Privacy preserving learning in negotiation

Machine learning techniques are widely used in negotiation systems. To get more accurate and satisfactory learning results, negotiation parties have the desire to employ learning techniques on the union of their past negotiation records. However, negotiation records are usually confidential and private, and owners may not want to reveal the details of these records. In this paper, we introduce a privacy preserving negotiation learning scheme that incorporate secure multiparty computation techniques into negotiation learning algorithms to allow negotiation parties to securely complete the learning process on a union of distributed data sets. As an example, a detailed solution for secure negotiation Qlearning is presented based on two secure multiparty computations: weighted mean and maximum. We also introduce a novel protocol for the secure maximum operation. Categories and Subject Descriptors H.3 [Information Storage and Retrieval]: Information Search and Retrieval General Terms Algo...

• Type: Conference
• Year: 2005
• Where: SAC
• Authors: Sheng Zhang, Fillia Makedon
• Keywords: Applied Computing, Negotiation, Negotiation Parties, Negotiation Records, SAC 2005