Êtes-vous un étudiant de l'EPFL à la recherche d'un projet de semestre?
Travaillez avec nous sur des projets en science des données et en visualisation, et déployez votre projet sous forme d'application sur Graph Search.
We study a new application of threshold-based secret sharing in a distributed online social network (DOSN), where users need a means to back up and recover their private keys in a network of untrusted servers. Using a simple threshold-based secret sharing in such an environment is insufficiently secured since delegates keeping the secret shares may collude to steal the user's private keys. Adversary can then take control of users' machines, infect them with malicious software, and use them for further attacks. This can lead to an epidemic that makes the whole system eventually collapse. To mitigate this problem, we propose using different techniques to improve the system security: by selecting only the most reliable delegates for keeping these shares and further by encrypting the shares with passwords. We develop a mechanism to select the most reliable delegates based on an effective trust measure. Specifically, relationships among the secret owner, delegate candidates and their related friends are used to estimate the trustworthiness of a delegate. This trust measure minimizes the likelihood of the secret being stolen by an adversary and is shown to be effective against various collusive attacks. Extensive simulations show that the proposed trust-based delegate selection performs very well in highly vulnerable environments where the adversary controls many nodes with different distributions and even with spreading of infections in the network. In fact, the number of keys lost is very low under extremely pessimistic assumptions of the adversary model.
Serge Vaudenay, Andrea Felice Caforio, Daniel Patrick Collins, Hailun Yan