Benchmarking Differential Privacy and Federated Learning for BERT Models

Basu, Priyam; Roy, Tiasa Singha; Naidu, R. China Appala; Müftüoğlu, Zümrüt; Singh, Sahib; Mireshghallah, Fatemehsadat

doi:10.48550/arxiv.2106.13973

Cited by 12 publications

(14 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Hoory et al (2021) achieve accuracy which is comparable to the non-private model, but additionally supplement the public pre-training data with additional domain-relevant material, while we use off-the-shelf pre-trained models. Basu et al (2021) observe significant drops in utility, compared to our parameter-efficient methods which do not. While Kerrigan et al (2020) consider public pre-training and private fine-tuning, their experiments are on much smaller architectures (i.e., feedforward networks with three hidden layers).…”

Section: Related Workcontrasting

confidence: 53%

“…Anil et al (2021) privately train BERT-Large from scratch, compared to our work which focuses on private fine-tuning. (Hoory et al, 2021;Basu et al, 2021) perform private full fine-tuning of BERT models. Hoory et al (2021) achieve accuracy which is comparable to the non-private model, but additionally supplement the public pre-training data with additional domain-relevant material, while we use off-the-shelf pre-trained models.…”

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

Differentially Private Fine-tuning of Language Models

Yu¹,

Naik²,

Bačkurs³

et al. 2021

Preprint

View full text Add to dashboard Cite

We give simpler, sparser, and faster algorithms for differentially private fine-tuning of large-scale pre-trained language models, which achieve the state-of-the-art privacy versus utility tradeoffs on many standard NLP tasks. We propose a meta-framework for this problem, inspired by the recent success of highly parameter-efficient methods for fine-tuning. Our experiments show that differentially private adaptations of these approaches outperform previous private algorithms in three important dimensions: utility, privacy, and the computational and memory cost of private training. On many commonly studied datasets, the utility of private models approaches that of non-private models. For example, on the MNLI dataset we achieve an accuracy of 87.8% using RoBERTa-Large and 83.5% using RoBERTa-Base with a privacy budget of ε = 6.7. In comparison, absent privacy constraints, RoBERTa-Large achieves an accuracy of 90.2%. Our findings are similar for natural language generation tasks. Privately fine-tuning with DART, GPT-2-Small, GPT-2-Medium, GPT-2-Large, and GPT-2-XL achieve BLEU scores of 38.5, 42.0, 43.1, and 43.8 respectively (privacy budget of ε = 6.8, δ = 1e-5) whereas the non-private baseline is 48.1. All our experiments suggest that larger models are better suited for private fine-tuning: while they are well known to achieve superior accuracy non-privately, we find that they also better maintain their accuracy when privacy is introduced.

show abstract

Section: Related Workcontrasting

confidence: 53%

Section: Related Workmentioning

confidence: 99%

Differentially Private Fine-tuning of Language Models

Yu¹,

Naik²,

Bačkurs³

et al. 2021

Preprint

View full text Add to dashboard Cite

show abstract

“…As an example, consider the problem of training a differentially private model [Dwork et al 2006] for a healthcare task. Training such a model "end-to-end" (i.e., without leveraging any pretraining) to a decent privacy-utility tradeoff currently requires vast amounts of privacy-sensitive data [McMahan et al 2018;Basu et al 2021]. In contrast, a foundation model pretrained on public data could potentially be adapted to the specific healthcare task with significantly less confidential data [Bommasani et al 2019;Tramèr and Boneh 2021].…”

Section: Opportunitiesmentioning

confidence: 99%

On the Opportunities and Risks of Foundation Models

Bommasani¹,

Hudson²,

Adeli³

et al. 2021

Preprint

653

652

View full text Add to dashboard Cite

AI is undergoing a paradigm shift with the rise of models (e.g., BERT, DALL-E, GPT-3) that are trained on broad data at scale and are adaptable to a wide range of downstream tasks. We call these models foundation models to underscore their critically central yet incomplete character. This report provides a thorough account of the opportunities and risks of foundation models, ranging from their capabilities (e.g., language, vision, robotics, reasoning, human interaction) and technical principles (e.g., model architectures, training procedures, data, systems, security, evaluation, theory) to their applications (e.g., law, healthcare, education) and societal impact (e.g., inequity, misuse, economic and environmental impact, legal and ethical considerations). Though foundation models are based on standard deep learning and transfer learning, their scale results in new emergent capabilities, and their effectiveness across so many tasks incentivizes homogenization. Homogenization provides powerful leverage but demands caution, as the defects of the foundation model are inherited by all the adapted models downstream. Despite the impending widespread deployment of foundation models, we currently lack a clear understanding of how they work, when they fail, and what they are even capable of due to their emergent properties. To tackle these questions, we believe much of the critical research on foundation models will require deep interdisciplinary collaboration commensurate with their fundamentally sociotechnical nature.

show abstract

“…SEFL [19] is a FedNLP framework that achieves data privacy without any trusted entities. [10] studies how FedNLP can orchestrate with differential privacy. None of above work addresses the high training cost of FedNLP.…”

Section: Related Workmentioning

confidence: 99%

AutoFedNLP: An efficient FedNLP framework

Cai¹,

Wu²,

Wang³

et al. 2022

Preprint

View full text Add to dashboard Cite

Transformer-based pre-trained models have revolutionized NLP for superior performance and generality. Fine-tuning pre-trained models for downstream tasks often requires private data, for which federated learning is the de-facto approach (i.e., FedNLP). However, our measurements show that FedNLP is prohibitively slow due to the large model sizes and the resultant high network/computation cost. Towards practical FedNLP, we identify as the key building blocks adapters, small bottleneck modules inserted at a variety of model layers. A key challenge is to properly configure the depth and width of adapters, to which the training speed and efficiency is highly sensitive. No silver-bullet configuration exists: the optimal choice varies across downstream NLP tasks, desired model accuracy, and client resources. To automate adapter configuration, we propose AutoFedNLP, a framework that enhances the existing FedNLP with two novel designs. First, AutoFedNLP progressively upgrades the adapter configuration throughout a training session; the principle is to quickly learn shallow knowledge by only training fewer and smaller adapters at the model's top layers, and incrementally learn deep knowledge by incorporating deeper and larger adapters. Second, AutoFedNLP continuously profiles future adapter configurations by allocating participant devices to trial groups. To minimize client-side computations, AutoFedNLP exploits the fact that a FedNLP client trains on the same samples repeatedly between consecutive changes of adapter configurations, and caches computed activations on clients. Extensive experiments show that AutoFedNLP can reduce FedNLP's model convergence delay to no more than several hours, which is up to 155.5× faster compared to vanilla FedNLP and 48× faster compared to strong baselines.

show abstract

Benchmarking Differential Privacy and Federated Learning for BERT Models

Cited by 12 publications

References 0 publications

Differentially Private Fine-tuning of Language Models

Differentially Private Fine-tuning of Language Models

On the Opportunities and Risks of Foundation Models

AutoFedNLP: An efficient FedNLP framework

Contact Info

Product

Resources

About