From Tweaks to Turmoil: Attacks against Text Summarization Models through Lead Bias and Influence Functions

Anonymous

Abstract: Large Language Models (LLMs) have introduced novel opportunities for text comprehension and generation. Yet, they are vulnerable to adversarial perturbations and data poisoning attacks, particularly in tasks like text classification and translation, as evidenced by numerous studies. However, the adversarial robustness of Text Summarization models remains less explored. In this work, we unveil a novel approach by exploiting the inherent lead bias in summarization models, to perform adversarial perturbations. Furthermore, we introduce an innovative application of influence functions, to execute data poisoning, which compromises models’ integrity. This approach not only shows a skew in the model’s behavior to produce desired outcomes, but also shows a new behavioral change, where models under attack tend to generate extractive summaries rather than abstractive summaries

Paper Type: long

Research Area: Summarization

Contribution Types: Model analysis & interpretability, NLP engineering experiment, Publicly available software and/or pre-trained models

Languages Studied: English