PT Unknown AU Doan, TS TI On Trustworthiness Recommendation PY 2017 PU Fernuniversität WP https://ub-deposit.fernuni-hagen.de/receive/mir_mods_00000940 LA en AB Nowadays, people use more and more the Internet and On-line Social Networks as their preferred media for communication as well as business and financial transactions. In particular, the technological progress signifcantly increases possibilities to interact with people over big distances. However, any first contacts come along with incalculable risks. Without the physical presence of a partner, the usual six human senses cannot give people a feeling of trust, safety and security. Even over longer periods, on-line partners are hard to evaluate. Therefore, many users look for any non-subjective possibility to get recommendations on trustworthiness of Internet partners for private communication and/or business. In this thesis, a new model for trustworthiness estimation is introduced. It calculates the trustworthiness of a user by an evaluation of his/her activities with all partners over a longer time. This practice directly corresponds to the human behaviour and psychology and may, therefore, insure a high acceptance among the user community. Basing on random walks, a decentralised method is derived to combine pairwise, locally kept trustworthiness evaluations into a global trustworthiness value or recommendation on trustworthiness for any participating user. A set of simulations shows the evidence and practicability of the introduced approach. Furthermore, a decentralised, P2P-based approach for its implementation is suggested, which may be employed in parallel to existing on-line social network platforms like Facebook or Google+. It allows to obtain the wanted trustworthiness recommendation for each participant. Differing from existing implementations, it safely keeps sensitive (since private) data, since they are stored in a distributed, local manner including a fraud as well as a privacy protection and (limited) owner control. PI Hagen ER