Identity theft through phishing attacks has become a major concern for Internet users. Typically, phishing attacks aim at luring the user to a faked web site to disclose personal information. Existing solutions proposed against this kind of attack can, however, hardly counter the new generation of sophisticated malware phishing attacks, e.g., pharming Trojans, designed to target certain services. This paper aims at making the first steps towards the design and implementation of a security architecture that prevents both classical and malware phishing attacks. Our approach is based on the ideas of compartmentalization for isolating applications of different trust level, and a trusted wallet for storing credentials and authenticating sensitive services. Once the wallet has been setup in an initial step, our solution requires no special care from users for identifying the right web sites while the disclosure of credentials is strictly controlled. Moreover, a prototype of the basic platform exists and we briefly describe its implementation.