Improving On-Device Speaker Verification Using Federated Learning with Privacy

Granqvist, Filip; Seigel, Matt; Dalen, Rogier C. van; Cahill, Áine; Shum, Stephen; Paulik, Matthias

doi:10.21437/interspeech.2020-2944

Cited by 40 publications

(14 citation statements)

References 23 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In local DP [Dwork et al, 2014], the statistics are obfuscated before leaving the device. However, models trained with local DP often suffer from low utility [Granqvist et al, 2020]. Instead, this paper will assume ( , δ)-central DP.…”

Section: Private Federated Learningmentioning

confidence: 99%

“…Finally, the resulting algorithm, termed Fair PFL or FPFL and described in Algorithm 1, inspired by the ideas from [McMahan et al, 2018b;Truex et al, 2019;Granqvist et al, 2020;Bonawitz et al, 2017] guarantees the users' privacy as follows:…”

Section: Extending the Algorithm To Private Federated Learningmentioning

confidence: 99%

“…Therefore, in order to provide guarantees of privacy to users, it is customary to ensure that the training of the model is differentially private, see e.g. [Bhowmick et al, 2018;McMahan et al, 2018a,b;Truex et al, 2019;Granqvist et al, 2020]. Differential privacy (DP) [Dwork et al, 2006[Dwork et al, , 2014] is a strong privacy standard for algorithms that operate on data.…”

Section: Introductionmentioning

confidence: 99%

“…For this reason, they are usually the model of choice for FL . In fact, DL models have been successfully trained with differentially private FL (or PFL) for tasks like next word prediction or speaker verification Granqvist et al, 2020]. Nonetheless, these models are susceptible to perpetuate societal biases existing in the data [Caliskan et al, 2017] or to discriminate against certain groups even when the data is balanced [Buolamwini and Gebru, 2018].…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Enforcing fairness in private federated learning via the modified method of differential multipliers

Rodríguez-Gálvez¹,

Granqvist²,

Rogier³

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

Federated learning with differential privacy, or private federated learning, provides a strategy to train machine learning models while respecting users' privacy. However, differential privacy can disproportionately degrade the performance of the models on under-represented groups, as these parts of the distribution are difficult to learn in the presence of noise. Existing approaches for enforcing fairness in machine learning models have considered the centralized setting, in which the algorithm has access to the users' data. This paper introduces an algorithm to enforce group fairness in private federated learning, where users' data does not leave their devices. First, the paper extends the modified method of differential multipliers to empirical risk minimization with fairness constraints, thus providing an algorithm to enforce fairness in the central setting. Then, this algorithm is extended to the private federated learning setting. The proposed algorithm, FPFL, is tested on a federated version of the Adult dataset and an "unfair" version of the FEMNIST dataset. The experiments on these datasets show how private federated learning accentuates unfairness in the trained models, and how FPFL is able to mitigate such unfairness.

show abstract

Section: Private Federated Learningmentioning

confidence: 99%

Section: Extending the Algorithm To Private Federated Learningmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Enforcing fairness in private federated learning via the modified method of differential multipliers

Rodríguez-Gálvez¹,

Granqvist²,

Rogier³

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…In FL, participating clients collaboratively learn a shared model under the supervision of a central server: each communication round often starts with the server broadcasting the global model to the participants; these participants then perform computations on their local data, and send their aggregated updates back to the server to update the global model [Kairouz et al, 2019]. While FL can be performed on a relatively small number of clients, many applications involve a large number of edge devices, such as mobile phones, or sensors [Ramaswamy et al, 2020;Sheller et al, 2020;Granqvist et al, 2020]. This setting is referred to as cross-device FL.…”

Section: Introductionmentioning

confidence: 99%

Efficient and Private Federated Learning with Partially Trainable Networks

Sidahmed¹,

Xu²,

Garg³

et al. 2021

Preprint

View full text Add to dashboard Cite

Federated learning is used for decentralized training of machine learning models on a large number (millions) of edge mobile devices. It is challenging because mobile devices often have limited communication bandwidth and local computation resources. Therefore, improving the efficiency of federated learning is critical for scalability and usability. In this paper, we propose to leverage partially trainable neural networks, which freeze a portion of the model parameters during the entire training process, to reduce the communication cost with little implications on model performance. Through extensive experiments, we empirically show that Federated learning of Partially Trainable neural networks (FedPT) can result in superior communication-accuracy trade-offs, with up to 46× reduction in communication cost, at a small accuracy cost. Our approach also enables faster training, with a smaller memory footprint, and better utility for strong differential privacy guarantees. The proposed FedPT can be particularly interesting for pushing the limitations of overparameterization in on-device learning.

show abstract