Homoglyph Detection Using Visual Similarity After Deobfuscation

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing computing systems struggle to detect homoglyphs, which are visually similar or identical characters with different Unicode code points, often used for malicious purposes such as phishing attacks and bypassing spam filters.

Innovation Solution

A method and system for detecting homoglyphs by deobfuscating a string of characters, rendering both the original and deobfuscated strings into images, calculating visual similarity metrics like SSIM, convolutional similarity, or OCR distance, and determining the presence of homoglyphs based on threshold values.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If traditional character comparison methods are used to detect homoglyphs, then the detection process is simple and fast, but the detection accuracy is low because homoglyphs have different Unicode code points but appear visually identical

Engineering Contradiction:

Improvehomoglyph detection accuracyVSAvoiddetection system complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The patent replaces traditional character-based comparison methods with image-based visual comparison. Characters are rendered as images and compared using image processing techniques (pixel comparison, structural similarity analysis) instead of Unicode code point comparison, enabling detection of visually similar characters that traditional methods miss

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The patent introduces an intermediary rendering step that converts characters into visual images before comparison. This intermediary representation allows the system to compare the actual visual appearance of characters rather than their abstract code point values, bridging the gap between digital representation and human perception

Inventive Principle:

Principle #24Intermediary (Mediator)

2Measurement precision

If visual rendering and image comparison methods are used to detect homoglyphs, then the detection accuracy improves, but the processing time and computational resources increase

Engineering Contradiction:

Improvehomoglyph detection accuracyVSAvoiddetection processing time

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The patent segments the character comparison process into distinct stages: rendering characters as images, comparing pixel values, and analyzing structural similarity. This segmentation allows for optimized processing at each stage and enables selective application of more computationally intensive methods only when necessary

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent employs a tiered comparison approach where simpler pixel-based comparison is used first for quick rejection of obviously different characters, and more computationally intensive structural similarity analysis is applied only when initial comparison yields inconclusive results, reducing overall processing time

Inventive Principle:

Principle #16Partial or excessive action

Data Source

PatentUS12586398B2Detecting a homoglyph in a string of characters

Publication Date: 2026.03.24 GOOGLE LLC

AI summary

A method for detecting a homoglyph in an input text includes receiving a first string of characters and deobfuscating the first string of characters to generate a second string of characters. When at least one character from the first string of characters has replaced another character in the second string of characters based on the deobfuscating, the method further includes determining a visual similarity value based on the first string of characters and the second string of characters and providing an indication whether a homoglyph is present in the first string of characters, based on the visual similarity value.