The Value of Collaboration in Convex Machine Learning with Differential Privacy

Abstract

In this paper, we apply machine learning to distributed private data owned by multiple data owners, entities with access to non-overlapping training datasets. We use noisy, differentially-private gradients to minimize the fitness cost of the machine learning model using stochastic gradient descent. We quantify the quality of the trained model, using the fitness cost, as a function of privacy budget and size of the distributed datasets to capture the trade-off between privacy and utility in machine learning. This way, we can predict the outcome of collaboration among privacy-aware data owners prior to executing potentially computationally-expensive machine learning algorithms. Particularly, we show that the difference between the fitness of the trained machine learning model using differentially-private gradient queries and the fitness of the trained machine model in the absence of any privacy concerns is inversely proportional to the size of the training datasets squared and the privacy budget squared. We successfully validate the performance prediction with the actual performance of the proposed privacy-aware learning algorithms, applied to: financial datasets for determining interest rates of loans using regression; and detecting credit card frauds using support vector machines.

When Machine Learning Meets Privacy

“…Training ML in a distributed manner can naturally provide a certain level of privacy protection, as the local training data points are usually not shared among users. Moreover, different privacy protection schemes in centralised learning, such as encryption, perturbation, can be easily extended to the distributed learning settings [169]. In this sense, private ML in distributed systems have a lot in common with that of centralised ML.…”

Joint Protection of Energy Security and Information Privacy for Energy Harvesting: An Incentive Federated Learning Approach

“…Some works focus on the application of DP in machine learning. In [28], the authors apply DP to the distributed machine learning and quantify the model value. The authors in [29] investigate the DP-enabled federated learning mechanism, which designs the algorithm and analyzes the performance.…”

Privacy-Driven Learning Analytics

“…More recent work proposed DP-based additions to existing machine learning algorithms to allow the learning of models with strong privacy assurance. In the context of federated learning, one example proposes a DP-enhanced gradient descent algorithm, which enables the distributed learning of a model from multiple datasets hosted by different data custodians (Wu et al, 2020).…”