Email Spam Detection via Visual Representation Comparison
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing email filters struggle to effectively detect spam messages that are not identical but have a similar 'look and feel', as they rely on one-to-one comparisons which are CPU and memory intensive and inadequate for larger email systems with varying message formats.
Innovation Solution
A method that processes email messages to create a visual representation, making them more fuzzy by averaging colors and computing a content identification code, which is then compared to a database of similar codes to classify messages as spam based on human perception, regardless of minor differences.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If one-to-one comparison of email message content is used, then detection accuracy for identical messages is improved, but CPU and memory usage becomes excessively high for large email systems
Solution Approach 1:
The patent segments the email comparison process into two stages: first comparing visual representations (low-cost, fast) to filter out obviously different messages, and then performing detailed content comparison only on messages with similar visual characteristics. This segmentation reduces the number of expensive one-to-one content comparisons needed while maintaining detection accuracy for spam variants.
Solution Approach 2:
The patent replaces the traditional mechanical content-comparison system with a visual representation-based system. Instead of directly comparing email text and content, the system converts emails into visual representations and compares those, which is computationally less intensive while still effective at detecting spam messages and their variants.
2Measurement precision
If exact content matching is required for spam detection, then precision in identifying identical spam is improved, but the system fails to detect spam messages with minor variations
Solution Approach 1:
The patent creates visual representations (copies) of email messages that capture their essential visual characteristics without requiring exact content matching. These visual copies can be compared efficiently to detect spam messages even when the original content has been slightly modified, thereby maintaining both precision and adaptability.
Solution Approach 2:
The patent changes the parameter being compared from exact text content to visual representation characteristics. This parameter change allows the system to detect spam messages based on their visual similarity rather than requiring identical text content, thus improving adaptability to spam variants while maintaining detection precision through the visual comparison mechanism.
3Adaptability or versatility
If visual representation processing is applied to all emails, then detection of similar spam messages is improved, but processing time increases for each message
Solution Approach 1:
The patent performs visual representation processing as a preliminary filtering step before detailed spam analysis. By quickly generating and comparing visual representations first, the system can eliminate obviously different messages without performing full spam detection analysis, thus reducing overall processing time while maintaining the ability to detect similar spam messages.
Solution Approach 2:
The patent applies visual representation processing selectively rather than uniformly to all emails. The system performs this processing primarily on messages that require spam detection analysis, rather than on every single email, thereby reducing the total processing time while still achieving effective detection of similar spam messages when needed.
Data Source
Figure 1
Figure 2
Figure 2
AI summary
A method of recognizing similarity of e-mail messages, by an e-mail processing device, wherein said method comprises the steps of receiving an e-mail message, processing a visual representation of said received e-mail message thereby resulting in a processed visual representation in which visual characteristics of said visual representation are highlighted, computing from said processed visual representation, a content identification code, wherein said content identification code is a measure for how a human would visually perceive said received e-mail message, and determining that said received e-mail message is similar to another e-mail message by finding a match between said computed content identification code and a plurality of content identification codes in a database.