US 12,265,637 B2
Identification of sensitive content in electronic mail messages to prevent exfiltration
Austin Grant Walters, Savoy, IL (US); Jeremy Edward Goodsitt, Champaign, IL (US); and Anh Truong, Champaign, IL (US)
Assigned to Capital One Services, LLC, McLean, VA (US)
Filed by Capital One Services, LLC, McLean, VA (US)
Filed on Dec. 28, 2021, as Appl. No. 17/563,337.
Prior Publication US 2023/0205906 A1, Jun. 29, 2023
Int. Cl. G06F 21/62 (2013.01); H04L 9/40 (2022.01)
CPC G06F 21/6218 (2013.01) [H04L 63/0245 (2013.01)] 19 Claims
OG exemplary drawing
 
1. A method performed by one or more computing resources, comprising:
with the one or more computing resources, scanning an outbound electronic mail message for patterns associated with sensitive content;
further processing the outbound electronic mail message with a neural network model running on the one or more computing resources and trained to yield a probability that the outbound electronic mail message contains sensitive content based on contextual information relating to the outbound electronic mail message, wherein:
the neural network model is trained, at least in part, using a training set comprising at least one training electronic mail message consisting of non-sensitive text of at least one sensitive mail message that exfiltrated at least one item of sensitive content, the non-sensitive text comprising text of the at least one sensitive mail message with an entirety of the at least one item of sensitive content removed to determine whether the non-sensitive text contains patterns indicative of containing sensitive content; and
the training set includes a specified number of electronic mail messages that were sent immediately before an exfiltration event; and
taking, responsive to the scanning locating one or more patterns associated with sensitive content or the probability that the outbound electronic mail message contains sensitive content exceeding a first threshold, a first remedial action regarding the outbound electronic mail message.