In ambient environments, new security challenges that are not adequately addressed by existing security models appear. In such context, intelligent communication devices participate to spontaneous and self-organized networks where unexpected interactions with unknown devices take place. Without centralized organization, security turns in a risk management problem. In this paper we propose and analyze a computational model of trust that captures trust dynamics of the human society. In our model, past experiences and recommendations are aggregated in the notion of history of past interactions which are protected by cryptographic material. To avoid the trust dissemination, each entity is viewed as an autonomous device and a trust level is computed based only upon selfish evaluation of common trustworthy nodes. Our proposal reduces the complexity of the decision-making process by providing proved data that can be the foundation of the final decision. The proposed trust model is described together with an overview of the cryptographic protocol and its security analysis. The trust function is analyzed through intensive simulations depending on the impact of the chosen parameters of the trust evaluation and on the dynamics of the studied groups.