AdaptiveWordBug: Generating adversarial texts with an adaptive scoring strategy against deep learning classifiers.

Deep learning models demonstrate vulnerability to textual adversarial attacks. Research on adversarial text generation methods contributes to the subsequent design of corresponding countermeasures. Current word-level adversarial text generation metho…

Preview only. Read the full abstract at the source