de Groot, D. R. A. and Brazier, F. M. T., Identity Management in Agent Systems, in: Proceedings of the 1st International Workshop on Privacy and Security in Agent-based Collaborative Environments (PSACE) at the Fif th International Joint Conference on Autonomous Agents and Multi-Agent Systems (AAMAS-06), pages 23-34, Future University, Hakodate, Japan, 2006.
Abstract: The Agent Operating System (AOS) provides the basic functionality needed for secure and reliable mobile agentplatforms: support for secure communication, secure agent storage and migration, and minimal primitives for agent life-cycle management. Designed as a layer between local operating systems and higher level agent platform middleware, it supports interoperability between agentplatforms and between different implementations of AOS itself. AOS has been tested on interoperability, both with regard to different higher-layer middleware platforms and interoperability between two implementations of AOS in C++ and Java.
Abstract: In most of today s agent systems migration of agents requires homogeneity in the programming language and/or agent platform in which an agent has been designed. In this paper an approach is presented with which heterogeneity is possible: agents can migrate between non-identical platforms, and need not be written in the same language. Instead of migrating the code (including data and state) of an agent, a blueprint of an agent s functionality is transferred. An agent factory generates new code on the basis of this blueprint. This approach of generative mobility not only has implications for interoperability but also for security, as discussed in this paper.
Abstract: Anonymity can be of great importance in distributed agent applications such as e-commerce & auctions. This paper proposes and analyzes a new approach for anonymous communication of agents based on the use of handles as pseudonyms. A novel naming scheme is presented that can be used by agentplatforms to provide automatic anonymity of communication for \emph{all} agents on its platform, or, alternatively, to provide anonymity \emph{on demand}. The paper furthermore introduces new approaches that provide authentication and anonymous payment schemes for agents. Performance measures for an anonymity service implemented for the AgentScape platform provides some insight in the overhead involved.
Abstract: In current agent systems agent migration is only possible between homogeneous systems supporting identical agentplatforms, limiting an agent?s possibilities considerably. This paper revisits the notion of generative migration. Generative migration entails migration of an agent blueprint, instead of complete code. This approach relies on homogeneity of libraries on different platforms to re-incarnate agents, but does not require homogeneity of platforms. Agent Factories are used to assemble agents at their destination, adapting an agent to its environment. This paper continues earlier work on generative migration by extending available theory and providing a demonstration and implementation of generative cross-platform agent migration using Agent Factories.
Abstract: Agentplatforms designed for Internet-scale, open networks need scalable and secure location services for agents and services. The location service based on the Fonkey public key distribution infrastructure presented in this paper has been designed and implemented for this purpose. It is scalable in the total number of published identifier--contact address pairs, the number of updates/changes, and the number of agentplatforms publishing and requesting contact addresses. This system also supports a signing mechanism to authenticate the publisher of an identifier--contact address pair. Experimental results show that the current implementation based on the Bunshin/Free Pastry overlay network exhibits good scaling behavior.
Abstract: Agents, and in particular mobile agents, offer a means for application developers to build distributed applications. Given homogeneity of agent platform and code base, agent migration is possible. However, many agentplatforms exist, differing substantially in the support for agents. Write once - run everywhere is not yet true for agents... Heterogeneity of agentplatforms, combined with heterogeneity in code-bases of agents, leads to an interesting question concerning agent mobility: can an agent migrate in a heterogeneous environment? The answer is relatively simple: an agent needs to be adapted to its destination agent platform and code-base, e.g. by an agent factory.
de Groot, D. R. A. and Brazier, F. M. T., Identity Management in Agent Systems, in: Proceedings of the 1st International Workshop on Privacy and Security in Agent-based Collaborative Environments (PSACE) at the Fif th International Joint Conference on Autonomous Agents and Multi-Agent Systems (AAMAS-06), pages 23-34, Future University, Hakodate, Japan, 2006.
Abstract: If agent-based applications are to be used in large scale, open environments, security is a main issue; digital identity management (DIDM) an essential element. DIDM is needed to be able to determine the rights and obligations of the four main computational entities in such systems: agentplatforms, hosts, agents, and services. The framework for evaluation of DIDM in agent systems proposed in this paper is based on four aspects of DIDM: representation, confidentiality, integrity and availability. Two agentplatforms (JADE-S and AgentScape) are used to illustrate the potential of this framework.
Abstract: If agent-based applications are to be used in large scale, open environments, security is a main issue; digital identity management (DIDM) an essential element. DIDM is needed to be able to determine the rights and obligations of the four main computational entities in such systems: agentplatforms, hosts, agents, and services. The framework for evaluation of DIDM in agent systems proposed in this paper is based on four aspects of DIDM: representation, confidentiality, integrity and availability. Two agentplatforms (JADE-S and AgentScape) are used to illustrate the potential of this framework.
Abstract: In general, Computer Science agentplatforms maintain a simple agent model. In contrast,
Artificial Intelligence agent technology maintains a strict agent model, requiring agents to
have mental notions such as beliefs, desires and intentions (BDI). This strict model offers
advantages for application developers in certain domains, as it represents a higher level of
abstraction than traditional programming languages, more akin to human reasoning.
To offer this strict BDI agent model to Computer Science agentplatforms, this thesis
describes the integration of the high-level BDI agent programming language Jason into the
AgentScape agent platform, which maintains a simple agent model. The selection of relevant
BDI languages is discussed, as well as the approach to integrate AgentScape and Jason.
The resulting integration succeeded: Jason’s BDI reasoning engine, internal actions, agent
creation and part of Jason’s communication features are all available in the integrated system.
Abstract: Software agents extend the current, information-based Internet to include autonomous mobile processing. In most countries such processes, i.e. software agents are, however, without an explicit legal status. Many of the legal implications of their actions (e.g. gathering information, negotiating terms, performing transactions) are not well understood. One important characteristic of mobile software agents is that they roam the Internet: they often run on agentplatforms of others. There often is no pre-existing relation between the ?owner? of a running agent?s process and the owner of the agent platform on which an agent process runs. When conflicts arise, the position of the agent platform owner is not clear: is he or she allowed to slow down the process or possibly remove it from the system? Can the interests of the user of the agent be protected? This article explores legal and technical perspectives in protecting the integrity and availability of software agents and agentplatforms.
Abstract: Anonymity is of great importance in distributed agent applications such as e-commerce & auctions.
This paper proposes and analyzes a new approach for organized anonymity of agents based on the use
of pseudonyms. A novel naming scheme is presented that can be used by agentplatforms to provide
anonymity for each individual agent. The paper introduces two distinct techniques, one based on handles
and another based on agent spawning. Both techniques can be integrated into agent platform middleware,
automatically guaranteeing anonymity for all individual agents. The applicability of this approach is
evaluated for three agentplatforms: AgentScape, JADE and SeMoA.
Abstract: Anonymity is of great importance in distributed agent applications such as e-commerce & auctions. This paper proposes and analyzes a new approach for organized anonymity of agents based on the use of pseudonyms. A novel naming scheme is presented that can be used by
agentplatforms to provide automatic anonymity for all agents on its platform, or, alternatively, to provide anonymity on demand. The paper introduces a new techniques, based on the use of handles. It can be integrated into the agent platform middleware, thereby ensuring organized anonymity for all agents. Performance measures for an anonymity middleware service implemented for AgentScape provides insight in the overhead involved.
