Scholarlink

The GoPhish Chrome extension aims to increase web security by providing users with an easy-to-use tool for identifying malicious URLs. With a focus on simplicity, the extension lets users start URL scans with a context menu interaction. This sets off a machine learning system that analyzes the input URL and produces a confidence score that indicates how malignant it is. The main functionality is underpinned by a structured manifest file outlining required rights and JavaScript components handling data processing and user interaction. For effective client-side execution, the machine learning model, which was initially created in Keras, is transformed into TensorFlow.js format.execution. After thorough evaluation and comprehensive testing, it shows an accuracy rate of 74.626% when classifying real-world URLs, this is below the targeted standard of 95% for dependable security applications. This research emphasizes the necessity of additional model optimization to improve its prediction efficacy in real-world situations, demonstrating the continuous challenges associated with implementing machine learning solutions in browser extensions for cybersecurity.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Implementing Deep Learning to Detect Malicious URLs

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Rhodri Thomas,
  • Sabeen Tahir,
  • Sheikh Tahir Bakhsh,
  • Reem Alotaibi

摘要

The GoPhish Chrome extension aims to increase web security by providing users with an easy-to-use tool for identifying malicious URLs. With a focus on simplicity, the extension lets users start URL scans with a context menu interaction. This sets off a machine learning system that analyzes the input URL and produces a confidence score that indicates how malignant it is. The main functionality is underpinned by a structured manifest file outlining required rights and JavaScript components handling data processing and user interaction. For effective client-side execution, the machine learning model, which was initially created in Keras, is transformed into TensorFlow.js format.execution. After thorough evaluation and comprehensive testing, it shows an accuracy rate of 74.626% when classifying real-world URLs, this is below the targeted standard of 95% for dependable security applications. This research emphasizes the necessity of additional model optimization to improve its prediction efficacy in real-world situations, demonstrating the continuous challenges associated with implementing machine learning solutions in browser extensions for cybersecurity.