Zobrazeno 1 - 10
of 17
pro vyhledávání: '"Panda, Ashwinee"'
Autor:
Panda, Ashwinee, Isik, Berivan, Qi, Xiangyu, Koyejo, Sanmi, Weissman, Tsachy, Mittal, Prateek
Existing methods for adapting large language models (LLMs) to new tasks are not suited to multi-task adaptation because they modify all the model weights -- causing destructive interference between tasks. The resulting effects, such as catastrophic f
Externí odkaz:
http://arxiv.org/abs/2406.16797
Autor:
Qi, Xiangyu, Panda, Ashwinee, Lyu, Kaifeng, Ma, Xiao, Roy, Subhrajit, Beirami, Ahmad, Mittal, Prateek, Henderson, Peter
The safety alignment of current Large Language Models (LLMs) is vulnerable. Relatively simple attacks, or even benign fine-tuning, can jailbreak aligned models. We argue that many of these vulnerabilities are related to a shared underlying issue: saf
Externí odkaz:
http://arxiv.org/abs/2406.05946
Autor:
Panda, Ashwinee, Choquette-Choo, Christopher A., Zhang, Zhengming, Yang, Yaoqing, Mittal, Prateek
When large language models are trained on private data, it can be a significant privacy risk for them to memorize and regurgitate sensitive information. In this work, we propose a new practical data extraction attack that we call "neural phishing". T
Externí odkaz:
http://arxiv.org/abs/2403.00871
Differentially private stochastic gradient descent (DP-SGD) allows models to be trained in a privacy-preserving manner, but has proven difficult to scale to the era of foundation models. We introduce DP-ZO, a private fine-tuning framework for large l
Externí odkaz:
http://arxiv.org/abs/2401.04343
Autor:
Qi, Xiangyu, Huang, Kaixuan, Panda, Ashwinee, Henderson, Peter, Wang, Mengdi, Mittal, Prateek
Recently, there has been a surge of interest in integrating vision into Large Language Models (LLMs), exemplified by Visual Language Models (VLMs) such as Flamingo and GPT-4. This paper sheds light on the security and safety implications of this tren
Externí odkaz:
http://arxiv.org/abs/2306.13213
In privacy-preserving machine learning, differentially private stochastic gradient descent (DP-SGD) performs worse than SGD due to per-sample gradient clipping and noise addition. A recent focus in private learning research is improving the performan
Externí odkaz:
http://arxiv.org/abs/2306.06076
In-context learning (ICL) is an important capability of Large Language Models (LLMs), enabling these models to dynamically adapt based on specific, in-context exemplars, thereby improving accuracy and relevance. However, LLM's responses may leak the
Externí odkaz:
http://arxiv.org/abs/2305.01639
An open problem in differentially private deep learning is hyperparameter optimization (HPO). DP-SGD introduces new hyperparameters and complicates existing ones, forcing researchers to painstakingly tune hyperparameters with hundreds of trials, whic
Externí odkaz:
http://arxiv.org/abs/2212.04486
Autor:
Zhang, Zhengming, Panda, Ashwinee, Song, Linyue, Yang, Yaoqing, Mahoney, Michael W., Gonzalez, Joseph E., Ramchandran, Kannan, Mittal, Prateek
Due to their decentralized nature, federated learning (FL) systems have an inherent vulnerability during their training to adversarial backdoor attacks. In this type of attack, the goal of the attacker is to use poisoned updates to implant so-called
Externí odkaz:
http://arxiv.org/abs/2206.10341
Autor:
Panda, Ashwinee, Mahloujifar, Saeed, Bhagoji, Arjun N., Chakraborty, Supriyo, Mittal, Prateek
Federated learning is inherently vulnerable to model poisoning attacks because its decentralized nature allows attackers to participate with compromised devices. In model poisoning attacks, the attacker reduces the model's performance on targeted sub
Externí odkaz:
http://arxiv.org/abs/2112.06274