EVALUATING ELECTRONIC MAIL MESSAGES BASED ON PROBABILISTIC ANALYSIS

A computer system can evaluate electronic messages based on probabilistic analysis, including Bayesian analysis. In one embodiment, a data processing system comprises logic configured for perform training a probabilistic filter using first properties of one or more first network resource identifiers obtained from a whitelist; training the probabilistic filter using second properties of one or more second network resource identifiers obtained from a blocklist; testing third properties of a third network resource identifier using the probabilistic filter, resulting in creating a probability output; and adding the third network resource identifier to the blocklist when the probability output is greater than a specified threshold. The blocklist can be communicated to messaging gateways in the field for local use in evaluating messages that contain spam or other threats.