Skip to content
DSML.KZ
Publications

Article · IEEE Access · December 10, 2025

A Web Protection Model Against Internationalized Domain Name Homograph Exploits

Sana Munir, Akmal Khan, Farkhanda Athar, Mansoor Qadir, Khalid Jamaan Alzahrani, Akbayan Bekarystankyzy

Akbayan Bekarystankyzy
Open via DOI

Abstract

Internationalized Domain Names (IDNs) enhance global accessibility by supporting Unicode characters in web addresses. However, this capability introduces new security vulnerabilities, particularly through homoglyphs—visually similar characters such as the Cyrillic ‘’ (U+0435) and the Latin ‘e’ (U+0065)—which enable attackers to craft deceptive domain names. These IDN homograph exploits are increasingly used in phishing attacks, creating URLs that are visually indistinguishable from legitimate sites. As the adoption of IDNs continues across browsers and domain registries, the risk of such exploits grows. According to the IDC 2023 Global DNS Threat Report, 90% of organizations faced domain-based attacks, with average losses of $1.1 million per incident. In this study, we propose a web protection model for the real-time detection and mitigation of IDN homograph threats. Central to our model is the construction of Unisimchar, a dynamic homoglyph database that extends prior work by incorporating Unicode block-level structures and performing pixel-level similarity analysis using 32×32 glyph images. Unlike static or manually maintained lists, Unisimchar enables automated and scalable identification of confusable character pairs through a threshold-based visual matching algorithm. The model processes over 1 million domain names sourced from Majestic Million, Zone Files, and Daily Domains, revealing that 80% of homoglyph pairs occur within high-traffic domains worldwide, with regional concentrations observed in Pakistan and Europe. With a multithreaded architecture, the model achieves domain-level detection within 0.07 seconds, making it suitable for real-time threat prevention. The comprehensive evaluation includes a human perception study to validate visual similarity thresholds and quantitative experiments using standard metrics such as precision, recall, and F1 score. The results demonstrate that our model significantly improves the accuracy of homoglyph detection compared to existing methods, offering a practical and extensible defense mechanism against modern IDN homograph exploits in the web ecosystem.

Spam and Phishing DetectionWeb Application Security VulnerabilitiesNetwork Security and Intrusion Detectioninternationalized domain namesphishingreal-time threat detection

More by Akbayan Bekarystankyzy