Secure Agent Mediated E-Commerce by Richard Ssekibuule Registration Number: 2007/HD18/4826U BSc (Computer Science, Mathematics) - MAK, MSc (Computer Science) - RUN, Oracle Certified Associate - Oracle Department of Computer Science Faculty of Computing and IT, Makerere University [email protected], +256-71-4284860 A Research Proposal Submitted to School of Graduate Studies for the Study Leading to a Thesis in Fulfilment of the Requirements for the Award of the Degree of Doctor of Philosophy in Computer Science of Makerere University Proposed Supervisors Dr. Jose Ghislain Quenum Department of Computer Science Faculty of Computing and IT, Makerere University, [email protected] Dr. Erik Poll Department of Computer Science Radboud University Nijmegen [email protected] April, 2008

List of Acronyms AI

Artificial Intelligence

AMEC

Agent Mediated E-Commerce

B2B

Business-to-Business

B2C

Business-to-Consumer

B2G

Business-2-Government

C2B

Consumer-to-Business

CBB

Consumer Behavior Buying model

CSC

Consumer-to-Consumer

ebXML

Electronic Business eXtensible Markup Language

E-Commerce

Electronic Commerce

IP

Internet Protocol

OWL

Web Ontology Language

RDF

Resource Description Framework

XACML

eXtensible Access Control Markup Language

XML

eXtensible Markup Language

1

Introduction

The ever evolving world of information and communication technologies has led to development of systems that have changed or affected ways in which organizations and individuals go about their businesses. As a result of advancements in web technologies and services, the process of carrying out business transactions on the Internet, commonly known as electronic commerce (abbreviated as e-commerce)[1] has become an important area of research and professional practice.

1.1

Background to the study

Electronic commerce1 transactions can be distinguished depending on the nature of transactions carried out between participating parties. Coppel et al. [1] and Eastin et al. [2] separately identified some of the transactions presented below that we use to provide various classifications for e-commerce transactions. • In circumstances where transactions are carried out directly between consumers e.g. Ebay[3], such a transaction is classified in the Consumer-to-Consumer (C2C) domain, such transactions are common in auction markets. • Additionally consumers are involved in selling information (e.g. pricing information) to businesses, then the transaction is classified as Consumer-to-Business(C2B). • The reverse of C2B is more common with retailing transactions to individual customers. This nature of transaction is referred to as Business-to-Customer (B2C). • Business-to-Business (B2B) domain on the other hand refers to transactions that occur between corporations. In that case both the seller and buyer are corporations. • Coppel et al. [1] suggest that governments also have large number of transactions that they carry out with other governments, corporations or individuals. Australia, Taiwan, Mexico, the United Kingdom and the United States governments are among those that have reorganized part of their management of public procurement processes over the Internet[1, 4, 5, 6]. Transactions involving government and corporations can be classified as either Government-to-Business (G2B) or Business2-Government (B2G) depending on the nature of transaction taking place. Technology embracing governments are using the Internet to process payment information and receipts from taxes and anyother services offered (G2B, G2C, C2G, B2G)[1, 5]. Figure 1 below, provides a summary of e-commerce domains with example areas of application. The current establishment of e-commerce systems provides clients (who are mostly human) with a well structured catalog of items on sale. The structure further provides a search or navigation facility using links through which customers reach their desirable items. This kind of service only scratches the surface for what is possible with e-commerce. More automation is needed on both the sellers’ and buyers’ side to reduce time needed 1

See subsection 1.2.1 for definition

1

Figure 1: E-commerce and Internet applications areas (adopted from [1]) by a client to get what they want and also to help the sellers reach target customers faster with customized advertisements, goods, prices and services. To achieve a high level of automation, a new paradigm of software is needed to handle complex tasks of varying offers (in terms of price, quality and warranty among others). In this business paradigm, it is envisioned that businesses will be able to monitor prices of competitors in order to make sure that their own prices remain competitive and performing inventory management in order to sell products at the right price and right time[7]. This kind of model is based on agent systems[8] defined in subsection 1.2.2. Agents being dynamic in nature, autonomous and proactive, are useful in ways that represent a computing paradigm natural to humans. An agent can act as a broker for its owner with minimal expenses as compared to the traditional human broker in market transactions. Information retrieval activities can be enhanced to the benefit of users if a profile concerning the searcher is known before hand. Agents provide this facility in a natural way since the agent will always serve needs of its owner. It is also important to note that agents do not act on their own, they have to interact with other systems and information repositories on the Internet. In order for agents to execute tasks that are assigned to them, they interact with other agents in the environment whose intentions could be malicious. This largely introduces trust questions among agents and platforms on which they execute. To explain the trust problem in the agent paradigm further, we consider an example of a person who wishes to purchase items from one of the shops in his/her new neighborhood. This kind of buyer would wish to choose a reliable or trustable shop among those available that meets the financial (issues concerning pricing, discounts, promotions etc.) and other set constraints such as quality, shopping ambious, customer care among others. It would be wise for such a consumer to consult with people who have stayed in that place long enough in order to obtain information about the shops or markets. Additionally the customer should have some level of trust in his/her neighbors to accept the information whole heartedly. In such a situation, the new consumer is in a trust predicament that is similar to what agents face whenever they seek reputation information from new sources that could have been compromised for malicious intent. 2

Due to the increasingly sophisticated nature of attackers, reputation information is sometimes altered to benefit individual agents and at times wrong information about agents is sent to reputation repositories. Attackers that carry out such actions are commonly referred to as strategic liars. Additionally, most trust models have been generic in nature, and thus giving rise to confusing trust contexts. Confusing trust contexts arise in cases where one entity assumes certain parameters for trust, yet those parameters are insignificant or irrelevant for the prevailing environment. Examples of strategic liars and confusing trust contexts have previously been presented in literature[9, 10]. Such challenges require improvement in trust management models to close existing gaps in agent-mediated e-commerce environments. Additionally, existing security techniques need to be improved to cater for the unique processes of agent-mediated e-commerce.

1.2

Definition of Key Terms

To put this proposal into context, this section defines key words used. 1.2.1

E-Commerce

E-commerce was defined by Till et al.[11] from the Electronic Commerce Association as “any form of business or administrative transaction that is executed using any information and communications technology”. However, this definition is very broad for our research work, we limit our interests in commercial activities carried out on the Internet. In that respect, we consider e-commerce activities that involve buying, selling, advertising and eventual delivery of goods and services. The presence of e-commerce in organizations comes with benefits, most especially related to automation of business operations. Given the possibilities of automation and easy communication using the Internet, e-commerce has undeniably reduced the time required by a customer to find needed products and services. E-commerce also plays a pivotal role in the business of suppliers, since they can easily market their goods and services using customized messages for target audiences to increase their sales. 1.2.2

Agent Systems

An agent is commonly defined as “a software program that can carry out a task autonomous on behalf of its owner”[8]. In order for agent systems to facilitate e-commerce services, they are are expected to exhibit generic characteristics discussed in the bulleting below. These characteristics were previously presented by Wooldridge et al.[12]. • Autonomy: Agents are are expected to have the ability of acting independently without constantly referring back to their owner or user. In the case of e-commerce services, an agent is supposed to have the ability to make choices concerning goods or services that an owner might need depending on the owner’s desires, budget and time of delivery using the initial information that was given. • Social ability: Agents should be able to communicate with other agents and possibly humans via some agent communication language. 3

• Reactivity: Agents should be able to respond appropriately to the prevailing circumstances in their environment and also respond in a timely fashion to changes that occur. • Proactivity: Agents are expected to have the ability to act in anticipation of future goals so that its owner’s objectives are met. Researchers in Artificial Intelligence (AI) impose stronger notions to agents in which they are expected to exhibit human like characteristics such as knowledge, belief, intention, and obligation [13, 12]. Some of these characteristics are essential to agent operations, yet they can be optionally used in some environments. 1.2.3

Semantic-Agents

In order to integrate agents in advanced e-commerce environment, there was a definite need of integrating intelligence in the agent paradigm. The concept of intelligence in relation to agent systems was indicated by Sycara et al. [13]. In order for agents to have intelligence that is needed for advanced e-commerce, scientist have employed the semantic web technologies[14, 15] in agent systems and hence the term semantic-agents. The term semantic-agents refers to agent systems that use semantic web technologies and frameworks in service provision. Essentially, the Semantic Web is about two things[16]: 1. It is used to define formats for integration and combination of data from various sources. Unlike the original web that mainly concentrated on the interchange of documents, the semantic web extends the current web by providing more meaning to data on the web. 2. It is also about the language for recording how the data relates to real world objects and facilitates interconnecting repositories that contain related information. 1.2.4

The Web Ontology Language

The Web Ontology Language (OWL)[17, 18] is a semantic markup language that was developed for publishing and sharing ontologies on the World Wide Web. The Oxford English Dictionary defines ontology as the science or study of being. However this definition is too broad and abstract to make meaning to this research. We consider a more narrowed down definition of ontology in the context of the semantic web. Ontology was defined by Hendler [19] as “a set of knowledge terms, including the vocabulary, the semantic interconnections, and some simple rules of inference and logic for some particular topic”. Ontologies are used in constructing relationships between people and objects that are readable by computer systems.

1.3

Security issues in AMEC

Application of agents in large-scale open distributed environment systems presents a host of new challenges. In the enumeration below, we present some of the challenges that are likely to be faced by agent systems in an open distributed environment. 4

1. Availability - Agents in an open distributed environment represent stake holders that have several objectives some of which could be malicious or depriving to other participants[20, 21, 10]. 2. Non-repudiation and Identity Management - Agents operating in open distributed environments have the ability to perform various tasks and later on obtaining a different identity to hide their previous actions[10]. In an e-commerce environment, agents could accept payments for goods that they may not deliver after receiving payments. Additionally, selling agents can even deny having ever received payment for goods to be delivered to customers. 3. Integrity - Information that is carried by agents in the open distributed environment needs to be protected from malicious attacks that may want to destroy it or even alter its meaning[22]. 4. Privacy - As agents traverse networks to fulfil their objectives, they carry information with them that usually represents needs of their owners. This information is used as a basis for interaction and service negotiation with other agent systems in the network. In many cases (and for several different reasons) agent owners would wish to keep all or part of this information private[23, 24].

1.4

Statement of the Problem

The deployment of agent systems in open distributed environments comes with a host of challenges; security being among the most notable ones. Security challenges for agent systems have been classified into two categories, based on the entity under attack. • Firstly, agents in e-commerce systems face threats on three fronts, namely; Agent platforms, Agents of the same or different type of application and other entities in the environment. “Other entities” in the environment refers to attackers that are not agents or mobile in nature. • Secondly, agent platforms face threats concerned with executing code from unknown or malicious sources. Problems associated with execution of code from unknown or untrusted sources have been well documented in literature[25]. The broad concept of security in this research is used in reference to issues concerning confidentiality, integrity and availability for information carried by agents, agents and the platforms on which they are executed. Being the nature of agent systems that they carry information and processes across networks, they provide an attractive target to people or agents with malicious intent. Such attackers may wish to either destroy them or even alter information that they carry (i.e. threat to service and information availability, information and agent code integrity and confidentiality). It is worth noting that strong trust management mechanisms need to be put in place to ensure that intruders do not abuse trust policies among collaborating agents. Research work has been done in this area, with several mechanisms and protocols presented. However, some problems have not been addressed yet. These include, strategic liars and 5

confusing trust contexts due to informal expectations in reputational models. As previously indicated by Matthews et al.[26], “the trust component has not progressed far beyond a vision of allowing people to ask questions concerning trustworthiness of the information on the web”.

1.5 1.5.1

Objectives General objective

The general objective of this research is to develop a security framework and propose improvements to on techniques relevant for agent mediated e-commerce security in open distributed environments. 1.5.2

Specific objectives

In order to achieve the main research objective, the specific objectives are: i. to investigate security requirements for agent mediated e-commerce case study. ii. to design a security framework for agent mediated e-commerce case study. iii. to implement the security framework in an agent mediated e-commerce case study. iv. to test and validate the proposed security framework.

1.6

Significance of the study

The theoretical and practical framework of this research is concerned with security challenges to agent mediated e-commerce. The current hindrances to wide deployment of agent platforms such as AgentScape [27, 28], JADE [29] and SEMOA [30] into large scale e-commerce systems is partly because of a weak integration of ontologies into these platforms. Consequently, most of the current interaction models between agents in these platforms are reasonably simplistic. This research work will contribute to the development of efficient models for better predictive capabilities and improved forms of secure interaction during brokering, supply chain management, negotiations and auctioning processes in open distributed environments.

1.7

Scope

The two categories of security problems in e-commerce, explained in problem statement subsection 1.4 raise two questions: • Firstly, how can we effectively protect agents and their data from attackers? Generally agents could face threats from platforms on which they execute and other entities in the environment such as non agent based software and network related attacks. 6

• A second research question is whether we can protect agent platforms against malicious agents? The research work will focus on the first question. There is hope of finding an answer for the second question in the successful research work being done on proof carrying code(PCC)[31]. The research will involve developing a security framework for agents in e-commerce applications for Business-to-Consumer (B2C) transactions.

2

Literature Review

In this section, the state of literature that is directly connected to the research project is presented. In subsection 2.1 the state of agent-mediated e-commerce and role of agent systems in e-commerce environments is presented. Subsections 2.1.1 and 2.1.2 present more details on the role of agents in both B2B and B2C environments. The state of the art in semantic agent mediated e-commerce is presented in subsection 2.2 while literature on security techniques for agents in open distributed environments are covered in subsection 2.4. The literature review is concluded with characteristics of a tool for semantic-agent mediated e-commerce in subsection 2.3.

2.1

Agent-Mediated E-Commerce (AMEC)

This subsection presents a few examples of services that can be facilitated by agents in second generation e-commerce. It is estimated by IDC[32] that the demand for application of agents in e-commerce will grow as B2B transactions become common place. As was indicated by He et al. [7], “second generation e-commerce is characterized with a higher degree of automation than first generation e-commerce that is commonly available”. However, there are overlapping areas between first generation and second generation e-commerce systems since agent-mediated e-commerce does not entirely rely on agent technology alone. Agent-mediated e-commerce is usually a mix of several other technologies most of which are from first generation e-commerce. The examples that we present in this section are based on previous work in literature [7, 33, 34, 35]. The presentation given in this section is not exhaustive, more detailed and integrated review is presented by He et al. [7]. 2.1.1

Agents in B2C E-Commerce

B2C e-commerce which involves selling of goods and services to a consumer by a business enterprise is very popular nowadays, since many people have recognized the convenience of shopping at the click of the mouse. However, as more people move towards online shopping or B2C e-commerce, so is the need to improve software technologies to meet the new demands of customization, quick retrieval of information and timely deliveries of goods and services. A Consumer Behavior Buying (CBB) model in which agents act as mediators in the five stages of need identification, product brokering, buyer coalition formation, merchant brokering and negotiation is shown in Figure 2 below.

7

Any other buyers?

Need identification

Merchant brokering

CBB Model

Product/service and evaluation H ow go is i ing t ?

Under what terms?

Need it?

Buyer coalition formation

to m ? ho om W y fr bu

Wh a buy t to ?

Product brokering

Negotiation

Purchase and delivery

s ard it c by d , e Cr y air B ship

Figure 2: Consumer Behavior Buying Model (Adopted from [33]; [36] ) To illustrate functional behavior of the CBB model, let us consider an example of an individual who wishes to buy a text book concerning Ugandan history using agent technology. The owner of the agent instructs this agent to look for a book on Ugandan history that was published less than 4 years ago, costing no more than $50 and should be new with hard cover. Armed with this information, the agent moves to available online markets dealing in book sales and searches for availability of books costing no more than $50 in addition to all the other attributes that the agent owner needs. On initial checks, the agent realizes that most shops are selling books at a price higher than $50, so the agent sets out to look for more markets that could probably be cheaper. Additionally, the agent monitors current shops and bargains for possible reductions in price and auction markets for probable cheaper options without actually performing a bid since such an action would require committing its master’s money. This agent also has the option of relaxing requirements of its master; these options may include considering books on Ugandan history that are more than 4 years old, books that are not hard-covered or even secondhand books that could be cheaper. The agent could also try to find out other agents with whom a coalition can be formed to buy a bigger package of books at a reduced price. The agent can then report back after performing negotiations on the availability of books concerning Ugandan history ranking them in the order of preference based on the master’s desires. The owner has possibilities of making a choice out of what the software agent has provided, or even rejecting them and instructing the agent with new requirements or constraints. 2.1.2

Agents in B2B E-Commerce

In this section we present an example in which agents can be employed in Business-toBusiness (B2B) transactions which are becoming more popular everyday as corporations build coalitions to leverage their businesses. The following B2B example presents a scenario in which an agent for an institution of learning obtains a request from a large organization in a country that wishes to train their technical and marketing departments 8

on issues concerning customer care. On receiving the request, the agent realizes that the resources currently available at the institution of learning are not sufficient for the number of employees that need to be trained since the training institution has small rooms and an inadequate number of training staff for the month in which the teaching is to be conducted. In such a situation, the agent sets out to look for other reputable institutions in the city that are probably less than 5 kilometers away, that can partner with its owner’s institution inorder to jointly offer the customer management course. After negotiations on terms of the contract with possible partner institutions, the agent returns results to the owner and a choice is made on either to accept the deal or to reject it.

2.2

The State of the Art in AMEC

Everyday, month and year that passes, the Internet grows with more information as many more information systems get connected on the Internet. As this growth and expansion goes on, it becomes more difficult and expensive for individuals and organizations to find what they want via e-commerce systems. For example, it was estimated in the year 2000 that over $5 trillion dollars was being spent on procurement of industrial parts each year[37] and it is believed that this figure has grown considerably as more businesses and industrial requirements have gone up. Of course many companies and organizations value this exercise and they invest heavily in information brokers or brokerage centres that act as intermediaries between the sellers and the buyers[33, 38]. The vision for state of the art e-commerce, is to provide agents or multi-agent systems that act as information brokers in a world rich with vast amounts of information. The information broker is expected to perform information retrieval and processing, a self-learning repository about the user or owner, monitoring on behalf of the owner for items that maybe of interest; commercial negotiation with service providers, vendors and collaborating with other brokers; and protecting the agent owner from compromising access. As indicated by He et al. [7] “most research effort in brokerage services has been focused on information search and matchmaking buyer’s and seller’s profiles as well as comparing the products in catalogs from different suppliers”. Advanced studies are yet to be done in the collaboration with brokers and protecting agent owners from intrusive access.

2.3

Characteristics of a Tool for AMEC

Agents (more particularly semantic-agents) were defined by Wooldridge et al. [12] as software programs that exhibit the following characteristics. i. Autonomy: agents operate without the direct intervention of humans or others, and have some kind of control over their actions and internal state. ii. Social ability: agents interact with other agents (and possibly humans) via some kind of agent-communication language. iii. Reactivity: agents perceive their environment, (which may be the physical world, a user via a graphical user interface, a collection of other agents, the Internet, and sometimes all of these combined), and respond in a timely fashion to changes that occur in it. 9

iv. Pro-activeness: agents do not simply act in response to their environment, they are able to exhibit goal-directed behavior by taking the initiative. It is also worth mentioning that agents are not just used in the domain of e-commerce, although its is the most popular domain of their application. For researchers in the Artificial Intelligence (AI) field, an agent is conceptualized using concepts that are directly applicable to human beings. Agents are more commonly characterized with notions such as knowledge, belief, intention and obligation[13]) in addition to the properties stated above. Since semantic-agents are able to act autonomous, socially, reactively and proactively, their use in electronic commerce presents several advantages and improvements in open distributed environments. In the listing below, some of the advantages that are attributed to semantic-agents in electronic commerce are presented. • Automation: Ability of semantic-agents to act autonomously present many opportunities for enabling automation in agent-mediated e-commerce. • Responsiveness: Since agent systems have reactive and pro-activeness properties, they are expected to facilitate timely responses to changes in their environment and business markets. • Flexibility: Research[13] from the AI field agrees with the fact that multi-agent systems can actually be conceptualized and developed with human like notions of knowledge, belief, intention and obligation which certainly implies that they (agents) can act flexibly in the environments in which they participate. • Scalability: As more information grows on the Internet currently available paradigms will most likely fall short of the new demands. Since semantic-agents can process information from the source and reason about its relevance to the agent owner and negotiating for goods and services without communicating very often with the sender, this paradigm scales well to the growing information in open distributed environments. • Economical: Resource utilization is low (in terms of bandwidth) for agent-mediated electronic commerce since agents minimize on the amount of communication and message exchanges along communication channels. Technologies for supporting development of semantic-agent system are already in place. Several languages are available for supporting interactions between agents are in place such as eXtensible Markup Language (XML)[39] and Agent Communication Languages (ACL)[40] that are used to code information and services in meaningful structures that agents can easily understand and process. In order to support semantic consistence for data tagged in XML, the Electronic Business eXtensible Markup Language (ebXML)[41, 42] has been developed based on international standards to provide an open technical framework for B2B and B2C e-commerce. However, there is a problem of semantic interaction among agents. As indicated by He et al. [7], “getting agents that have not been pre-designed to meaningfully interact 10

with one another in a sophisticated manner is one of the major challenges of computer science”. However, with the increasing convergence between the semantic web, grid computing and agent technology, there is commendable progress towards achieving a solution to this problem.

2.4

Security Techniques for Agents

This section presents a review of techniques that have been developed for protecting and detecting attackers on agents in open distributed environments. 2.4.1

Partial result encapsulation

Partial result encapsulation [43] is a technique that can be used to detect tampering by malicious hosts on the agent. Results of an agents actions are encapsulated at each platform visited for future verification using cryptography. This technique presents an overhead of results verification at several locations where it is done. Partial encapsulation additionally provides confidentiality and integrity by use of encryption and digital signatures. Open challenges There is need for providing better optimization techniques to be used during verification and routing of agents in order to reduce processing overheads. 2.4.2

Itinerary recording with replication and voting

This technique provides a mechanism for detecting faulty behavior of an agent platform by use of replication and voting[43, 44]. The technique provides fault tolerance to counter effects of malicious environments that could attempt to alter computational results. This approach would be well suited for environments in which availability of agents is highly required, and where replication does not present heavy performance penalty. Open challenges The technique imposes an overhead on agent platforms that may have to handle replicated computations of the same task. There is need for developing replication frameworks that provide the necessary fault tolerance with less performance overhead. 2.4.3

Execution Tracing

This approach uses faithful recording of agents behavior during execution to detect unauthorized modifications of an agent[43, 44]. Each execution platform is required to maintain a non-repudiable log (trace) of operations performed by an agent resident there. A cryptographic hash of the trace is submitted by the platform upon conclusion as a trace summary or fingerprint.

11

Open challenges Efficient generation of a non-repudiable trace for agents executing on various agent platforms and appropriate usage of fingerprints is still an open area of research. It would be interesting to investigate how this techniques performs in multi-threaded agents. 2.4.4

Environmental key generation

Environmental key generation intends to provide agents with the ability to take predefined actions whenever predefined environmental conditions are met. This technique has had indications of success in areas that are closely related to agent systems. Filiol et al. [45] illustrated the use of an environmental key as a virus cryptographic key. Open challenges Use of the environmental key generation techniques also presents risks that go with having an agent embedded with a key. There are still open questions concerning provision of effective protection for keys embedded within agents traversing insecure networks. The techniques that can used to define and protect environmental keys do present an open area of further research. 2.4.5

Computing with encrypted functions

Computing with encrypted functions is intended to provide a safe mechanism through which agents can execute cryptographic tasks in insecure environments without being compromised [46, 47]. Open challenges Citations in literature [43, 48] indicate that this technique though powerful, does not protect agents against several attacks such as replay, experimental extraction, and denial of service among others. Furthermore, computing with encrypted functions has been seen successful with rational functions, but similar success is yet to be seen with more general functions. 2.4.6

Obfuscated code

Obfuscation is another mechanism for protecting agents against attackers. Obfuscation can be used to scramble the code such that it is not easily read by an attacker [49]. Recent work by Kuzurin et al.[50] provides an optimistic picture of this technique towards security problems in computer security. Open challenges This technique has open challenges in the design of appropriate algorithms needed for performing obfuscation in applications, and formalizing of security requirements.

12

2.4.7

Trust estimation

Trust estimation is used in agent systems to provide a measure for correct behavior and interaction acceptance among agents. Various approached to trust estimation exist, some of which are based on a trusted third party, metrics and inference rules, and reputation models[9]. Reputational models are used in mobile agents and other areas of distributed computing to keep record of behavior for an entity. Open challenges Techniques used in literature[9, 10] have not dealt with possibilities of subverted reputation models. Moreover these studies have also been too generic, and thus failing to meet or consider metrics in reputation models that would be relevant in e-commerce for example. Reputation models still face problems of strategic liars and informal expectations in trust models that result in confusing trust contexts. There is still a need for mechanisms of combating strategic liars and addressing informal expectations.

2.5

Conclusion

The mobile computing paradigm is novel for state of the art electronic commerce. Ground breaking research has already been done to provide for functional requirements. As indicated in previous sections of literature review, security challenges are still of concern to the wide adoption of semantic-agent mediated e-commerce. There is hope in alleviating existing security problems by improving existing security techniques and proposing new approaches. This research work will contribute to solutions for addressing security problems that have been presented in subsection 1.3 for business-to-consumer transactions and to improve on existing techniques whose open challenges have been presented in subsection 2.4.

3

Methodology

To achieve the stated objectives and anticipated results, the following methodology will be used.

3.1

Analytical Studies

Investigations will be carried out to obtain in-depth knowledge of security requirements and limitations for semantic-agents, agent systems operations and semantic web technologies. The acquired knowledge will be used to reason about security requirements and challenges that will have to be considered during the stage of implementing a prototype system. In order to develop comprehensive security framework for agent mediated e-commerce in open distributed environments, this research work will analyze and improve on mechanisms for facilitating trust, authentication, authorization, privacy and anonymity among others. The security specifications will be used in designing a framework in which semantic-agent systems can be used for e-commerce business-to-consumer transactions. 13

The developed security framework will be used as a basis for improving existing security techniques that are discussed in subsection 2.4 of the literature review. Figure 3 shows a model that will be used to investigate security requirements of agent mediated e-commerce applications. On the left hand side of the model (Figure 3) are security goals

Figure 3: Conceptual Model that are generic to a large number of agent applications. On the right hand side are security techniques (discussed in literature review subsection 2.4) that have been found relevant for secure agent-mediated e-commerce applications. The conceptual model aims to find the best combination of security goals and security techniques. The research work will also reflect on the process of coming up with the security framework.

3.2

Case Study

In order to test and validate the security framework for agent-mediated e-commerce applications, a prototype e-commerce application will be developed to be used as a case study. A prototype application will be implemented to investigate and demonstrate support for secure business-to-consumer transactions. Implementation of a booktrading application based on the JADE agent platform[29] has already been started for this purpose. In this application, agent based bookshops and book-buyers are expected to negotiate for prices and to choose best offers based on the consumer behavior buying model in Figure 2. 3.2.1

Requirements Analysis

During this phase, user requirements will be defined and documented based on findings from the literature review and analytical studies. The requirements document will mainly

14

define security requirements expected by the users. The output of this phase will be a user requirements document. 3.2.2

System Design

This phase will aim to achieve both the high-level and low-level design of the agentmediated e-commerce case study. The designs will be necessary for realization of the system requirements within the proposed budget and schedule. The high-level designs will define methods by which requirements will be achieved. At the lower-level this phase will be used to define modules, their functionalities, relationships and dependencies. A software specifications document for the development phase and system testing document will be generated. The software specifications document will then be used to break down the system into smaller programmable modules. Several agent development methodologies have been identified in literature[51, 52]. These methodologies aim to simplify the development process for intelligent agents and their recommendations will be considered during the design phase. 3.2.3

Implementation

In order to perform a prototype implementation, research will be done to evaluate existing middleware solutions for appropriateness in facilitation of the required system requirements. Preliminary evaluations have already been done on three middleware solutions namely, AgentScape[27, 28], JADE[29] and SEMOA[30] that will be considered for probable support in the development process. Fortunately, these middleware platforms are all open source and features that will be found missing and unsupported at the time of our systems development will have to be added to the middleware. The language of choice for systems development will be mainly Java[53], Web Ontology Language (OWL)[18] and scripting languages such as python which will be used if need arises. The main reason for using Java is because of its wide based of deployment in agent system and middleware development. Java being platform independent and widely supported on a variety of ubiquitous devices, we expect to face few challenges in achieving interoperability. 3.2.4

Testing

System testing for the proposed security framework will be done using the e-commerce case study. The testing process will be carried out during and after the development stage using component-based testing techniques[54]. Since the e-commerce case study will be developed incrementally over a period of three years, component-based testing is the most appropriate technique for enabling testing of components based on both adhoc and formal approaches. A comprehensive system test will be done at the end of the implementation process using the system testing document developed during the system design phase (3.2.2).

15

3.3

Concluding Remarks

In order to develop a solid security framework, security analysis will be done using case study (booktrading application) to guide the process for indentifying key security requirements that are important for a typical agent mediated e-commerce application. As noted in subsection 2.4, several security techniques exist for dealing with generic agent security problems. However, applying security techniques in a generic manner does not always guarantee a secure system. The implemented case study will be useful in understanding security techniques that are relevant for agent mediated e-commerce. After identifying security techniques relevant for secure agent mediated e-commerce, the scope of the project will narrow down to improving the identified techniques.

16

References [1] J. Coppel, “E-commerce: impacts and policy challenges,” Organisation for Economic Co-operation and Development Paris, 2000. [2] M.S. Eastin, “Diffusion of e-commerce: an analysis of the adoption of four ecommerce activities,” Telematics and Informatics, vol. 19, no. 3, pp. 251–267, 2002. [3] eBay Inc, “eBay,” http://www.ebay.com/, accessed on 20th December 2006. [4] J. Gebauer, C. Beam, and A. Segev, “Impact of the Internet on Procurement,” Acquisition Review Quarterly, vol. 5, no. 2, pp. 167–184, 1998. [5] TS Liao, MT Wang, and HP Tserng, “A framework of electronic tendering for government procurement: a lesson learned in Taiwan,” Automation in Construction, vol. 11, no. 6, pp. 731–742, 2002. [6] A. Davila, M. Gupta, and R. Palmer, “Moving procurement systems to the internet: the adoption and use of e-procurement technology models,” European Management Journal, vol. 21, no. 1, pp. 11–23, 2003. [7] M. He, N.R. Jennings, and H.F. Leung, “On agent-mediated electronic commerce,” IEEE Transactions on Knowledge and Data Engineering, vol. 15, no. 4, pp. 985–1003, 2003. [8] N.R. Jennings, “An agent-based approach for building complex software systems,” Communications of ACM, vol. 44, no. 4, pp. 35–41, 2001. [9] L.D. Molm, N. Takahashi, and G. Peterson, “Risk and Trust in Social Exchange: An Experimental Test of a Classical Proposition,” AJS, vol. 105, no. 5, pp. 1396–1427, 2000. [10] SD Ramchurn, D Huynh, and NR Jennings, “Trust in multi-agent systems,” The Knowledge Engineering Review, vol. 19, no. 01, pp. 1–25, 2005. [11] R. Till, “Transforming the Way We Do Business,” Electronic Commerce, T. Nash, ed, pp. 9–12, 1998. [12] M. Wooldridge and N.R. Jennings, “Intelligent Agents: Theory and Practice,” Knowledge Engineering Review, vol. 10, no. 2, pp. 115–152, 1995. [13] K.P. Sycara, “Multiagent Systems,” AI Magazine, vol. 19, no. 2, pp. 79–92, 1998. [14] N. Shadbolt, W. Hall, and T. Berners-Lee, “The semantic Web revisited,” IEEE Intelligent Systems, vol. 21, no. 3, pp. 96–101, 2006. [15] T. Berners-Lee, J. Hendler, and O. Lassila, “The semantic Web,” Scientific American, vol. 284, no. 5, pp. 28–37, 2001. [16] W3 Consortium, “Semantic web activity,” http://www.w3.org/2001/sw, 2007.

17

[17] M. Dean and G. Schreiber, “OWL Web Ontology Language Reference,” W3C Recommendation, vol. 10, 2004. [18] W3 Consortium, features/, 2004.

“Owl web ontology language,”

http://www.w3.org/TR/owl-

[19] J. Hendler, “Agents and the Semantic Web,” Intelligent Systems, IEEE [see also IEEE Intelligent Systems and Their Applications], vol. 16, no. 2, pp. 30–37, 2001. [20] D.H. McKnight, V. Choudhury, and C. Kacmar, “Developing and Validating Trust Measures for e-Commerce: An Integrative Typology,” Information Systems Research, vol. 13, no. 3, pp. 334–359, 2003. [21] T.D. Huynh, N.R. Jennings, and N.R. Shadbolt, “An integrated trust and reputation model for open multi-agent systems,” Autonomous Agents and Multi-Agent Systems, vol. 13, no. 2, pp. 119–154, 2006. [22] C. Schl¨ager, “A Reference Model for Authentication and Authorisation Infrastructures Respecting Privacy and Flexibility in b2c eCommerce,” The First International Conference on Availability, Reliability and Security (ARES 2006), pp. 20–22, 2006. [23] W. Hong and K. Zhu, “Migrating to internet-based e-commerce: factors affecting ecommerce adoption and migration at the firm level,” Information and Management, vol. 43, no. 2, pp. 204–221, 2006. [24] Martijn Warnier, David de Groot, and Frances Brazier, “Organized Anonymity in Agent Systems,” in Informal Proceedings of the Fourth European Workshop on Multi-Agent Systems (EUMAS’06), 2006. [25] Joann j. Ordille, “When agents roam, who can you trust?,” Proceedings, First Annual Conference on Emerging Technologies and Applications in Communications, 1996. [26] B. Matthews, “Semantic Web Technologies,” E-learning, vol. 6, no. 6, pp. 8, 2005. [27] F.M.T. Brazier, D.G.A. Mobach, B.J. Overeinder, E. Posthumus, S. van Splunter, M. van Steen, and N.J.E. Wijngaards, “Agentscape demonstration,” in Proceedings of the Fourteenth Belgium-Netherlands Conference on Artificial Intelligence (BNAIC2002), H. Blockeel and M. Denecker, Eds., October 2002, pp. 513–514. [28] Vrije University Amsterdam, http://www.iids.org/research/aos, 2006.

“Agentscape

operating

system,”

[29] Telecom Italia Lab, “Java Agent DEvelopment Framework,” http://jade.tilab.com, 2006. [30] Fraunhofer IGD, “Secure Mobile Agents Project, Department Security Technology ,” http://www.semoa.org/, 2006. [31] G.C. Necula and P. Lee, “Safe, untrusted agents using proof-carrying code,” Special Issue on Mobile Agent Security, pp. 61–91, 1997. 18

[32] IDC, “The premier global market intelligence firm,” http://www.idc.com/, accessed on 5th January 2007. [33] R.H. Guttman, A.G. Moukas, and P. Maes, “Agent-mediated electronic commerce: a survey,” The Knowledge Engineering Review, vol. 13, no. 02, pp. 147–159, 2001. [34] H.S. Nwana, J. Rosenschein, T. Sandholm, C. Sierra, P. Maes, and R. Guttmann, “Agent-mediated electronic commerce: issues, challenges and some viewpoints,” Proceedings of the second international conference on Autonomous agents, pp. 189–196, 1998. [35] C. Preist, J. Esplugas-Cuadrado, S.A. Battle, S. Grimm, and S.K. Williams, “Automated Business-to-Business Integration of a Logistics Supply Chain using Semantic Web Services Technology,” 4th International Semantic Web Conference, Galway, Ireland, 2005. [36] P. Maes, R. Guttman, and A. Moukas, “Agents that buy and sell: Transforming commerce as we know it,” Communications of the ACM, vol. 42, no. 3, 1999. [37] S. Tully, “The B2B tool that really is changing the world,” Fortune, vol. 141, no. 6, pp. 132–45, 2000. [38] S.H. Ha and S.C. Park, “Matching buyers and suppliers: an intelligent dynamic exchange model,” Intelligent Systems, IEEE [see also IEEE Intelligent Systems and Their Applications], vol. 16, no. 4, pp. 28–40, 2001. [39] R.J. Glushko, J.M. Tenenbaum, and B. Meltzer, “An XML framework for agentbased E-commerce,” Communications of the ACM, vol. 42, no. 3, 1999. [40] N. Gibbins, S. Harris, and N. Shadbolt, “Agent-based Semantic Web Services,” Web Semantics: Science, Services and Agents on the World Wide Web, vol. 1, no. 2, pp. 141–154, 2004. [41] OASIS Open 2006, “ebXML - Enabling A Global Electronic Market,” http://www.ebxml.org/, accessed on 13th January 2007. [42] Hans Weigand Wilhelm Hasselbring, “Languages for electronic business communication: state of the art,” Industrial Management & Data Systems, vol. 101, no. 5, pp. 217–227, 2001. [43] WA Jansen, “Countermeasures for mobile agent security,” Computer Communications, vol. 23, no. 17, pp. 1667–1676, 2000. [44] M. Alfalayleh and L. Brankovic, “AN OVERVIEW OF SECURITY ISSUES AND TECHNIQUES IN MOBILE AGENTS,” Communications And Multimedia Security: 8th IFIP TC-6 TC-11 Conference on Communications and Multimedia Security, Sept. 15-18, 2004, Windermere, The Lake District, United Kingdom, 2005. [45] E. Filiol, “Strong Cryptography Armoured Computer Viruses Forbidding Code Analysis: the,” Proceedings of the 14th EICAR Conference, pp. 216–227, 2005. 19

[46] IBM, “Mobile Code Security and Computing with Encrypted Functions,” http://www.zurich.ibm.com/security/mobile/, 2007. [47] H. Reiser and G. Vogt, “Security Requirements for Management Systems using Mobile Agents,” Tohme, S. und M. Ulema (Herausgeber): Proceedings of the Fifth IEEE Symposium on Computers & Communications, pp. 160–165. [48] S. Silver, “Implementation and Analysis of Software-Based Fault Isolation,” Tech. Rep., Dartmouth College Technical Report PCS-TR96-287, 1996. [49] T. Sander and C.F. Tschudin, “Protecting Mobile Agents Against Malicious Hosts,” Mobile Agents and Security, vol. 60, 1998. [50] N. Kuzurin, A. Shokurov, N. Varnovsky, and V. Zakharov, “On the Concept of Software Obfuscation in Computer Security?,” . [51] M. Wooldridge, N.R. Jennings, and D. Kinny, “The Gaia Methodology for AgentOriented Analysis and Design,” Autonomous Agents and Multi-Agent Systems, vol. 3, no. 3, pp. 285–312, 2000. [52] L. Padgham and M. Winikoff, “Prometheus: A Methodology for Developing Intelligent Agents,” Agent-Oriented Software Engineering III: Third International Workshop AOSE 2002, Bologna, Italy, July 15, 2002: Revised Papers and Invited Contributions, 2003. [53] Sun Developer Network (SDN), “Java technology,” http://java.sun.com, 2006. [54] M. van der Bijl, A. Rensink, and J. Tretmans, “Compositional Testing with IOCO,” FATES, vol. 2931, pp. 86–100, 2003.

20