The purpose of trust and reputation systems is to strengthen the quality of markets and communities by providing an incentive for good behaviour and quality services, and by sanctioning bad behaviour and low quality services. However, trust and reputation systems will only be able to produce this effect when they are sufficiently robust against strategic manipulation or direct attacks. Currently, robustness analysis of TRSs is mostly done through simple simulated scenarios implemented by the TRS designers themselves, and this can not be considered as reliable evidence for how these systems would perform in a realistic environment. In order to set robustness requirements it is important to know how important robustness really is in a particular community or market. This paper discusses research challenges for trust and reputation systems, and proposes a research agenda for developing sound and reliable robustness principles and mechanisms for trust and reputation systems.