The invention provides a
phishing mail dynamic defense decision-making method and
system based on
reinforcement learning, and relates to the technical field of
network security, and the method comprises the steps: obtaining historical
phishing mail data, carrying out the
feature extraction, generating multi-dimensional features, constructing a high-quality
phishing mail
feature data set, constructing a phishing mail dynamic defense decision-making model, and carrying out the training, packaging the trained phishing mail dynamic defense
decision model into a standardized interface for integration, obtaining a newly received mail, extracting initial mail features, carrying out initial judgment, carrying out deep
feature extraction, carrying out secondary discrimination, generating an accurate identification result, and if the mail is a phishing mail, generating a phishing mail
attack portrait and an
active defense strategy; and outputting a defense action instruction, issuing the defense action instruction to a defense execution node, obtaining
user feedback and behavior data in real time, evaluating a global phishing mail
attack situation and a defense effect, and combining and analyzing feedback opinions to update the phishing mail dynamic defense
decision model.