Crowd-ML: A Privacy-Preserving Learning Framework for a Crowd of Smart Devices

By Jihun Hamm, Adam C. Champion, Guoxing Chen, Mikhail Belkin, Dong Xuan. 2015

65 citations recorded by the source corpus.

Smart devices with built-in sensors, computational capabilities, and network connectivity have become increasingly pervasive. Crowds of smart devices offer opportunities to collectively sense and perform computing tasks at an unprecedented scale. This paper presents Crowd-ML, a privacy-preserving machine learning framework for a crowd of smart devices, which can solve a wide range of learning problems for crowd sensing data with differential privacy guarantees. Crowd-ML endows a crowd sensing system with the ability to learn classifiers or predictors online from crowd sensing data privately with minimal computational overhead on devices and servers, suitable for practical large-scale use of the framework. We analyze the performance and scalability of Crowd-ML and implement the system with off-the-shelf smartphones as a proof of concept. We demonstrate the advantages of Crowd-ML with real and simulated experiments under various conditions.

Subfields: Artificial Intelligence, Computer Science Applications, Transportation

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Jihun Hamm
2. Quanquan Gu
3. Jun Zhao
4. Donald E. Brown
5. Yang Wang
6. Y. Yang
7. Chris J. M. Yoon
8. Clement Fung

In Computer Science Applications

1. Jihun Hamm
2. Donald E. Brown
3. Yang Wang
4. Quanquan Gu
5. Chris J. M. Yoon
6. Jun Zhao
7. Siu Cheung Hui
8. Ge Yu

In Transportation

1. Chris J. M. Yoon
2. Choong Seon Hong
3. Nguyen H. Tran
4. Clement Fung
5. Ivan Beschastnikh

Citations in the corpus

• Collecting and Analyzing Multidimensional Data with Local Differential Privacy (2019, methodology): “Towards this end, existing studies [14], [22] have suggested that the aggregator asks the selected user in each iteration to submit a noisy version of ∇ ′, by using the Laplace mechanism or Duchi et al.”
• Learning privately from multiparty data (2016, methodology): “, 2010), and private exchange of gradient information to minimize empirical risks incrementally (Rajkumar & Agarwal, 2012; Hamm et al., 2015).”
• Differentially Private Hypothesis Transfer Learning (2018, background): “gradients) among multiple parties during the iterative hypothesis training process [14,36,37,1,40].”
• Privacy-Preserving Collaborative Deep Learning with Irregular Participants. (2018, methodology): “Privacypreserving techniques such as secure multi-party computation or differential privacy have been applied to linear and logistic regression analysis [8], [9], k-means clustering [10], support vector machines [11], and crowd machine learning [12].”
• The Limitations of Federated Learning in Sybil Settings (2020, background): “Clients maintain a basic level of privacy by computing their model updates locally and independently, enabling collaborative ML in resource-constrained settings such as over a mobile network or in IoT(Internet of Things) deployments [24, 29, 49, 57].”
• A Survey on Security, Privacy, and Trust in Mobile Crowdsourcing (2018, background): “[90] proposed to utilize perturbation to support differential privacy of sensed data.”
• Privacy-preserving Machine Learning through Data Obfuscation (2018, background): “[12] applied the same idea to privacy machine learning in distributed IoT sensing systems.”
• Collecting and Analyzing Data from Smart Device Users with Local Differential Privacy (2016, methodology): “Towards this end, existing work [7,13] has suggested that the aggregator may ask the selected user in each iteration to submit a noisy version of ∇ ℓ ′ , by using the Laplace mechanism or Duchi et al. ’s method (i.e., Algorithm 1).”
• Model inversion attacks against collaborative inference (2019, background): “It has been shown that the privacy of training data is better protected if each participant in the distributed system uses his own dataset and never shares it with other participants [15, 40].”
• Federated Learning over Wireless Networks: Optimization Model Design and Analysis (2019, methodology): “Due to Big Data applications and complex models such as Deep Learning, training machine learning models needs to be distributed over multiple machines, giving rise to the researches on decentralized machine learning [6]–[8].”