Trust is a widespread mechanism to establish and conduct successful social collaborations, expecially in today's Internet, where people perform more and more activities on-line, from social networking to e-learning, e-commerce and others. Several algorithms are available for trust assessment, such as the well-known EigenTrust to PowerTrust, GossipTrust, TrustWebRank to cite some. An evaluation of such algorithms must take into account several factors, e.g. performances, scalability and robustness against malicious attacks, which plays a key role in ensuring algorithm's effectiveness. In this paper we study the effects of malicious peers on the EigenTrust algorithm, focusing on the (mis)behaviour of its pre-trusted peers, using a proper simulator to conduct several experiments under different conditions. Results show that EigenTrust effectiveness may be significantly affected by inactive and/or malicious pre-trusted peers