A Comprehensive Analysis of Adversarial Attacks against Spam Filters

TL;DR

This work analyzes the vulnerability of six deep learning–based spam filters to adversarial text attacks across word, character, sentence, and AI-generated paragraph levels using three real-world datasets. It introduces two novel scoring functions, spam weights and attention weights, to efficiently identify impactful text segments for perturbation under a black-box setting. AI-generated paragraph-level attacks reveal DistilBERT's robustness and expose weaknesses in other architectures, emphasizing the need for defenses in adversarial environments. Overall, the study provides a comprehensive framework for evaluating and improving the security of spam filters against evolving NLP-based threats.

Abstract

Deep learning has revolutionized email filtering, which is critical to protect users from cyber threats such as spam, malware, and phishing. However, the increasing sophistication of adversarial attacks poses a significant challenge to the effectiveness of these filters. This study investigates the impact of adversarial attacks on deep learning-based spam detection systems using real-world datasets. Six prominent deep learning models are evaluated on these datasets, analyzing attacks at the word, character sentence, and AI-generated paragraph-levels. Novel scoring functions, including spam weights and attention weights, are introduced to improve attack effectiveness. This comprehensive analysis sheds light on the vulnerabilities of spam filters and contributes to efforts to improve their security against evolving adversarial threats.

Figures (2)

• Figure 1: Word Deletion Attack Results on the Dense Filter
• Figure 2: Character Attack Results on the Dense Filter