| CPC G06F 16/215 (2019.01) [G06F 16/285 (2019.01); G06F 16/9017 (2019.01); G06F 16/90344 (2019.01); G06F 16/9566 (2019.01); G06N 7/01 (2023.01); G06N 20/00 (2019.01)] | 20 Claims |
|
1. A computer system, comprising:
a processor; and
a non-transitory computer-readable medium having stored thereon instructions that are executable by the processor to cause the computer system to perform operations comprising:
analyzing uniform resource locator (URL) access data corresponding to a particular user of an electronically provided service, the URL access data comprising a plurality of URLs accessible by the particular user of the electronically provided service, the plurality of URLs each including a plurality of character strings that include one or more de-limiting characters;
tokenizing, for each URL of the plurality of URLs, each of the character strings for that URL into one or more respective sub-string tokens that are separated from one another by the one or more de-limiting characters;
using a token probability lookup table, cleaning the plurality of URLs to produce a plurality of cleaned URLs, wherein cleaning the plurality of URLs comprises, for each URL of the plurality of URLs, discarding sub-string tokens that do not meet a threshold probability level and retaining sub-string tokens that do meet the threshold probability level, wherein the token probability lookup table indicates a relative occurrence of sub-string tokens within the plurality of URLs;
generating a processed strings table that associates each of the plurality of character strings with a particular transaction or a particular web software function, wherein an electronic action is associated with the particular web software function;
receiving, from the particular user of the electronically provided service, a user request to perform the electronic action that is associated with the particular web software function via the electronically provided service, the user request including a first URL that comprises a plurality of URL segments;
processing the first URL at least in part by removing a subset of the URL segments that do not appear in the processed strings table;
inputting the plurality of cleaned URLs and the processed first URL into a machine learning model;
generating, based at least in part on the machine learning model and the processed strings table, a model classification output, wherein the model classification output categorizes the processed first URL as being associated with the particular transaction or with the particular web software function; and
based on the model classification output, approving or denying the user request to perform the electronic action that is associated with the particular web software function or requesting an additional authentication action from the particular user.
|