and give it to a trusted service agent. The agents are protected from communicating with the outside world and from leaving the agent server. The service agent checks for probable private data in the computation result, creates multi-signature on the result and send it back to the users. Participating agents are disposed after the completion of the job. We also analyze probable covert channel through the result sending process. The presented architecture can be very effective in protecting privacy in multi-party cooperative computing.
一般社団法人情報処理学会の論文
2006-05-18