Mechanisms for Secure and Robust Agent-Based Distributed Computing

一个移动的代理是一个程序，能够自主地从一个节点迁移到另一个节点，代表人类用户执行计算。 在当前和未来的以网络为中心的应用中，特别是在下一代互联网的背景下，代理模式可以发挥重要作用。 移动Agent的安全性和鲁棒性问题是目前阻碍基于Agent的应用广泛部署和使用的最大障碍，本文的主要目的是研究在开放的分布式环境（如Internet）中构建移动的Agent编程系统的安全性和鲁棒性问题。 提出的研究解决了与编程原语和系统级机制有关的问题，用于构建基于代理的 应用. 在这个建议中提出的方法是基于支持使用移动的对象的mobileagent范例。 这就提出了一系列研究挑战，涉及主机资源的保护、移动的代理的保护以及代理所携带信息的保密性和真实性。它还需要对代理进行远程监视和控制以及从故障中恢复的机制。另一个需要解决的相关挑战是建立可扩展的机制来命名和定位网络中的对象。 本研究将以Java语言为基础，利用密码学技术设计协议，解决其中的几个安全问题。 Java提供的功能（如对象序列化、反射、安全特性等）将被用于设计代理移动性机制和支持代理在远程主机上执行的保护域。一个移动的代理系统的原型实现将被建立，既作为一个研究基础设施，也作为一个演示的机制，在这方面的努力将被开发。 该系统将使用理论和实验分析进行评估。

A mobile agent is a program capable of migrating autonomously fromnode to node, to perform computation on behalf of a human user. Themobile agent paradigm can play a significant role in current and futurenetwork-centric applications, especially in the context of the NextGeneration Internet. Security and robustness concerns about mobileagents are currently identified as the biggest obstacle preventing thewidespread deployment and use of agent-based applications.The primary objective of the proposed research is to investigate securityand robustness issues in building a mobile agent programming system,in an open distributed environment, such as the Internet. The proposedresearch addresses issues related to programming primitives andsystem-level mechanisms for building agent-based applications. Theapproach presented in this proposal is based on supporting the mobileagent paradigm using mobile objects. This raises a number ofresearch challenges related to protection of host resources, protectionof mobile agents, and secrecy and authenticity of information carriedby agents. It also requires mechanisms for remote monitoring andcontrol of agents, and recovery from failures. Another related challengeto be addressed by this effort is in building scalable mechanisms fornaming and locating objects in the network. The proposed research will bebased on the Java language and will use cryptographic techniques to designprotocols to solve several of the security problems. The facilitiesprovided by Java (such as object serialization, reflection, securityfeatures, etc.) will be used in devising mechanisms for agent mobility andsupporting protection domains for agent execution on remote hosts.A prototype implementation of a mobile agent system will be built toserve both as a research infrastructure as well as a demonstrationof the mechanisms to be developed during this effort. The systemwill be evaluated using both theoretical and experimental analyses.