Over the years, many anti-phishing software packages have been developed that can reliably and accurately detect and delete phishing emails as they are received. As communication on the internet evolves, however, these existing anti-phishing systems are becoming less effective. As more users migrate away from email and into emerging technologies such as Slack, Zoom, and Microsoft Teams, new effective anti-phishing filters must be created for each new communication platform. Developers are therefore fighting an uphill battle to keep users safe. An anti-phishing mechanism that positions itself instead directly between the user and the websites they visit is therefore proposed. This positioning allows the system to protect the user against phishing attacks no matter the communication medium. Existing research in this area suffers from impractical processing overhead, secure logic failures, and unreliability in the long term. This thesis overcomes these issues by using corporate branding color as a visual similarity measurement within a supervised learning algorithm to perform phishing identification. Since it has been shown that corporate branding colors change much less often than other design choices like HTML layout, this visual similarity comparison is able to maintain high accuracy over long periods of time. This principle, combined with a fast machine learning algorithm, allows the application to be accurate, effective, and adaptable with little to no added overhead, overcoming the shortcomings in currently proposed solutions.

Library of Congress Subject Headings

Phishing--Prevention; Industrial design coordination; Trademarks; Corporate image; Branding (Marketing); Color in marketing; Colorimetry; Machine learning

Publication Date


Document Type


Student Type


Degree Name

Computer Science (MS)

Department, Program, or Center

Computer Science (GCCIS)


Rajendra Raj

Advisor/Committee Member

Carol Romanowski

Advisor/Committee Member

Reynold Bailey


RIT – Main Campus

Plan Codes