secure computation
Distributed Learning without Distress: Privacy-Preserving Empirical Risk Minimization
Distributed learning allows a group of independent data owners to collaboratively learn a model over their data sets without exposing their private data. We present a distributed learning approach that combines differential privacy with secure multi-party computation. We explore two popular methods of differential privacy, output perturbation and gradient perturbation, and advance the state-of-the-art for both methods in the distributed learning setting. In our output perturbation method, the parties combine local models within a secure computation and then add the required differential privacy noise before revealing the model. In our gradient perturbation method, the data owners collaboratively train a global model via an iterative learning algorithm. At each iteration, the parties aggregate their local gradients within a secure computation, adding sufficient noise to ensure privacy before the gradient updates are revealed. For both methods, we show that the noise can be reduced in the multi-party setting by adding the noise inside the secure computation after aggregation, asymptotically improving upon the best previous results. Experiments on real world data sets demonstrate that our methods provide substantial utility gains for typical privacy requirements.
Distributed Learning without Distress: Privacy-Preserving Empirical Risk Minimization
Distributed learning allows a group of independent data owners to collaboratively learn a model over their data sets without exposing their private data. We present a distributed learning approach that combines differential privacy with secure multi-party computation. We explore two popular methods of differential privacy, output perturbation and gradient perturbation, and advance the state-of-the-art for both methods in the distributed learning setting. In our output perturbation method, the parties combine local models within a secure computation and then add the required differential privacy noise before revealing the model. In our gradient perturbation method, the data owners collaboratively train a global model via an iterative learning algorithm. At each iteration, the parties aggregate their local gradients within a secure computation, adding sufficient noise to ensure privacy before the gradient updates are revealed. For both methods, we show that the noise can be reduced in the multi-party setting by adding the noise inside the secure computation after aggregation, asymptotically improving upon the best previous results. Experiments on real world data sets demonstrate that our methods provide substantial utility gains for typical privacy requirements.
Optimizing Privacy-Preserving Primitives to Support LLM-Scale Applications
Jandali, Yaman, Zhang, Ruisi, Sheybani, Nojan, Koushanfar, Farinaz
Privacy-preserving technologies have introduced a paradigm shift that allows for realizable secure computing in real-world systems. The significant barrier to the practical adoption of these primitives is the computational and communication overhead that is incurred when applied at scale. In this paper, we present an overview of our efforts to bridge the gap between this overhead and practicality for privacy-preserving learning systems using multi-party computation (MPC), zero-knowledge proofs (ZKPs), and fully homomorphic encryption (FHE). Through meticulous hardware/software/algorithm co-design, we show progress towards enabling LLM-scale applications in privacy-preserving settings. We demonstrate the efficacy of our solutions in several contexts, including DNN IP ownership, ethical LLM usage enforcement, and transformer inference.
Compact: Approximating Complex Activation Functions for Secure Computation
Islam, Mazharul, Arora, Sunpreet S., Chatterjee, Rahul, Rindal, Peter, Shirvanian, Maliheh
Secure multi-party computation (MPC) techniques can be used to provide data privacy when users query deep neural network (DNN) models hosted on a public cloud. State-of-the-art MPC techniques can be directly leveraged for DNN models that use simple activation functions (AFs) such as ReLU. However, DNN model architectures designed for cutting-edge applications often use complex and highly non-linear AFs. Designing efficient MPC techniques for such complex AFs is an open problem. Towards this, we propose Compact, which produces piece-wise polynomial approximations of complex AFs to enable their efficient use with state-of-the-art MPC techniques. Compact neither requires nor imposes any restriction on model training and results in near-identical model accuracy. We extensively evaluate Compact on four different machine-learning tasks with DNN architectures that use popular complex AFs SiLU, GeLU, and Mish. Our experimental results show that Compact incurs negligible accuracy loss compared to DNN-specific approaches for handling complex non-linear AFs. We also incorporate Compact in two state-of-the-art MPC libraries for privacy-preserving inference and demonstrate that Compact provides 2x-5x speedup in computation compared to the state-of-the-art approximation approach for non-linear functions -- while providing similar or better accuracy for DNN models with large number of hidden layers
A Survey of Secure Computation Using Trusted Execution Environments
Li, Xiaoguo, Zhao, Bowen, Yang, Guomin, Xiang, Tao, Weng, Jian, Deng, Robert H.
As an essential technology underpinning trusted computing, the trusted execution environment (TEE) allows one to launch computation tasks on both on- and off-premises data while assuring confidentiality and integrity. This article provides a systematic review and comparison of TEE-based secure computation protocols. We first propose a taxonomy that classifies secure computation protocols into three major categories, namely secure outsourced computation, secure distributed computation and secure multi-party computation. To enable a fair comparison of these protocols, we also present comprehensive assessment criteria with respect to four aspects: setting, methodology, security and performance. Based on these criteria, we review, discuss and compare the state-of-the-art TEE-based secure computation protocols for both general-purpose computation functions and special-purpose ones, such as privacy-preserving machine learning and encrypted database queries. To the best of our knowledge, this article is the first survey to review TEE-based secure computation protocols and the comprehensive comparison can serve as a guideline for selecting suitable protocols for deployment in practice. Finally, we also discuss several future research directions and challenges.
Secure Quantized Training for Deep Learning
We implement training of neural networks in secure multi-party computation (MPC) using quantization commonly used in said setting. We are the first to present an MNIST classifier purely trained in MPC that comes within 0.2 percent of the accuracy of the same convolutional neural network trained via plaintext computation. More concretely, we have trained a network with two convolutional and two dense layers to 99.2% accuracy in 3.5 hours (under one hour for 99% accuracy). We have also implemented AlexNet for CIFAR-10, which converges in a few hours. We develop novel protocols for exponentiation and inverse square root. Finally, we present experiments in a range of MPC security models for up to ten parties, both with honest and dishonest majority as well as semi-honest and malicious security.
DP-Cryptography
On Feb 15, 2019, John Abowd, chief scientist at the U.S. Census Bureau, announced the results of a reconstruction attack that they proactively launched using data released under the 2010 Decennial Census.19 The decennial census released billions of statistics about individuals like "how many people of the age 10-20 live in New York City" or "how many people live in four-person households." Using only the data publicly released in 2010, an internal team was able to correctly reconstruct records of address (by census block), age, gender, race, and ethnicity for 142 million people (about 46% of the U.S. population), and correctly match these data to commercial datasets circa 2010 to associate personal-identifying information such as names for 52 million people (17% of the population). This is not specific to the U.S. Census Bureau--such attacks can occur in any setting where statistical information in the form of deidentified data, statistics, or even machine learning models are released. That such attacks are possible was predicted over 15 years ago by a seminal paper by Irit Dinur and Kobbi Nissim12--releasing a sufficiently large number of aggregate statistics with sufficiently high accuracy provides sufficient information to reconstruct the underlying database with high accuracy. The practicality of such a large-scale reconstruction by the U.S. Census Bureau underscores the grand challenge that public organizations, industry, and scientific research faces: How can we safely disseminate results of data analysis on sensitive databases? An emerging answer is differential privacy. An algorithm satisfies differential privacy (DP) if its output is insensitive to adding, removing or changing one record in its input database. DP is considered the "gold standard" for privacy for a number of reasons. It provides a persuasive mathematical proof of privacy to individuals with several rigorous interpretations.25,26 The DP guarantee is composable and repeating invocations of differentially private algorithms lead to a graceful degradation of privacy.
Fairness in the Eyes of the Data: Certifying Machine-Learning Models
Segal, Shahar, Adi, Yossi, Pinkas, Benny, Baum, Carsten, Ganesh, Chaya, Keshet, Joseph
We present a framework that allows to certify the fairness degree of a model based on an interactive and privacy-preserving test. The framework verifies any trained model, regardless of its training process and architecture. Thus, it allows us to evaluate any deep learning model on multiple fairness definitions empirically. We tackle two scenarios, where either the test data is privately available only to the tester or is publicly known in advance, even to the model creator. We investigate the soundness of the proposed approach using theoretical analysis and present statistical guarantees for the interactive test. Finally, we provide a cryptographic technique to automate fairness testing and certified inference with only black-box access to the model at hand while hiding the participants' sensitive data.
Microsoft Unveils Falcon To Secure Computation of AI Models
Researchers from Microsoft, Princeton University, Technion and Algorand Foundation recently introduced a new framework known as Falcon. Falcon is an end-to-end 3-party protocol that can be used for fast and secure computations of deep learning algorithms on larger networks. Today, a vast amount of private data and sensitive information is continuously being generated. According to the researchers, combining this data with deep learning algorithms can transform the current social and technological scenario. Falcon is a deep learning framework that provides support for both training and inference with malicious security guarantees.
Distributed Learning without Distress: Privacy-Preserving Empirical Risk Minimization
Jayaraman, Bargav, Wang, Lingxiao, Evans, David, Gu, Quanquan
Distributed learning allows a group of independent data owners to collaboratively learn a model over their data sets without exposing their private data. We present a distributed learning approach that combines differential privacy with secure multi-party computation. We explore two popular methods of differential privacy, output perturbation and gradient perturbation, and advance the state-of-the-art for both methods in the distributed learning setting. In our output perturbation method, the parties combine local models within a secure computation and then add the required differential privacy noise before revealing the model. In our gradient perturbation method, the data owners collaboratively train a global model via an iterative learning algorithm.