Related Experiment Video
Updated: Jan 14, 2026
09:09
Foreign Accent and Forensic Speaker Identification in Voice Lineups: The Influence of Acoustic Features Based on Prosody
Published on: September 27, 2024
807
Phishing detection on webpages in European non-English languages based on machine learning.
1Department of Telecommunications, Brno University of Technology, Brno, Czech Republic. komosny@vut.cz.
Scientific Reports
|October 28, 2025
Summary
This study enhances machine learning phishing detection for minor European languages, achieving 99% accuracy and significantly reducing false positives on local webpages. The new method boosts cybersecurity for underserved language communities.
Area of Science:
- Cybersecurity
- Machine Learning
- Natural Language Processing
Background:
- Current machine learning phishing detection is effective for English but lacks accuracy for minor languages.
- Zero-day phishing attacks pose a significant threat, necessitating robust detection methods.
Purpose of the Study:
- To improve phishing detection accuracy for webpages in minor European languages.
- To reduce the false positive rate in detecting phishing attempts on local language websites.
Main Methods:
- Development of a novel language-based phishing detection model.
- Testing the model on approximately two million local webpages from 16 European countries.
- Statistical validation using Shapiro-Wilk and Paired T-tests to ensure robustness.
Main Results:
- Achieved 99% accuracy for phishing detection on local webpages in 12 European countries.
- Reduced the false positive rate by up to a factor of 10 for minor language webpages.
- Demonstrated statistically significant and robust performance across diverse webpage sets.
Conclusions:
- The proposed language-based phishing detection significantly outperforms existing methods for minor languages.
- This advancement enhances global cybersecurity by providing more inclusive protection.
- Publicly releasing code and data promotes reproducibility and further research.
