Estimate an evaluation of trust, for any member in a non centralized network/graph.
- a member knows only his list of contacts.
- The member evaluates the trust for each of his contacts.
- The evaluation given to his contacts is wieghted by his own evaluation :
- if the member has a low evaluation in the network, his evaluation of his contacts does not matter so much
- a member must be able to estimate the evaluation of all members (contacts of contacts, recursively).
- members receive/send tokens "permanently". These tokens capture information of the network/graph.
- When a member receives a token, the member :
- adds a triplet to the token : member_id , evaluation , contact_id. Then he sends this token updated to the contact.
- estimates evaluation of members, using the tokens already received
- detects potential frauds, using the tokens already received
- limits useless "circular" movements of tokens in the network
- evaluate_graph.py : Contribution from Vincent Yernaux. This python-script creates/display a complete graph (nodes, notes, edges oriented/valued between nodes). It makes a calculation of the trust for each node. It shows that the calculation can be stable after few iterations. This scripts needs to know the whole graph. To be done : for each node C - which knows only his contacts Ld(c) - , C should be able to discover by himself the whole graph (sending a token ?), calculate it like proposed in the script, and verify its result against Ld(C) (sending a token ?)