System, method, and logic for classifying communications

In accordance with particular embodiments, a method includes intercepting a communication and extracting metadata associated with the communication. The extracted metadata comprises a plurality of different fields from communication metadata and file metadata. The method further includes determining a score, based on previous communications, for each field of the extracted metadata. The score is indicative of a likelihood that the communication is a malicious communication. The method additionally includes combining the scores to generate a combined score for the communication based on an algorithm developed from the previous communications. The method also includes generating, based on the combined score at a first time, a predicted classification as to whether the communication is a malicious communication. The method further includes receiving, at a second time subsequent to the first time, an indication of whether the communication is a malicious communication and updating the algorithm based on the indication.