As the technical infrastructure to support Grid environments matures, attention must be focused on integrating such technical infrastructure with technologies to support more dynamic access to services, and ensuring that such access is appropriately monitored and secured. Current approaches for securing organisations through conventional firewalls are insufficient; access is either enabled or disabled for a given port, whereas access to Grid services may be conditional on dynamic factors. This paper reports on the Semantic Firewall (SFW) project, which investigated a policy-based security mechanism responsible for mediating interactions with protected services given a set of dynamic access policies, which define the conditions in which access may be granted to services. The aims of the project are presented, and results and contributions described.

As the technical infrastructure to support Grid environments matures, attention must be focused on integrating such technical infrastructure with technologies to support more dynamic access to services, and ensuring that such access is appropriately monitored and secured. Such capabilities will be key in providing a safe environment that allow the creation of virtual organisations at run-time. This paper addresses this issue by analysing how work from within the field of Electronic Institutions (EIs) can be employed to provide security support for Grid environments, and introduces the notion of a Semantic Firewall (SFW) responsible for mediating interactions with protected services given a set of access policies. An overarching guideline is that such integration should be pragmatic, taking into account the real-life lessons learned whilst developing, deploying and using the GRIA infrastructure for Grid environments.

One of the main challenges for the successful application of agent-based systems in mobile and embedded devices is enabling application developers to reconcile the needs of the user to the capabilities and limitations of agents in the context of environments with changing and often limited resources. In this paper we present an attempt to move towards a solution through a framework for defining and reasoning about agents in a manner that is modular and reconfigurable at run-time. Departing from the theoretical basis afforded by the SMART framework, we extend it to enable the definition of fully re-configurable component-based agent architectures. The guiding principle of this approach is an architecturally-neutral model that supports a separation between the description, behaviour and structure of an agent.

In order for agent-oriented software engineering to prove effective it must use principled notions of agents and enabling specification and reasoning, while still considering routes to practical implementation. This paper deals with the issue of individual agent specification and construction, departing from the conceptual basis provided by the smart agent framework. smart offers a descriptive specification of an agent architecture but omits consideration of issues relating to construction and control. In response, we introduce two new views to complement smart: a behavioural specification and a structural specification which, together, determine the components that make up an agent, and how they operate. In this way, we move from abstract agent system specification to practical implementation. These three aspects are combined to create an agent construction model, actsmart, which is then used to define the AgentSpeak(L) architecture in order to illustrate the application of actsmart.

Current mechanisms for evaluating the trustworthiness of an agent within an electronic marketplace depend either on using a history of interactions or on recommendations from other agents. In the first case, these requirements limit what an agent with no prior interaction history can do. In the second case, they transform the problem into one of trusting the recommending agent. However, these mechanisms do not consider the relationships between agents that arise through interactions (such as buying or selling) or through overarching organisational structures (such as hierarchical or flat), which can also aid in evaluating trustworthiness. In response, this paper outlines a method that enables agents to evaluate the trustworthiness of their counterparts, based solely on an analysis of such relationships. Specifically, relationships are identified using a generic technique in conjunction with a basic model for agent-based marketplaces. They are then interpreted through a trust model that enables the inference of trust valuations based on the different types of relationships. In this way, we provide a further component for a trust evaluation model that addresses some of the limitations of existing work.

A central requirement for achieving the vision of run-time discovery and dynamic composition of services is the provision of appropriate descriptions of the operation of a service, that is, how the service interacts with agents or other services. In this paper, we use experience gained through the development of real-life Grid applications to produce a set of requirements for such descriptions and then attempt to match those requirements against the offerings of existing work, such as OWL-S [1] and IRS-II [2]. Based on this analysis we identify which requirements are not addressed by current research and, in response, produce a model for describing the interaction protocol of a service in response. The main contributions of this model are the ability to describe the interactions of multiple parties with respect to a single service, distinguish between interactions initiated by the service itself and interactions that are initiated by clients or other cooperating services, and capture within the description service state changes relevant to interacting parties that are either a result of internal service events or interactions. The aim of the model is not to replace existing work, since it only focuses on the description of the interaction protocol of a service, but to inform the further development of such work.

Introduction to Agents. Agent Architectures. Agent Methodologies. Agent Toolkits. Agent Standards. Related Technologies.

Current mechanisms for evaluating the trustworthiness of an agent within an electronic marketplace depend either on using a history of interactions or on recommendations from other agents. In the first case, these requirements limit what an agent with no prior interaction history can do, and in the second case, they transform the problem into one of trusting the recommending agent. However, these mechanisms do not consider the relationships between agents that arise through the interactionsbetween them, such as buying or selling, or through overarching organisationalstructures, which can also aid in evaluating trustworthiness. In response, this paper outlines a method that enables agents to evaluate the trustworthiness of their counterparts, based solely on an analysis of such relationships: relationships are identified using a generic technique in conjunction with an ontological model for agent-based marketplaces; they are then interpreted through a trust model that enables the inference of trust valuations based on the different types of relationships. In this way, we provide a further component for a trust evaluation model that addresses some of the limitations of existing work.

The types of relationships that arise between interacting agents in a multi-agent system can greatly influence the effectiveness of the entire system. However, the behaviour of agents cannot always be anticipated, especially when dealing with open and complex systems. Such systems must incorporate relationship management mechanisms that attempt to guide the behaviour of agents towards desired outcomes. To effectively design such mechanisms, we must first be able to identify the types of relationships that may emerge and how they can be understood, when faced with the constraints and opportunities presented by inter-agent relationships. We have previously addressed this in the limited context of restricting malicious behaviour through the application of regulations. In this paper we generalise that model to identify and characterise a much broader range of relationships. This results in a generic analysis tool, which can be used to achieve two crucial tasks: to identify opportunities for more effective coordination between agents; and to enable the analysis of the control that agents can exert over their own goals in the context of such relationships. Combining these types of analysis allows us to identify conflicts and opportunities for cooperation in multiagent systems both at design time and at run-time, allowing for finer-grained system control.

The move towards supporting more autonomous systems, where decisions are made without direct user intervention, and more complex operating scenarios, where services from multiple organisations form temporary ties to solve particular problems, creates new security challenges. This paper argues that the answers should combine the use of conventional security solutions, such as cryptographic mechanisms, with the ability to reason about security at the semantic level, using appropriate descriptions of security policies and the required tasks. Such reasoning can enable software entities aiming to interact to determine whether their respective security requirements and capabilities will allow them to proceed. Furthermore, it can support the enforcement of security policies based on the context of the interactions. We motivate the need for such reasoning about security through an example and discuss a set of requirements to support the implementation of a specialised security device, termed a Semantic Firewall.

Cooperation is the fundamental underpinning of multi-agent systems, allowing agents to interact to achieve their goals. However, agents must manage the risk associated with interacting with others who have different objectives, or who may fail to fulfill their commitments. In this paper, we consider the role of trust and norms in a motivation-based view of agency. Motivations provide a means for representing and reasoning about overall objectives, trust offers a mechanism for modeling and reasoning about reliability, honesty, etc., and norms provide a framework within which to apply them.

This paper describes AbIMA, an agent-based intelligent mobile assistant for supporting users prior to and during the execution of their tasks. The agent is based on the well-known AgentSpeak(L) agent architecture and programming language, which provides explicit representations of agents' beliefs, desires and intentions (BDI). AbIMA is implemented using Java 2 Mobile Edition and is tested on a hand-held computer. We also provide conceptual foundations and discuss various challenges relating to the use of cognitive agent architectures for intelligent mobile user support.

The ability of agents to interact in order to collectively achieve goals is one of the central arguments for the utility of multi-agent systems. Such interactions take place whenever one agent performs an action which, intentionally or not, affects one or more other agents. Thus, when agents interact we can say that they are related by virtue of the fact that they are affecting each other. Identifying, analysing and understanding the implications of the various types of such relationships is of critical importance, since they can have both beneficial and adverse effects on the performance of the overall system and the individuals within it. Therefore, if coordination and regulation of agents is to be achieved as an agent society evolves, either by external intervention or through interventions by the agents themselves, we require some means of identifying the relationships between agents at run-time. Of course, this information is only useful if we are also able to determine how the identified relationships may impact on individual agent operation and the system as a whole. Thus, we also require a principled and comprehensive typology for characterising agent relationships. In this paper we outline just such a model.

This paper describes work-in-progress towards the development of AbIMA, an agent-based intelligent mobile assistant for supporting users prior to and during the execution of their tasks. The agent is based on the well-known AgentSpeak(L) agent architecture and programming language, which provides explicit representations of agents’ beliefs, desires and intentions (BDI). AbIMA is implemented using Java 2 Mobile Edition and is tested on a Palm Pilot mobile computing device. We discuss challenges facing us with regards to: (i) the characterisation of the opportunities BDI agents may offer for providing useful intelligent support for users of mobile devices; and (ii) the technical challenge of implementing a well established agent architecture on a mobile device with limited computational power, input and output capabilities.

Multi-agent systems result from interactions between individual agents. Through these interactions different kinds of relationships are formed, which can impact substantially on the overall system performance. However, the behaviour of agents cannot always be anticipated, especially when dealing with open and complex systems. Open agent systems must incorporate relationship management mechanisms to constrain agent actions and allow only desirable interactions. In consequence, in this paper we tackle two important issues. Firstly, in addressing management, we identify the range of different control mechanisms that are required and when they should be applied. Secondly, in addressing relationships, we present a model for identifying and characterising relationships in a manner that is application-neutral and amenable to automation.

Automated negotiation is gaining interest, but issues relating to the construction of negotiating agent architectures have not been addressed sufficiently. Towards this end, we present a novel agent construction model that enables the development of a range of agent architectures based on a common set of building blocks. In this paper we identify the fundamental components needed for two generic classes of negotiating agents: simple negotiators and argumentative negotiators, and use our model to describe them. We demonstrate how the model allows us to reuse fundamental components across these negotiation architectures.

Networked computing environments are going through radical changes. The possibility to connect practically any device through the Internet, from mobile phones to home appliances and computing devices embedded in cars to powerful personal computers, creates numerous possibilities for new kinds of applications that can lead to enhanced services for users. At the same time, application developers are faced with new challenges that need to be addressed so as to develop applications that operate in this emerging computing environment. The challenges relate to both the difficulties of creating applications that operate in dynamic, heterogeneous environments, where devices vary in capabilities and connectivity is not always guaranteed, and the need for solutions to the problem of coordinating the operation of these devices so that they can work towards common goals.

As the field of agent-based computing has continued to develop, there have been several contributions to its theoretical underpinnings, and several others to supporting the efforts of practical systems development. Yet the connection between the two has been limited at best. In this paper we aim to address these limitations through a consideration of appropriate agent infrastructure that can support principled development of agent systems based on a strong conceptual framework. As well as a general discussion of infrastructure requirements in this context, we also describe the Paradigma implementation environment, based on the smart agent framework, which represents our initial efforts in this direction.