FedAUXfdp: Differentially Private One-Shot Federated Distillation
Authors:
Haley Hoech,
Roman Rischke,
Karsten Müller,
Wojciech Samek
Abstract:
Federated learning suffers in the case of non-iid local datasets, i.e., when the distributions of the clients' data are heterogeneous. One promising approach to this challenge is the recently proposed method FedAUX, an augmentation of federated distillation with robust results on even highly heterogeneous client data. FedAUX is a partially $(ε, δ)$-differentially private method, insofar as the cli…
▽ More
Federated learning suffers in the case of non-iid local datasets, i.e., when the distributions of the clients' data are heterogeneous. One promising approach to this challenge is the recently proposed method FedAUX, an augmentation of federated distillation with robust results on even highly heterogeneous client data. FedAUX is a partially $(ε, δ)$-differentially private method, insofar as the clients' private data is protected in only part of the training it takes part in. This work contributes a fully differentially private modification, termed FedAUXfdp. We further contribute an upper bound on the $l_2$-sensitivity of regularized multinomial logistic regression. In experiments with deep networks on large-scale image datasets, FedAUXfdp with strong differential privacy guarantees performs significantly better than other equally privatized SOTA baselines on non-iid client data in just a single communication round. Full privatization of the modified method results in a negligible reduction in accuracy at all levels of data heterogeneity.
△ Less
Submitted 21 June, 2022; v1 submitted 30 May, 2022;
originally announced May 2022.
Auditing the Biases Enacted by YouTube for Political Topics in Germany
Authors:
Hendrik Heuer,
Hendrik Hoch,
Andreas Breiter,
Yannis Theocharis
Abstract:
With YouTube's growing importance as a news platform, its recommendation system came under increased scrutiny. Recognizing YouTube's recommendation system as a broadcaster of media, we explore the applicability of laws that require broadcasters to give important political, ideological, and social groups adequate opportunity to express themselves in the broadcasted program of the service. We presen…
▽ More
With YouTube's growing importance as a news platform, its recommendation system came under increased scrutiny. Recognizing YouTube's recommendation system as a broadcaster of media, we explore the applicability of laws that require broadcasters to give important political, ideological, and social groups adequate opportunity to express themselves in the broadcasted program of the service. We present audits as an important tool to enforce such laws and to ensure that a system operates in the public's interest. To examine whether YouTube is enacting certain biases, we collected video recommendations about political topics by following chains of ten recommendations per video. Our findings suggest that YouTube's recommendation system is enacting important biases. We find that YouTube is recommending increasingly popular but topically unrelated videos. The sadness evoked by the recommended videos decreases while the happiness increases. We discuss the strong popularity bias we identified and analyze the link between the popularity of content and emotions. We also discuss how audits empower researchers and civic hackers to monitor complex machine learning (ML)-based systems like YouTube's recommendation system.
△ Less
Submitted 21 July, 2021;
originally announced July 2021.
