Software applications that employ secure multi-party computation (MPC) can empower individuals and organizations to benefit from privacy-preserving data analyses when data sharing is encumbered by confidentiality concerns, legal constraints, or corporate policies. MPC is already being incorporated into software solutions in some domains; however, individual use cases do not fully convey the variety, extent, and complexity of the opportunities of MPC. This position paper articulates a role-based perspective that can provide some insight into how future research directions, infrastructure development and evaluation approaches, and deployment practices for MPC may evolve. Drawing on our own lessons from existing real-world deployments and the fundamental characteristics of MPC that make it a compelling technology, we propose a role-based conceptual framework for describing MPC deployment scenarios. Our framework acknowledges and leverages a novel assortment of roles that emerge from the fundamental ways in which MPC protocols support federation of functionalities and responsibilities. Defining these roles using the new opportunities for federation that MPC enables in turn can help identify and organize the capabilities, concerns, incentives, and trade-offs that affect the entities (software engineers, government regulators, corporate executives, end-users, and others) that participate in an MPC deployment scenario. This framework can not only guide the development of an ecosystem of modular and composable MPC tools, but can make explicit some of the opportunities that researchers and software engineers (and any organizations they form) have to differentiate and specialize the artifacts and services they choose to design, develop, and deploy. We demonstrate how this framework can be used to describe existing MPC deployment scenarios, how new opportunities in a scenario can be observed by disentangling roles inhabited by the involved parties, and how this can motivate the development of MPC libraries and software tools that specialize not by application domain but by role.
more »
« less
Polymath: Low-Latency MPC via Secure Polynomial Evaluations and Its Applications
Abstract While the practicality of secure multi-party computation (MPC) has been extensively analyzed and improved over the past decade, we are hitting the limits of efficiency with the traditional approaches of representing the computed functionalities as generic arithmetic or Boolean circuits. This work follows the design principle of identifying and constructing fast and provably-secure MPC protocols to evaluate useful high-level algebraic abstractions; thus, improving the efficiency of all applications relying on them. We present Polymath, a constant-round secure computation protocol suite for the secure evaluation of (multi-variate) polynomials of scalars and matrices, functionalities essential to numerous data-processing applications. Using precise natural precomputation and high-degree of parallelism prevalent in the modern computing environments, Polymath can make latency of secure polynomial evaluations of scalars and matrices independent of polynomial degree and matrix dimensions. We implement our protocols over the HoneyBadgerMPC library and apply it to two prominent secure computation tasks: privacy-preserving evaluation of decision trees and privacy-preserving evaluation of Markov processes. For the decision tree evaluation problem, we demonstrate the feasibility of evaluating high-depth decision tree models in a general n -party setting. For the Markov process application, we demonstrate that Poly-math can compute large powers of transition matrices with better online time and less communication.
more »
« less
- Award ID(s):
- 2055605
- NSF-PAR ID:
- 10414271
- Date Published:
- Journal Name:
- Proceedings on Privacy Enhancing Technologies
- Volume:
- 2022
- Issue:
- 1
- ISSN:
- 2299-0984
- Page Range / eLocation ID:
- 396 to 416
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Secure multi-party computation (MPC) allows multiple parties to jointly compute the output of a function while preserving the privacy of any individual party's inputs to that function. As MPC protocols transition from research prototypes to real-world applications, the usability of MPC-enabled applications is increasingly critical to their successful deployment and wide adoption. Our Web-MPC platform, designed with a focus on usability, has been deployed for privacy-preserving data aggregation initiatives with the City of Boston and the Greater Boston Chamber of Commerce. After building and deploying an initial version of this platform, we conducted a heuristic evaluation to identify additional usability improvements and implemented corresponding application enhancements. However, it is difficult to gauge the effectiveness of these changes within the context of real-world deployments using traditional web analytics tools without compromising the security guarantees of the platform. This work consists of two contributions that address this challenge: (1) the Web-MPC platform has been extended with the capability to collect web analytics using existing MPC protocols, and (2) this capability has been leveraged to conduct a usability study comparing the two version of Web-MPC (before and after the heuristic evaluation and associated improvements). While many efforts have focused on ways to enhance the usability of privacy-preserving technologies, this study can serve as a model for using a privacy-preserving data-driven approach in evaluating or enhancing the usability of privacy-preserving websites and applications deployed in real-world scenarios. The data collected in this study yields insights about the interplay between usability and security that can help inform future implementations of applications that employ MPC.more » « less
-
There is great demand for scalable, secure, and efficient privacy-preserving machine learning models that can be trained over distributed data. While deep learning models typically achieve the best results in a centralized non-secure setting, different models can excel when privacy and communication constraints are imposed. Instead, tree-based approaches such as XGBoost have attracted much attention for their high performance and ease of use; in particular, they often achieve state-of-the-art results on tabular data. Consequently, several recent works have focused on translating Gradient Boosted Decision Tree (GBDT) models like XGBoost into federated settings, via cryptographic mechanisms such as Homomorphic Encryption (HE) and Secure Multi-Party Computation (MPC). However, these do not always provide formal privacy guarantees, or consider the full range of hyperparameters and implementation settings. In this work, we implement the GBDT model under Differential Privacy (DP). We propose a general framework that captures and extends existing approaches for differentially private decision trees. Our framework of methods is tailored to the federated setting, and we show that with a careful choice of techniques it is possible to achieve very high utility while maintaining strong levels of privacy.more » « less
-
Abstract Background Logistic regression (LR) is a widely used classification method for modeling binary outcomes in many medical data classification tasks. Researchers that collect and combine datasets from various data custodians and jurisdictions can greatly benefit from the increased statistical power to support their analysis goals. However, combining data from different sources creates serious privacy concerns that need to be addressed. Methods In this paper, we propose two privacy-preserving protocols for performing logistic regression with the Newton–Raphson method in the estimation of parameters. Our proposals are based on secure Multi-Party Computation (MPC) and tailored to the honest majority and dishonest majority security settings. Results The proposed protocols are evaluated against both synthetic and real-world datasets in terms of efficiency and accuracy, and a comparison is made with the ordinary logistic regression. The experimental results demonstrate that the proposed protocols are highly efficient and accurate. Conclusions Our work introduces two iterative algorithms to enable the distributed training of a logistic regression model in a privacy-preserving manner. The implementation results show that our algorithms can handle large datasets from multiple sources.more » « less
-
Malkin, T. (Ed.)Existing approaches to secure multiparty computation (MPC) require all participants to commit to the entire duration of the protocol. As interest in MPC continues to grow, it is inevitable that there will be a desire to use it to evaluate increasingly complex functionalities, resulting in computations spanning several hours or days. Such scenarios call for a dynamic participation model for MPC where participants have the flexibility to go offline as needed and (re)join when they have available computational resources. Such a model would also democratize access to privacy-preserving computation by facilitating an “MPC-as-a-service” paradigm—the deployment of MPC in volunteer-operated networks (such as blockchains, where dynamism is inherent) that perform computation on behalf of clients. In this work, we initiate the study of fluid MPC, where parties can dynamically join and leave the computation. The minimum commitment required from each participant is referred to as fluidity, measured in the number of rounds of communication that it must stay online. Our contributions are threefold: We provide a formal treatment of fluid MPC, exploring various possible modeling choices. We construct information-theoretic fluid MPC protocols in the honest-majority setting. Our protocols achieve maximal fluidity, meaning that a party can exit the computation after receiving and sending messages in one round. We implement our protocol and test it in multiple network settings.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Journal Article:
- https://doi.org/10.2478/popets-2022-0020
