We suggest the concept of multitrust, which is aimed at computing trust by collectively involving a group of trustees at the same time: the trustor needs the concurrent support of multiple individuals to accomplish its task. We propose Soft Constraint Logic Programming based on semirings as a mean to quickly represent and evaluate trust propagation for this scenario. To attain this, we model the Web of Trust adapting it to a weighted and-or graph, where the weight on a connector corresponds to the trust feedback value among the connected peers. Semirings are the parametric and flexible structures used to appropriately represent trust metrics. (Corresponding author: Santini, Francesco)

Recommender Systems (RS) aim at predicting items or ratings of items that the user are interested in. Collaborative Filtering (CF) algorithms such as user- and item-based methods are the dominant techniques applied in RS algorithms. To improve recommendation quality, metadata such as content information of items has typically been used as additional knowledge. With the increasing popularity of the collaborative tagging systems, tags could be interesting and useful information to enhance RS algorithms. Unlike attributes which are "global" descriptions of items, tags are "local" descriptions of items given by the users. To the best of our knowledge, there hasn't been any prior study on tag-aware RS. In this paper, we propose a generic method that allows tags to be incorporated to standard CF algorithms, by reducing the three-dimensional correlations to three two-dimensional correlations and then applying a fusion method to re-associate these correlations. Additionally, we investigate the effect of incorporating tags information to different CF algorithms. Empirical evaluations on three CF algorithms with real-life data set demonstrate that incorporating tags to our proposed approach provides promising and significant results. (Corresponding author: Tso-Sutter, Karen H.L.)

Recommendation systems, based on collaborative filtering, offer a means of sifting through the enourmous amounts of content on the web, by composing user ratings in order to generate predicted ratings for other users. This kind of system can be viewed as a network of interacting peers, where each user is a node and the links to all other nodes are weighted according to how similar the corresponding users are. Predicted ratings are generated by a user for unknown items by requesting and aggregating rating information from the surrounding neighbors. However, the different methods of computing user similarity, or weighting the network links, very often do not agree with each other as to the quality of recommendations that a pair of users could exchange, and in doing so they change the entire structure of the network of recommenders. In this work we perform an analysis of a range of similarity measures, comparing their performance in terms of prediction accuracy and coverage, based on the network perspective of a collaborative filtering environment. Using the results we obtain, we argue that user-similarity may not sufficiently capture the relationships that recommenders could otherwise share in order to maximise the performance of these collaborative communities. (Corresponding author: Lathia, Neal)

Ad Hoc networks do not rely on any centralized administration or fixed network infrastructure and their nodes establish a routing structure in a self-organized way, by means of an ad hoc routing protocol such as OLSR. Ad hoc route discovery and maintenance introduce specific security problems for routing protocols to prevent, detect or respond. Solutions to secure these routing protocols using some centralized units or trusted third-parties actually constrain the self-organization of ad hoc networks. In this paper, we propose for OLSR the integration of trust reasonings into each node behavior, so as to allow a self-organized trust-based control to help nodes to detect misbehavior attacks. Our analysis of OLSR brings out the trust rules that characterize this protocol and allows us to express formally the trust-related properties that can be verified by each node to assess the correct behavior of the other nodes. Simulation of OLSR with nodes reasoning on trust allows us to demonstrate the effectiveness of our approach and to compare trust-based routing choices with the bare OLSR reachability-based choices. (Corresponding author: Adnane, Asmaa)

Generating adequate recommendations for newcomers is a hard problem for a recommender system (RS) due to lack of detailed user profiles and social preference data. Empirical evidence suggests that the incorporation of a trust network among the users of the RS can leverage such ‘cold start’ (CS) recommendations. Hence, new users should be encouraged to connect to the network as soon as possible. But whom should new users connect to? Given the impact this choice has on the delivered recommendations, it is critical to guide newcomers through this early stage connection process. In this paper, we identify key figures in the trust network (in particular mavens, connectors and frequent raters) and investigate their influence on the coverage and accuracy of a collaborative filtering RS. Using a dataset from Epinions.com, we demonstrate that the generated recommendations for a new user are more beneficial if they connect to an identified key figure compared to a random user. (Corresponding author: Victor, Patricia)

Reputation systems are proved mechanisms used to help nodes to decide whom to trust, to maintain the overall credibility of the system and to promote collaboration. This paper presents Surework, a reputation framework based on Super-peers. In Surework, peers form clusters around Super-reputation-peers (Sure-peers) who help to increase the reputation knowledge. Surework introduces incentives in order to promote that nodes with higher capabilities become Super-peers and assume more tasks than normal peers. Reciprocity is also promoted by encouraging peers to provide better services to most reputable client peers. (Corresponding author: Rodriguez-Perez, Manuel)

The requirements for spontaneous interactions in open and dynamic systems create security issues and necessitate the incorporation of trust management into each software entity to make decisions. Trust encompasses various quality attributes (e.g., security, competence, honesty) and helps in making appropriate decisions. In this paper, we present CAT, an interaction-based Context-Aware Trust model for open and dynamic systems by considering services as contexts. We identify a number of trust properties including context and risk awareness and address those in the proposed model. A context-similarity parameter is proposed to make decisions in similar situations. A time-based ageing parameter is introduced to change trust values over time without any further interaction. We present direct and indirect recommendations and apply path-based ageing on indirect recommendations. A mechanism to calculate the accuracy of recommendations is described. This accuracy is used to differentiate between reliable and unreliable recommendations in the total trust calculation. (Corresponding author: Uddin, Mohammad)

This work investigates the use of keywords and classes to represent user’s profiles in order to improve a content-based recommender system. The techniques were implemented and tested in a recommender system for a website that gathers commercial ads. Ads are posted by individuals and contain a title and a textual description. Profiles are created and maintained through the analysis of ads seen by the user during a certain period of time and may be represented by classes, keywords or both kinds. Keywords are automatically extracted from the textual description of the ads. Classes come from a taxonomy defined by the website. Ads must be posted within a leaf class of the taxonomy. The items to be recommended are ads containing keywords associated to the user in his/her profile and/or ads classified in the leaf-classes present in the user’s profile. The paper demonstrates that the combination of both techniques (keywords and classes) outperforms the use of each one separately. (Corresponding author: Loh, Stanley)

In large-scale open systems such as eBay one of the key concerns in increasing the utility of users is having a trustworthy method for users to determine which interactions will be satisfactory and which are liable to lead to disappointment. Rather than starting from the point of assuming there are "good" and "bad" users we will examine why we can make such a distinction in this context and how humans mitigate some of the problems which seem endemic to such a system through game modification. We then demonstrate a use of this model of behaviour in simulating a particular agent choice in order to show the conditions under which different reputation systems affect an agent's trustworthiness, before briefly describing possible future directions of research to deal with the truly disenfranchised agents. (Corresponding author: Martin-Hughes, Rowan)