BitAbuse: A Dataset of Visually Perturbed Texts for Defending Phishing Attacks

2025-02-06

AI Summary

This work addresses the challenge of defending against phishing attacks induced by visually perturbed text. We introduce the first real-world phishing-based dataset of visually perturbed text—comprising 325,000 perturbed sentences with original-text annotations—thereby filling a critical gap in authentic adversarial examples for this domain. Methodologically, we propose a hybrid data generation strategy that integrates real phishing corpora with controllable synthetic perturbations, enabling the training of an end-to-end sequence-to-sequence text restoration model. Experiments demonstrate that our model achieves 96% accuracy on perturbed-text recovery, substantially outperforming baselines trained solely on synthetic data. Crucially, we empirically reveal a significant performance gap between synthetic and real-world perturbations—a previously unquantified limitation—thereby establishing both a reliable empirical foundation and an effective technical pathway for building robust language models resilient to visual-textual phishing attacks.

Technology Category

Computer Vision: Adversarial Attacks & RobustnessMachine Learning: Adversarial Learning & RobustnessNatural Language Processing: Safety and Robustness

Application Category

Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for rankingWeb Mining and Content Analysis: Large pretrained models with web dataUser Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems

Abstract

Phishing often targets victims through visually perturbed texts to bypass security systems. The noise contained in these texts functions as an adversarial attack, designed to deceive language models and hinder their ability to accurately interpret the content. However, since it is difficult to obtain sufficient phishing cases, previous studies have used synthetic datasets that do not contain real-world cases. In this study, we propose the BitAbuse dataset, which includes real-world phishing cases, to address the limitations of previous research. Our dataset comprises a total of 325,580 visually perturbed texts. The dataset inputs are drawn from the raw corpus, consisting of visually perturbed sentences and sentences generated through an artificial perturbation process. Each input sentence is labeled with its corresponding ground truth, representing the restored, non-perturbed version. Language models trained on our proposed dataset demonstrated significantly better performance compared to previous methods, achieving an accuracy of approximately 96%. Our analysis revealed a significant gap between real-world and synthetic examples, underscoring the value of our dataset for building reliable pre-trained models for restoration tasks. We release the BitAbuse dataset, which includes real-world phishing cases annotated with visual perturbations, to support future research in adversarial attack defense.

Problem

• Defend phishing attacks with real-world data
• Improve language models' accuracy on perturbed texts
• Address limitations of synthetic datasets in phishing research

Innovation

• Real-world phishing case dataset
• Visual perturbation annotation
• Enhanced language model accuracy