Poisoning Language Models During Instruction Tuning [PDF]
International Conference on Machine Learning, 2023 Instruction-tuned LMs such as ChatGPT, FLAN, and InstructGPT are finetuned on datasets that contain user-submitted examples, e.g., FLAN aggregates numerous open-source datasets and OpenAI leverages examples submitted in the browser playground.
Alexander Wan +3 more
semanticscholar +1 more source
Poisoning Web-Scale Training Datasets is Practical [PDF]
IEEE Symposium on Security and Privacy, 2023 Deep learning models are often trained on distributed, web-scale datasets crawled from the internet. In this paper, we introduce two new dataset poisoning attacks that intentionally introduce malicious examples to a model’s performance.
Nicholas Carlini +8 more
semanticscholar +1 more source
FLDetector: Defending Federated Learning Against Model Poisoning Attacks via Detecting Malicious Clients [PDF]
Knowledge Discovery and Data Mining, 2022 Federated learning (FL) is vulnerable to model poisoning attacks, in which malicious clients corrupt the global model via sending manipulated model updates to the server.
Zaixi Zhang +3 more
semanticscholar +1 more source
Nightshade: Prompt-Specific Poisoning Attacks on Text-to-Image Generative Models [PDF]
IEEE Symposium on Security and Privacy, 2023 Trained on billions of images, diffusion-based text-to-image models seem impervious to traditional data poisoning attacks, which typically require poison samples approaching 20% of the training set.
Shawn Shan +4 more
semanticscholar +1 more source
Back to the Drawing Board: A Critical Evaluation of Poisoning Attacks on Production Federated Learning [PDF]
IEEE Symposium on Security and Privacy, 2021 While recent works have indicated that federated learning (FL) may be vulnerable to poisoning attacks by compromised clients, their real impact on production FL systems is not fully understood.
Virat Shejwalkar +3 more
semanticscholar +1 more source
Wild Patterns Reloaded: A Survey of Machine Learning Security against Training Data Poisoning [PDF]
ACM Computing Surveys, 2022 The success of machine learning is fueled by the increasing availability of computing power and large training datasets. The training data is used to learn new models or update existing ones, assuming that it is sufficiently representative of the data ...
A. E. Cinà +9 more
semanticscholar +1 more source
Poisoning Attacks in Federated Learning: A Survey
IEEE Access, 2023 Federated learning faces many security and privacy issues. Among them, poisoning attacks can significantly impact global models, and malicious attackers can prevent global models from converging or even manipulating the prediction results of global ...
Geming Xia +3 more
semanticscholar +1 more source
Manipulating the Byzantine: Optimizing Model Poisoning Attacks and Defenses for Federated Learning
Network and Distributed System Security Symposium, 2021 —Federated learning (FL) enables many data owners (e.g., mobile devices) to train a joint ML model (e.g., a next-word prediction classifier) without the need of sharing their private training data.
Virat Shejwalkar, Amir Houmansadr
semanticscholar +1 more source
FedRecover: Recovering from Poisoning Attacks in Federated Learning using Historical Information [PDF]
IEEE Symposium on Security and Privacy, 2022 Federated learning is vulnerable to poisoning attacks in which malicious clients poison the global model via sending malicious model updates to the server.
Xiaoyu Cao +3 more
semanticscholar +1 more source
Poisoning with Cerberus: Stealthy and Colluded Backdoor Attack against Federated Learning
AAAI Conference on Artificial Intelligence, 2023 Are Federated Learning (FL) systems free from backdoor poisoning with the arsenal of various defense strategies deployed? This is an intriguing problem with significant practical implications regarding the utility of FL services.
Xiaoting Lyu +6 more
semanticscholar +1 more source