This article concerns a security system which enables the detection of undesirable agents, intruders in a multi agent system. The intruders are identified on the basis of their behavior which is determined by actions they undertook. The process of behavior evaluation is distributed each agent makes autonomous behavior evaluation of other agents. In order to distinguish if an agent is good or bad (an intruder), results of behavior evaluations of this agent have to be collected and processed. The problem of collecting and processing of the results of distributed behavior evaluation is the main topic of this article. The conception of storing agents' behavior evaluation results obtained during earlier periods of time is presented. The decision about an agent is undertaken on the basis of the analysis of the gathered results of his behavior evaluation.