Approved for public release; distribution is unlimited / This research concerns information security and computer-network defense. It addresses how to handle the information of log files and intrusion-detection systems to recognize when a system is under attack. But the goal is not the usual one of denying access to the attacker but providing a justification for deceptive actions to fool the attacker. We implemented a simple demonstration of how two different kinds of open-source intrusion-detection systems can efficiently pool data for this purpose. / Lieutenant Commander, Brazilian Navy

The aim of this dissertation is to develop a suitable bidding strategy for an internet bidding agent that allows the agent to obtain the lowest possible price for a desired product at an internet auction. The bidding strategy is obtained under the constraint of limited information available about the strategies of the opponents. The agent will operate in an internet auction environment. Therefore classic auction theory is researched and explained. Auctions are widely used to bring buyers and sellers of products together and to create a market to buy and sell goods. The buyer wants to pay the lowest possible price and the seller wants to receive the highest possible price. However, the seller has no influence on the final selling price of the product. Instead the price is determined by the buyers. The agent will place bids on the auction site on behalf of the human instructor. The bidding agent will make use of the theory behind auctions to influence the other bidders on the auction to make the lowest possible bids. The model suggested in this dissertation, the regressive bidding agent model (RBA model), will incorporate auction theory to create a suitable agent. The agent will predict the future bids of opponents on the auction, basing its predictions on a regressive function. The agent will base its own bids placed at the auction on the bidding time remaining at the auction together with the bids placed by other bidders on the auction. / Prof. E.M. Ehlers

M.Sc.(Computer Science) / The dissertation investigates trust and reputation as a specialisation of agent technology. The research presented herein aims to establish and demonstrate how it is possible for one rational agent to trust another entity. Furthermore, the research presented herein aims to determine the extent of the limitations of trust and reputation models, and of the demonstrable solution in particular. To this end, the dissertation investigates theoretical aspects of trust. The dissertation investigates several existing trust models and establishes criteria for a qualitative analysis. Supplementary techniques aimed at enhancing trust evaluation are also investigated. The research also identifies architectural abstractions suitable for developing agents capable of intelligent trust evaluation. The main focus of the research is enhancing agent protection through a trust-based approach. A particular problem is the threats posed to mobile agents from malicious agent hosts. Therefore, a solution is sought that can be used to augment existing mechanisms aimed at mobile agent protection and agent protection in general. Thus, the research also examines mobile agents and mobile agent systems in an effort to produce a general trust-based solution that can be applied in most mobile agent systems. The solution presented in the dissertation proposes the concept of an evaluator agent as an add-on to existing mobile agent systems. The evaluator agent is presented as a rational agent with an embedded intelligent trust evaluation capability. The intelligent trust evaluation capability is provided via a set of reusable components. The solution demonstrates how a rational agent may evaluate the trustworthiness of other entities. The dissertation further analyses the strengths and limitations of the approach. The dissertation provides results that quantitatively demonstrate the extent of the limitations of the trust-based approach. The contribution of the dissertation partly lies in the service orientation of the evaluator agent approach. The service orientation of the solution provides an abstraction and a degree of heterogeneity suitable for handling the challenges of open environments. The solution can be deployed in most mobile agent systems to provide a trust evaluation service without the need to redesign existing mobile agent systems. More broadly, the research is another step towards the development of cognitive social agents.

The growing popularity of online virtual communities such as Second Life and ActiveWorlds demands the presence of intelligent agents to assist users in their daily online activities (e.g., exploring, shopping, and socializing). As these virtual environments become more crowded, multiple agents are needed to support the increasing number of users. Multi-agent environments, however, can suffer from the problem of resource competition among agents. It is therefore necessary that agents within multi-agent environments include a coordination mechanism to prevent unrealistic behaviors. Moreover, it is essential that these agents exhibit some form of intelligence, or the ability to learn, to support realism as well as to eliminate the need for developers to write separate scripts for each task the agents are required to perform. This research presents a coordinated reinforcement learning framework which can be used to develop task-oriented intelligent agents in multi-agent virtual environments. The framework contains a combination of a "next available agent" coordination model and a reinforcement learning model consisting of existing temporal difference reinforcement learning algorithms. Furthermore, the framework supports evaluations of reinforcement learning algorithms to determine which methods are best suited for task-oriented intelligent agents in dynamic, multi-agent virtual environments. To assess the effectiveness of the temporal difference reinforcement algorithms used in this study (Q-learning and Sarsa), experiments were conducted that measured an agent's ability to learn three tasks commonly performed by workers in a café environment. These tasks were basic sandwich making (BSM), complex sandwich making (CSM), and dynamic sandwich making (DSM). The BSM task consisted of four steps. The CSM and DSM tasks contained an additional fifth step. The agent learned the BSM and CSM tasks from scratch while the DSM task was learned after the agent became skillful in BSM. The measurements used to evaluate the efficiency of the Q-learning and Sarsa algorithms were the percentage of successful and optimally successful episodes performed by the agent and the average number of time steps taken by the agent to complete a successful episode. The experiments were run using both a fixed (FEP) and variable (VEP) ε-greedy probability rate. Results showed that the Sarsa reinforcement learning algorithm, on average, outperformed the Q-learning algorithm in almost all experiments except when measuring the percentage of successfully completed episodes using FEP for CSM and DSM, in which Sarsa performed almost equally as well as Q-learning. Overall, experiments utilizing VEP resulted in higher percentages of successes and optimal successes, and showed convergence to the optimal policy when measuring the average number of time steps per successful episode.

Agent-oriented software engineering (AOSE) covers issues on developing systems with software agents. There are many techniques, mostly agent-oriented and object-oriented, ready to be chosen as building blocks to create agent-based systems. There have been several AOSE methodologies proposed intending to show engineers guidelines on how these elements are constituted in having agents achieve the overall system goals. Although these solutions are promising, most of them are designed in ad-hoc manner without truly obeying software developing life-cycle fully, as well as lacking of examinations on agent-oriented features. To address these issues, we investigated state-of-the-art techniques and AOSE methodologies. By examining them in different respects, we commented on the strength and weakness of them. Toward a formal study, a comparison framework has been set up regarding four aspects, including concepts and properties, notations and modeling techniques, process, and pragmatics. Under these criteria, we conducted the comparison in both overview and detailed level. The comparison helped us with empirical and analytical study, to inspect the issues on how an ideal agent-based system will be formed.

No description available.

Negotiation is one of the most important activities for organisations in conducting electronic business. Traditional purchasing and selling have been conducted through complex processes involving negotiation that includes coordination and cooperation. To conduct automated negotiation for electronic business, a multi-agent system is needed where agents interact with each other. To perform this activity effectively and efficiently agents need to be able to negotiate, coordinate and cooperate with each other within the system. The research detailed in this thesis investigated the negotiation process in business-to business (B2B) transactions in supply chain management for multi-agent based electronic business (e business). Specifically, it answers the following research question: How can the negotiation process in B2B transactions be formulated and applied in multi-agent based e-business? The research strategy utilized an exploratory case study framework, with methods from decision theory, game theory, fuzzy logic and simulation for analysis. A series of integrated studies were undertaken to develop: an automated negotiation protocol; negotiation strategies; and a coordination and cooperation model. These were analysed in the context of the case study, Trading Agent Competition Supply Chain Management (TAC SCM) game scenario. Currently, the TAC SCM is the only international competition involving an electronic marketplace (e market). The studies involved the negotiation strategy between two agents, where the agents will be able to solve a problem by finding the best feasible strategy to bind an agreement for negotiation. By adopting a maximin and minimax strategy, this research proposes that agents will reach a reasonable positive intention approach towards negotiation, and will increase the agreement binding rate. A negotiation strategy was also examined by Fuzzy Logic using possibility theory and linguistic variables in which it also proposes a negotiation strategy in an uncertain situation for the TAC SCM. This will aid in binding the agreement to achieve the agent’s expected profit. Next, this research reviewed the TAC SCM game and explored the procurement performance of agents.The monotonic concession negotiation protocol, which determines the rules in which the agents can offer and counter-offer in the negotiation process was investigated. The author proposes two types of protocols. The first protocol is a Non-Monotonic Protocol with theoretical analysis. The second protocol is an Extended and Flexible Iterated Negotiation Process. This research also developed an Extended Bilateral Negotiation Model based on OMG (1999). The Negotiation Mechanism involving Offering and Counter-Offering models were also developed. Next, this research reviewed the cooperation and coordination process. This study identified problems in conducting e-business and supply chain management and expected benefits for supply chains with agents working together in coordinated and cooperative processes. The utilization of the multi-agent system in supply chain management with the Enhanced and Effective Cooperative Processing Stages is discussed. To apply these stages, the author proposes an architecture of Effective Cooperative Processing for Agents, and some characteristics in modelling coordination and cooperation for TAC SCM have been outlined. The research detailed how the negotiation process in B2B transactions can be formulated and applied in multi-agent based e-business. Through the proposal of a Flexible and Iterated Negotiation Framework, consisting of an Extended Bilateral Negotiation Model and a Cooperation and Coordination Model, the research community moves further towards the ultimate goal of an efficient, economic and automated negotiation process. In summary, the main contributions of the thesis include: a theoretical analysis of the negotiation process with coordination and cooperation; proposed models for an automated negotiation process; development of strategies and protocols for automated negotiation; and the coordination and cooperation model that can be used not only in supply chain management but also in any type of e-business. / Doctor of Philosophy

The field of agent control is concerned with the design and implementation of components that form an agent's control architecture. The interaction between these components determines how an agent?s sensor data and internal state combine to direct how the agent will act. Rule-based systems couple sensing and action in the form of condition-action rules and one class of such systems, learning classifier systems, has been extensively used in the design of adaptive agents. An adaptive agent explores an often unknown environment and uses its experience in its environment with the aim of improving its performance over time. The data an adaptive agent receives regarding the current state of its environment might be limited and ambiguous. In learning classifier systems, three different approaches to the problem of limited and ambiguous data from the environment have been: (1) to enable the agent to learn from its past experience, (2) to develop sequences of rules (in which rules may be linked implicitly or explicitly) and (3) multiagent LCSs. This thesis investigates the use of an adaptive agent?s past experience as a resource with which to perform a number of functions internal to the agent. These functions involve developing explicit sequences of rules, detecting and escaping from infinite loops, and firing and reinforcing rules. The first part of this thesis documents the design, implementation and evaluation of a control system that incorporates these functions. The control system is realised as a learning classifier system and is evaluated through experiments in a number of environments that provide limited and ambiguous stimuli. These experiments test the impact of explicit sequences of rules on the performance of a learning classifier system more thoroughly than previous research achieved. The use of explicit sequences of rules results in mixed performance in these environments and it is shown that while the use of these sequences in simple environments enables the rule space to be more effectively explored, in complex environments the behaviours developed by these sequences result in the agent stagnating more often in corners of the environment. Rather than endowing the rule-base with more rules, as in previous research with explicit sequences of rules, it is proposed that multiple interacting agents may enhance the exploration of the rule space in more complex environments. This approach is taken in the second part of this thesis, where the control system is used with multiple agents that interact by sharing rules. The aim of this interaction is to enhance the rule discovery process through cooperation between agents and thus improve the performance of the agents in their respective environments. It is shown that the benefit obtained from rule sharing is dependent on the environment and the type and amount of rule sharing used and that rule sharing is generally more beneficial in complex environments compared to simple environments. The properties of the rule-bases developed in these environments are examined in order to account for these results and it is shown that the type and amount of rule sharing most useful in each environment are dependent on these properties.

Intelligent agents are goal-oriented software entities which exhibit a number of desirable characteristics, such as flexibility and robustness, which are suitable for complex, dynamic, and failure-prone environments. However, these characteristics of individual agents are not exhibited by their interactions with each other since traditional approaches to interaction design are message-centric, and these message-centric approaches force the intelligent agents to follow prescribed message sequences in order to achieve their interactions, thus usually resulting in interactions which have limited flexibility and robustness. In this thesis an alternative to the traditional message-centric interaction design approaches is presented. In this approach, the interactions are designed based on interaction goals, and message sequences are not prescribed. Instead, message sequences emerge from the interactions as the intelligent agents attempt to achieve the interaction goals. The main contribution of this work is Hermes, a methodology for the design and implementation of goal-oriented interactions. An important motivation for Hermes is to not only allow for the design and implementation of goal-oriented interactions, but to also be pragmatic and usable by practicing software engineers. To that end, Hermes has a clear and guided design process with a notation explicitly created for the design of goal-oriented interactions. Furthermore, Hermes, which covers the design and implementation of agent interactions only, has been integrated with Prometheus, a full agent system design methodology. Guidelines for the integration are provided so that, in future, Hermes may also be integrated with other existing methodologies if desired. Hermes also provides guidelines for mapping its design artifacts to an implementation. As Hermes is goal-oriented, the implementation platform should be one that is goal-based. The guidelines help developers map the design to skeleton code. This contributes to the pragmatism of Hermes. To further ensure that Hermes is pragmatic, two prototype software support tools have been developed. The design support tool allows for the graphical design of Hermes artifacts and the implementation support tool produces skeleton code for the Jadex agent platform based on a structured textual representation of Hermes design artifacts. Although only the Jadex agent platform is currently supported, the implementation tool can be extended to accommodate other goal-based agent platforms. An empirical evaluation was carried out, and its results show that Hermes designs are significantly more flexible and robust than message-centric designs, although more time is required to design Hermes interactions. This suggests that Hermes is suitable for interactions which are complex and/or error-prone.

Manufacturing intelligent agent simulation has not been widely applied in industry because of its application complexity. This complexity, which includes choosing priority machines or jobs, determining machine maintenance schedules, and allocating working shifts and breaks, requires intelligent decision making. Manufacturing systems are strongly influenced by intelligent decision makers. Especially for a fixed manufacturing layout, system performance improvement depends on intelligent manufacturing decision making. As a result, a manufacturing simulation can not be truly complete if intelligent decision making processes are not represented. This thesis describes an architecture which includes the representation of intelligent agents in manufacturing simulation model. An intelligent agent simulation environment (IASE) is developed under the concepts of distributed artificial intelligence and object oriented methodology. As an extension to an existing simulation environment, IASE inherits primary manufacturing simulation elements and material handling systems from object oriented manufacturing architecture (Beaumariage, 1990) and AGV simulation system (Beaumariage and Wang, 1995). In IASE, production operators, maintenance technicians and job releasers are created to represent manufacturing intelligent agents. Several basic elements such as the blackboard structure and knowledge base for supporting intelligent agent simulation are also developed. In contrast to traditional simulation environments designed for and in procedural programming languages, future extensions or modifications for IASE are eased since IASE is developed in an object oriented fashion. This paper introduces IASE structure both in the conceptual design and implementation methodology levels. At the end, two case studies are performed. The first case study is to verify IASE's implementation and results by comparing it with a model developed in SLAM II. The second case study, a mixed intelligent agent decision making example, demonstrates the intelligent agent simulation ability of IASE. / Graduation date: 1997