A Taxonomy to Express Open Challenges in Trust and Reputation Systems

Effective and reliable trust and reputation modeling systems are central to the success of decentralized e-commerce systems where autonomous agents are relied upon to conduct commercial transactions. However, the subjective and social-based qualities that are inherent to trust and reputation introduce many complexities into the development of a reliable model. Existing research has successfully demonstrated how trust systems can be decentralized and has illustrated the importance of sharing trust information, or rather, modeling reputation. Still, few models have provided a solution for developing an initial set of advisors from whom to solicit reputation rankings, or have taken into account all of the social criteria used to determine trustworthiness. To meet these objectives, we propose the use of two new parameters in trust and reputation modeling: agent lifetime and total transaction count. We describe a model that employs these parameters to calculate an agent’s seniority, then apply this information when selecting agents for soliciting and ranking reputation information. Experiments using this model are described. The results are then presented and discussed to evaluate the effect of using these parameters in reputation modeling. We also discuss the value of our particular model in contrast with related work and conclude with directions for future research.

In recent years, we have witnessed an increasing use of trust and reputation systems in different areas of ICT. The idea at the base of trust and reputation systems is of letting users to rate the provided services after each interaction. Other users may use aggregate ratings to compute reputation scores for a given party. The computed reputation scores are a collective measure of parties trustworthiness and are used to drive parties interactions. Due to the widespread use of reputation systems, research work on them is intensifying and several models have been proposed. This calls for a methodology for the analysis and the evaluation of trust and reputation systems that can help researcher and developers in studying, designing and implementing such systems. In this thesis we propose different kinds of theoretical results and software tools that could be useful means for researchers and developers in area of trust and reputation systems. Our work addresses the three main stages of trust and reputation systems development, namely study, design and implementation. We provide: 1) a general framework based on Bayesian decision theory for the assessment of trust and reputation models, 2) an analysis methodology for reputation systems based on a coordination language, 3) a software tool for network-aware evaluation of reputation systems and their rapid prototyping.

Trust and reputation systems represent a significant trend in decision support including selection of best match cloud providers to process Big Data. Reputation is often considered as a collective measure of trustworthiness based on the referrals or ratings from members in a community. Reputation systems have been applied in various applications such as online service provision. However, reputation models do not reflect user's quality of service (QoS) preferences and thus they might not be satisfied with the recommendations from others. In this paper, we propose a de-centralized reputation-based trust model that incorporates the user QoS preferences to select the best match Cloud Service Provider to process Big Data. Our trust model relies on three multi-attribute decision-making (MADM) methods including Simple Additive Weighting (SAW), Weighted Product Method (WPM), and Technique for Order Preference by Similarity to Ideal Solution (TOPSIS). We conducted several experiments using simulated cloud environment to validate our trust model and assess the three MADM methods. The results show that the proposed model is pliable to users' requirements and efficiently evaluate trust of cloud providers.

Trust and reputation systems are widely used, facilitating interactions among agents based on trust evaluation. These systems may have robustness issues, which should be paid attention. Various attacks in existing trust and reputation systems have been identified. Designers of trust systems propose methods to defend against these attacks. However, they typically verify the robustness of their defense mechanisms (or trust models) only under specific attacks. This raises problems: First, the robustness of their models is not guaranteed as they do not consider all attacks. Second, the comparison between two trust models depends on the choice of specific attacks, introducing bias. We propose to quantify the strength of attacks, and to quantify the robustness of systems based on the strength of the attacks it can resist. Our quantification is based on information theory. Our approach provides designers of trust systems a fair measurement of the robustness.

The Cloud Computing (CC) is an Internet-based technology which offers a shared pool of highly available, virtualized, dynamically scalable, and configurable computing resources (e.g., networks, servers, computing power, storage, applications, data, services, etc.). In the CC, the authentication as well as trust and reputation of the cloud users and service provider is a critical issue for the security of the cloud. Lots of articles previously present different trust models, but unfortunately, these trust models are not suitable for all cases as the trust is a non-measurable and subjective parameter that differs person to person. The trust involves many multi-dimensional factors such as evidence, reputation, expectations, feedback, history, risk, motivation, session behaviours, and so on. To addresses the above-mentioned issues, the paper present a novel multi-dimensional trust and reputation calculation model for the cloud computing, in which we integrate multiple trust factors that increase the novelty of the model. Additionally, we dynamically assign a weight for each trust and reputation factor using the Weighted Moving Average and Ordered Weighted Averaging (WMA-OWA) combination algorithm. The proposed model overcome the issues of the previous existing models. The detail presentation and analysis of the model shows that the model is appropriate and accurate for trust calculation in the cloud.

Trust and reputation systems provide a foundation for security, stability, and efficiency in the online environment because of their ability to stimulate quality and to sanction poor quality. Trust and reputation scores are assumed to represent and predict future quality and behaviour and thereby to provide valuable decision support for relying parties. This assumption depends on two factors, primarily that trust and reputation scores faithfully reflect past observed quality, and secondly that future quality will be truly similar to that represented by the scores. Unfortunately, poor robustness of trust and reputation systems often makes it relatively easy to manipulate these factors, so that the fundamental assumption behind trust and reputation systems becomes questionable. On this background we discuss to what degree robustness against strategic manipulation is important for the usefulness of trust and reputation systems in general.

With the increasing popularity of self-organized communication systems, distributed trust and reputation systems in particular have received increasing attention. By formalizing trust relationships, trust between parties within the community can be derived by analyzing the trust paths linking the parties together. This paper treats trust network as an emergent property. The emergence interpretation concerns both the maintenance and usage of trust network inference. Specifically, in P2P system, trust network is maintained by individual agents at micro level, and it is used (i.e., queried) as one entity at macro level. In this paper, we first discuss micro level activities, that is, we classifies trust into functional trust and referral trust to facilitate trust inference, and extend the referral trust to include factors of similarity and truthfulness, so that our approach can accommodate the personalized feature of reputation rating, and reduce trust inference error effectively; then we discuss macro level properties of trust network. Specifically, we investigate the emergence of network structural properties of trust and reputation system in terms of efficiency and cost. That is, efficiency measures how well information propagates over trust system, and cost measures how expensive it is to build this system. Preliminary simulation results show the performance improvement of P2P community and the emergence of economical small-world trust network, namely relatively high efficiency and low cost.

It is argued that group-based trust metric is effective resisting attacks, which evaluates groups of assertions in tandem, and generally computes trust ranks for sets of individuals according to peers' social position trust network. Thus, the group-based trust value should be called reputation rank. But, unfortunately, most group-based trust metrics are vulnerable to the attack of front peers, which represent these malicious colluding peers always cooperate with others order to increase their reputation, and then provide misinformation to promote actively malicious peers. In this paper, we proposed adaptive spreading activation approach to mitigating the effect of front peer attack, which adaptive spreading factor is used to reflect the peer's recommendation ability according to behaviors of the peer's direct/indirect children trust network; Simulation results show that the adaptive spreading activation approach can identify and mitigate the attack of front peer.

Trust and reputation systems have been regarded as a powerful tool to defend against insider attacks caused by the captured nodes in wireless sensor networks (WSNs). However, trust systems are vulnerable to on-off attacks, in which malicious nodes can opportunistically behave good or bad, compromising the network with the hope that bad behavior will be undetected. Thus, malicious nodes can remain trusted while behaving badly. In this paper, we propose O2Trust, On-Off attack mitigation for Trust systems in wireless sensor networks. O2Trust adopts the penalty policy against the misbehavior history of each node in the network as a reliable factor that should influence on the calculation of the trust value. This punishment future helps to perceive malicious node that aim to launch intelligent attacks against trust-establishment and consequently on-off attack is mitigated efficiently.

    Peer-to-peer (P2P) E-commerce systems are more and more popular in recent years with opportunities and threats both existing. A lot of malicious nodes threaten the system’s security via cheating, selfish and attacking or else. Trust mechanisms can help estimating the trustworthiness. This paper presents a distributed trust model named P2PETrust (peer-to-peer E-commerce trust model) based on social networks’ principles. A transaction history vector with multi dimensions is designed to describe the history before the distributed storage structure. In order to compute the trust, three sub trust models are defined in P2PETrust, which are local model, global model and correlation model. The local trust model is the base trust mechanism to compute history between two nodes after defining the nonlinear time factor, transaction amount factor, frequency factor and success rate for the history while the global one is based on local model to calculate a node’s whole transaction histories with some global factors defined. On the basis of these jobs, a correlation trust model is presented by calculating nodes’ correlative similarity multiplied with global trust degree. We present a simulation environment to validate the proposed trust model and report the set of initial experiments, showing the feasibility and benefit of our approach.   Key words: Peer-to-peer E-commerce, trust model, P2P security, social network, network security.  

User reputation systems are widely used in Ecommerce website and social networks. In present most of the user reputation systems use the rule-based method or the voting systems to calculate user reputations. These systems heavily depend on the experience of experts. In this paper we try to use machine learning method to automatically learn user reputation in social networks. The social network we selected is a financial forum. A social network is seen as a directed graph, every user in the networks is a node in the graph, and the interactions between the users are the directed edges. Then we extract features of users from the social network graph. We translate the reputation learning problem into the document ranking problem, and use the listwise based rank learning method to build the reputation model. The reputation prediction model is represented as a linear model. We use the model to predict user reputation. The experimental results show that using rank learning method to predict user reputation is effective.

Today, WSNs are spread in both industry and academia; they are focusing their research efforts in order to enhance their appliances. One of the first concerns to solve in order to acquire that expected enrichment is to assure relieve a minimum level of security in such a prohibitive environment. This study concentrates on trust and reputation system management. The proposed approach titled enhanced bio-inspired trust and reputation model (EBTRM) is Bio-inspired extending Trust and Reputation Model. The aim of the proposed algorithm is to provide an adequate security solution to collusion network of BTRM, which can provide a high level of security and energy preserving ability

Traders in electronic marketplaces may behave dishonestly, cheating other agents. A multitude of trust and reputation systems have been proposed to try to cope with the problem of cheating. These systems are often evaluated by measuring their performance against simple agents that cheat randomly. Unfortunately, these systems are not often evaluated from the perspective of security---can a motivated attacker defeat the protection? Previously, it was argued that existing systems may suffer from vulnerabilities that permit effective, profitable cheating despite the use of the system. In this work, we experimentally substantiate the presence of these vulnerabilities by successfully implementing and testing a number of such 'attacks', which consist only of sequences of sales (honest and dishonest) that can be executed in the system. This investigation also reveals two new, previously-unnoted cheating techniques. Our success in executing these attacks compellingly makes a key point: security must be a central design goal for developers of trust and reputation systems.

The growing development of Web based trust and reputation systems in the 21st century will have powerful social and economic impact on all business entities, and will make transparent the quality assessment and customer assurance realities in the distributed Web based service oriented environments. The Web based trust and reputation systems will be the foundation for Web intelligence in the future. Trust and reputation systems help capture business intelligence through establishing customer relationships, learning consumer behavior, capturing market reaction on products and services, disseminating customer feedback, buyers' opinions and end-user recommendations, and revealing dishonest services, unfair trading, biased assessment, discriminatory actions, fraudulent behaviors, and untrue advertising. The continuing development of these technologies will help in the improvement of professional business behavior, sales, reputation of sellers, providers, products and services. However there has been no effort in defining ontology for reputation. In this paper we attempt to define ontology for reputation.

Reputation management is a powerful security tool that helps establish the trustworthiness of users in online applications. One of the most successful uses of reputation systems is on e-commerce web sites such as eBay.com and Amazon.com, which use reputation systems to root out fraudulent sellers. Reputation systems can also play an important role in social networks to enforce various security requirements. For example, a reputation system can help filter fake user profiles. However, a major challenge in developing reputation systems for social networks is that users often hesitate to publicly rate fellow users or friends due to the fear of retaliation. This trend prevents a reputation system from accurately computing reputation scores. Privacy preserving reputation systems hide the individual ratings of users about others and only reveal the aggregated community reputation score thus allowing users to rate without the fear of retaliation. In this chapter, we describe privacy preserving reputation management in social networks and the associated challenges. In particular, we look at privacy preserving reputation management in decentralized social networks, where there is no central authority or trusted third parties, thus making the task of preserving privacy particularly challenging.