Phishing URL Detection using Bi-LSTM

TL;DR

The paper addresses phishing URL detection by applying a Bi-LSTM model to character-level URL sequences, classifying URLs into four categories: benign, phishing, defacement, and malware. It leverages a Kaggle dataset of over 650{,}000 URLs and demonstrates a lightweight deployment pipeline with embedding, Bi-LSTM, dropout, and a softmax output, achieving about $0.98$ accuracy on the test set. A Flask backend paired with a UI enables real-time predictions and confidence scores, illustrating practical applicability in real-world settings. The work highlights strong performance while acknowledging areas for improvement in phishing recall and outlines future directions to enrich features and broaden coverage of malicious URL types.

Abstract

Phishing attacks threaten online users, often leading to data breaches, financial losses, and identity theft. Traditional phishing detection systems struggle with high false positive rates and are usually limited by the types of attacks they can identify. This paper proposes a deep learning-based approach using a Bidirectional Long Short-Term Memory (Bi-LSTM) network to classify URLs into four categories: benign, phishing, defacement, and malware. The model leverages sequential URL data and captures contextual information, improving the accuracy of phishing detection. Experimental results on a dataset comprising over 650,000 URLs demonstrate the model's effectiveness, achieving 97% accuracy and significant improvements over traditional techniques.

Figures (2)

• Figure 1: Phishing attack process
• Figure 2: Confusion Matrix of the Classification Model