Granted Patent B2

US 12,413,621 · App. 18/649,734 · Granted Sep 9, 2025

Inventor: Shashi KIran N (Bangalore, IN)

Assignee: McAfee, LLC

H04L63/1483G06F16/955G06N20/00H04L63/1416H04L63/1433H04L63/166

Abstract

Claims (34)

1. A computer-implemented method of protecting a user from phishing attacks, comprising:

detecting that the user has opened a website associated with a uniform resource locator (URL), and determining that the URL does not have a known good reputation, comprising querying a cloud-based reputation store for the URL;

in response to the determining, rendering a test version of the website in a headless web browser, including abstracting the website into at least one visual element of the test version;

visually inspecting the test version of the website with a digital eye, the digital eye comprising a machine learning model visually trained on known legitimate websites, and determining that the test version of the web looks like a known legitimate website not associated with the URL; and

based on the visual inspection, warning the user that the website is or may be a phishing website.

2. The method of claim 1 , further comprising querying a cloud reputation service to determine that the URL does not have a known good reputation.

3. The method of claim 1 , further comprising querying a local reputation cache to determine that the URL does not have a known good reputation.

4. The method of claim 1 , further comprising determining that the user has not previously visited the URL to determine that the URL does not have a known good reputation.

5. The method of claim 1 , wherein abstracting the at least one visual element includes a corporate logo.

6. The method of claim 1 , wherein abstracting the at least one visual element includes a font.

7. The method of claim 1 , wherein abstracting the at least one visual element includes a background color.

8. The method of claim 1 , wherein abstracting the at least one visual element includes a background pattern.

9. The method of claim 1 , wherein abstracting the at least one visual element includes a color scheme.

10. The method of claim 1 , wherein determining that the test version of the website looks like the known legitimate website not associated with the URL comprises determining that the URL is a non-secure hypertext transport protocol (HTTP) URL.

11. The method of claim 1 , wherein determining that the test version of the website looks like the known legitimate website not associated with the URL comprises determining that the URL uses transport layer security (TLS), but that a certificate of the URL is not associated with the known legitimate website.

12. The method of claim 1 , wherein determining that the test version of the website looks like the known legitimate website not associated with the URL comprises determining that the URL belongs to a blogs or personal webpages subdomain of a known third-party service not associated with the known legitimate website.

13. One or more tangible, nontransitory computer-readable storage media having stored thereon executable instructions to:

detect that a user has opened a website associated with a uniform resource locator (URL), and determine that the URL does not have a known good reputation, comprising querying a cloud-based reputation store for the URL;

in response to the determining, render a test version of the website in a headless web browser, including abstracting the website into at least one visual element of the test version;

visually inspect the test version of the website with a digital eye, the digital eye comprising a machine learning model visually trained on known legitimate websites, and determine that the test version of the web looks like a known legitimate website not associated with the URL; and

based on the visual inspection, warn the user that the website is or may be a phishing website.

14. The one or more tangible, nontransitory computer-readable storage media of claim 13 , wherein the instructions are further to query a cloud reputation service to determine that the URL does not have a known good reputation.

15. The one or more tangible, nontransitory computer-readable storage media of claim 13 , wherein the instructions are further to query a local reputation cache to determine that the URL does not have a known good reputation.

16. The one or more tangible, nontransitory computer-readable storage media of claim 13 , wherein the instructions are further to determine that the user has not previously visited the URL to determine that the URL does not have a known good reputation.

17. A computing apparatus, comprising:

a hardware platform comprising a processor circuit and a memory; and

instructions encoded within the memory to instruct the processor circuit to:

detect that a user has opened a website associated with a uniform resource locator (URL), and determine that the URL does not have a known good reputation, comprising querying a cloud-based reputation store for the URL;

in response to the determining, render a test version of the website in a headless web browser, including abstracting the website into at least one visual element of the test version;

visually inspect the test version of the website with a digital eye, the digital eye comprising a machine learning model visually trained on known legitimate websites, and determine that the test version of the web looks like a known legitimate website not associated with the URL; and

based on the visual inspection, warn the user that the website is or may be a phishing website.

18. The computing apparatus of claim 17 , wherein the instructions are further to query a cloud reputation service to determine that the URL does not have a known good reputation.

19. The computing apparatus of claim 17 , wherein the instructions are further to query a local reputation cache to determine that the URL does not have a known good reputation.

20. The computing apparatus of claim 17 , wherein the computing apparatus is a gateway.

Continuity (3)

References Cited (48)