Visual-Based Spam Filtering for Obfuscated Emails

This paper addresses the growing problem of spam emails that use visual obfuscation techniques to bypass traditional text-based spam filters. The proposed VBSF architecture offers a novel approach by mimicking human visual processing, rendering emails and analyzing both the extracted text and the visual appearance. The high accuracy reported (over 98%) suggests a significant improvement over existing methods in detecting these types of spam.

• •Addresses the problem of spam emails using visual obfuscation.
• •Proposes a novel visual-based spam detection architecture (VBSF).
• •Employs a multi-step process mimicking human visual processing.
• •Combines OCR, Naive Bayes, Decision Trees, and CNNs.
• •Achieves high accuracy (over 98%) on the designed dataset.