Scalable Privacy-Preserving Distributed Extremely Randomized Trees for Structured Data With Multiple Colluding Parties

Abstract

Today, in many real-world applications of machine learning algorithms, the data is stored on multiple sources instead of at one central repository. In many such scenarios, due to privacy concerns and legal obligations, e.g., for medical data, and communication/computation overhead, for instance for large scale data, the raw data cannot be transferred to a center for analysis. Therefore, new machine learning approaches are proposed for learning from the distributed data in such settings. In this paper, we extend the distributed Extremely Randomized Trees (ERT) approach w.r.t. privacy and scalability. First, we extend distributed ERT to be resilient w.r.t. the number of colluding parties in a scalable fashion. Then, we extend the distributed ERT to improve its scalability without any major loss in classification performance. We refer to our proposed approach as k-PPD-ERT or Privacy-Preserving Distributed Extremely Randomized Trees with k colluding parties.

Extremely Randomized Trees With Privacy Preservation for Distributed Structured Health Data

“…Several studies focus on tree-based algorithms to train a model based on decentralized data. For instance, in [53], [23], [42], [54], [55], [56] the ID3 [57], random forest [58], and extremely randomized trees [59] algorithms are extended to be utilized for scenarios in which the training data is decentralized. These methods employ encryption or SMC to protect the privacy of data holders.…”

“…In our previous studies in [7], [4], [6], [9], [103], [104], [105], we have developed energy-efficient ML techniques for mobile-health and wearable technologies, including in the distributed and federated learning [106], [12], [102]. On the other hand, we have looked into privacy-preserving distributed and federated learning techniques considering tree-based algorithms [55], [54], [56], [107], but not considering the resourceconstraints of mobile-health and wearable technologies. To address this gap, in this article, we propose a framework that jointly considers prediction performance, computation and communication overheads, and privacy concerns, which are all essential for resource-constrained mobile-health systems involving sensitive medical/personal data.…”

Towards adaptive technology in routine mental health care

“…In the second alternative, we conduct privacy-preserving data analysis without sharing the data, for example, in the literary works. [72][73][74][75][76] Several mining algorithms can learn machine-learning models by receiving partial information about the data in the learning process instead of raw data. The party/parties holding data can participate in the learning process by merely sharing partial information instead of the raw data.…”