| CPC G06Q 10/107 (2013.01) [G06F 40/205 (2020.01); G06N 7/01 (2023.01); G06Q 30/018 (2013.01); G06Q 40/02 (2013.01); H04L 63/1416 (2013.01); H04L 63/1425 (2013.01)] | 18 Claims |

|
1. A method, comprising:
receiving an email that is directed to an email recipient and is from an email sender;
extracting, from the email, a triple, the extracted triple including a display name of the email sender, an email address of the email sender, and an email domain of the email sender;
identifying a triple identifier associated with the extracted triple;
determining a name score for the email based on the triple identifier by accessing the name score for the triple identifier in a database, wherein the name score indicates a probability that the email is from a legitimate sender, and wherein the name score is updated based on frequency counting;
inputting message content of the email into a plurality of classifiers, wherein each of the plurality of classifiers is machine-learned to recognize a particular message type;
determining a message type of the email based on output of the plurality of classifiers; and
determining, based on at least the determined name score for the email, the determined message type of the email, and a reputation score associated with an IP address of the email sender, whether the email is associated with a business email compromise (BEC) attack.
|