Collaborative Research: CT-ISG: Incentive-Compatible Protocols

合作研究：激励兼容协议丽贝卡·n·赖特，史蒂文斯研究所和钟生，纽约州立大学布法罗分校互联网的迅速发展改变了我们的交流方式，我们日常生活中越来越多的方面涉及计算和数据通信。然而，许多通信协议设计假设所有参与者都将遵循指定的协议，如果不这样做，则不会保持其理想的属性。解决这个问题的一种方法是设计一个世界，在这个世界中，一些各方可能会以任意方式恶意偏离他们的协议。然而，最终的设计通常是昂贵的或不可能实现的，并且在许多情况下高估了可能遇到的不当行为的种类。较弱的安全目标更容易实现，但往往低估了参与者作弊的意愿。这个项目研究了安全的一个中间概念，称为激励兼容性。在许多实际环境中，可以合理地假设用户倾向于按照自己的最佳利益行事。因此，如果一个人能够设计和使用与激励兼容的协议，因为遵循协议对参与者产生的结果至少与偏离协议一样好，那么在许多情况下，一个人就可以避免不当行为。激励兼容性借鉴了经济学和博弈论的思想，是一种比传统安全模型更可微调的安全概念，允许不同类型的参与者受到不同的激励。与仅针对被动对手的安全协议相比，此类解决方案可以提供更好的信任和保证，而且成本可能低于针对任意恶意对手的安全协议。这使得激励兼容协议能够以“合适”的成本提供“合适”的安全级别。研究的四个领域是：激励兼容的特设网络，激励兼容的数据挖掘（两者都可以从引入激励兼容的解决方案中显著受益），激励兼容性的测试工具，以及激励兼容性的基础。

NSF 0524139 / 0524030Collaborative Research: Incentive-Compatible ProtocolsRebecca N. Wright, Stevens Institute and Sheng Zhong, State University of New York at BuffaloThe rapid expansion of the Internet has changed the way we communicate, with ever-increasing aspects of our daily life involving computation and data communication. However, many communications protocol designs assume that all participants will follow the specified protocols, and do not maintain their desirable properties if they don't. One approach to this problem is to design for a world in which some parties may maliciously deviate from their protocols in arbitrary ways. However, the resulting designs are typically expensive or impossible to achieve, and in many cases overestimate the kinds of misbehavior one is likely to encounter. Weaker security goals are more achievable, but often underestimate the willingness of participants to cheat.This project studies an intermediate notion of security, known as incentive compatibility. In many practical settings, it is reasonable to assume that users will tend to act in their own best interests. Therefore, if one can design and use protocols that are incentive-compatible, in that following the protocol produces outcomes for participants at least as good as deviating from the protocol, then one can in many settings avoid misbehavior. Incentive compatibility, which borrows ideas from economics and game theory, is a more finely tunable notion of security than traditional security models, allowing different kinds of participants motivated by different incentives. Such solutions can provide better trust and assurance than protocols secure only against passive adversaries, at a potentially lower cost than protocols secure against an arbitrarily malicious adversary. This allows incentive-compatible protocols to provide the "right" level of security, at the "right" cost.Four areas studied are: incentive-compatible ad hoc networks, incentive-compatible data mining, (both of which can significantly benefit from introducing incentive-compatible solutions), testing tools for incentive compatibility, and foundations of incentive compatibility.