Papers with data poisoning
From Shortcuts to Triggers: Backdoor Defense with Denoised PoE (2024.naacl-long)
Copied to clipboard
| Challenge: | Existing backdoor defense methods focus on specific triggers, leaving a universal defense unexplored. |
| Approach: | They propose an ensemble-based backdoor defense framework that denies backdoor attacks by capturing backdoor shortcuts and preventing learning them. |
| Outcome: | The proposed framework significantly improves defense performance against backdoor attacks . it is also effective under a more challenging but practical setting . |
Instructions as Backdoors: Backdoor Vulnerabilities of Instruction Tuning for Large Language Models (2024.naacl-long)
Copied to clipboard
| Challenge: | et al., 2021) show that instruction models can be trained on crowdsourced datasets with task instructions to achieve superior performance. |
| Approach: | They examine security concerns of emergent instruction tuning paradigm that models are trained on crowdsourced datasets with task instructions to achieve superior performance. |
| Outcome: | The proposed model can achieve 90% success rate across four commonly used datasets. |
Backdoor NLP Models via AI-Generated Text (2024.lrec-main)
Copied to clipboard
| Challenge: | Existing attacks disregard fluency and semantic fidelity of poisoned text, rendering it easily detectable. |
| Approach: | They propose to use AI-generated poisoned text to attack NLP models by establishing covert associations between trigger patterns and target labels without affecting normal accuracy. |
| Outcome: | The proposed method achieves effective attacks while maintaining fluency and semantic similarity across all scenarios. |
Pre-training CLIP against Data Poisoning with Optimal Transport-based Matching and Alignment (2025.emnlp-main)
Copied to clipboard
| Challenge: | Recent studies have shown that Contrastive Language-Image Pre-training (CLIP) models are vulnerable to data poisoning and backdoor attacks due to massive training image-caption pairs crawled from the Internet. |
| Approach: | They propose an Optimal Transport-based framework to reconstruct image-caption pairs and propose an optimal transport-based distance measure to re-assign new captions based on the proposed optimal transport distance. |
| Outcome: | The proposed framework reduces the attack success rates of poisoning attacks to 0% in most cases. |
Attacks against Abstractive Text Summarization Models through Lead Bias and Influence Functions (2024.findings-emnlp)
Copied to clipboard
| Challenge: | Large Language Models (LLMs) are vulnerable to adversarial perturbations and data poisoning attacks, especially in tasks like text classification and translation. |
| Approach: | They propose a novel approach exploiting the inherent lead bias in large language models to perform adversarial perturbations and an innovative application of influence functions to execute data poisoning attacks. |
| Outcome: | The proposed approach shows that the models under attack tend to generate extractive summaries rather than abstractive summarizations. |