A Semantic, Syntactic, And Context-Aware Natural Language Adversarial Example Generator
Published in IEEE Transactions on Dependable and Secure Computing, 2024
The paper presents SSCAE, a novel method for crafting high-quality adversarial examples (AEs) in natural language processing (NLP). SSCAE utilizes a masked language model to identify key words and generate substitutions, which are then evaluated by two language models for semantic and syntactic accuracy. Incorporating dynamic thresholds and local greedy search, SSCAE efficiently generates imperceptible AEs that maintain semantic and syntactic consistency.
Recommended citation: Javad Rafiei Asl, Mohammad H. Rafiei, Manar Alohaly, and Daniel Takabi. "A Semantic, Syntactic, And Context-Aware Natural Language Adversarial Example Generator." IEEE Transactions on Dependable and Secure Computing (2024). https://ieeexplore.ieee.org/abstract/document/10416371