Two-server Model Research Articles

Hawk: Accurate and Fast Privacy-Preserving Machine Learning Using Secure Lookup Table Computation

Training machine learning models on data from multiple entities without direct data sharing can unlock applications otherwise hindered by business, legal, or ethical constraints. In this work, we design and implement new privacy-preserving machine learning protocols for logistic regression and neural network models. We adopt a two-server model where data owners secret-share their data between two servers that train and evaluate the model on the joint data. A significant source of inefficiency and inaccuracy in existing methods arises from using Yao’s garbled circuits to compute non-linear activation functions. We propose new methods for computing non-linear functions based on secret-shared lookup tables, offering both computational efficiency and improved accuracy. Beyond introducing leakage-free techniques, we initiate the exploration of relaxed security measures for privacy-preserving machine learning. Instead of claiming that the servers gain no knowledge during the computation, we contend that while some information is revealed about access patterns to lookup tables, it maintains epsilon-dX-privacy. Leveraging this relaxation significantly reduces the computational resources needed for training. We present new cryptographic protocols tailored to this relaxed security paradigm and define and analyze the leakage. Our evaluations show that our logistic regression protocol is up to 9x faster, and the neural network training is up to 688x faster than SecureML. Notably, our neural network achieves an accuracy of 96.6% on MNIST in 15 epochs, outperforming prior benchmarks that capped at 93.4% using the same architecture.

Secure and efficient multi-key aggregation for federated learning

Federated learning (FL) is a distributed machine learning framework that aims to provide privacy for local datasets while learning a global machine learning model. However, the updates exchanged in FL may indirectly reveal information about the local training datasets. To protect the confidentiality of updates, various solutions using cryptographic schemes such as secret sharing, differential privacy, and homomorphic encryption have been developed. The solutions using secret sharing often have high communication costs, while those using differential privacy require a trade-off between accuracy and privacy. Homomorphic encryption has been used to address these challenges to reduce communication costs and provide provable security. However, existing solutions based on homomorphic encryption require clients to use the same public-private key pair, which may lead to local updates disclosure. To address the existing challenges, we propose a secure and efficient multi-key aggregation protocol (MKAgg) that utilizes homomorphic encryption. The protocol allows clients to drop out at any point during the process. We construct MKAgg based on a two-server model and adopt a proxy re-encryption scheme with additively homomorphic properties to implement secure and efficient ciphertext transformation and calculation. We provide security proof to demonstrate that our MKAgg protocol meets the required security standards. Furthermore, we perform an efficiency analysis of MKAgg and evaluate its performance on various datasets. The results affirm that MKAgg is both effective and efficient for aggregation in the multi-key setting. We then apply MKAgg in FL and develop a multi-key privacy-preserving neural network scheme called MKPNFL. We analyze the security of MKPNFL and conduct tests using real-world datasets. The results demonstrate that MKPNFL is secure and practical for real-world applications.

Privacy-preserving outsourcing decision tree evaluation from homomorphic encryption

A decision tree is a common algorithm in machine learning, which performs classification prediction based on a tree structure. In real-world scenarios, input attribute values may be sensitive and tree models are usually private, thus the computation potentially incurs security and privacy risks. In this paper, we focus on how to retain privacy when outsourcing decision tree evaluation. Leveraging homomorphic encryption, our construction achieves data confidentiality against semi-honest adversaries while enabling both the client (who provides the attributes) and the model holder to be offline during evaluation. Based on an unbalanced-computational-ability two-server model, we achieve single-branch evaluation, which is exponentially less computation than previous schemes that evaluated the entire decision tree. A proof-of-concept implementation demonstrates this. Additionally, with multi-key encryption and joint decryption, our protocol easily supports multi-client scenarios.

EPPSQ: Achieving efficient and privacy-preserving statistics queries over encrypted data in smart grids

Outsourcing encrypted data and query services to clouds have been widely adopted by data owners, such as in the energy, healthcare, and transportation sectors, usually for economic considerations. However, this is usually prone to data privacy breaches. In smart grids, although many privacy-preserving data query solutions have been presented in the literature, they either suffer from low query efficiencies or limited capabilities of statistics queries (say the aggregation, variance, or extrema only). To mitigate this gap, in this paper we propose an efficient and privacy-preserving statistics query scheme over encrypted data in smart grids, coined EPPSQ, which can achieve diverse statistics queries in an efficient and privacy-preserving way. Specifically, we first develop a public key encryption based secure data collection protocol for data owners. Then, we design a two-server model based privacy-preserving statistics query protocol for data requesters. In addition, four algorithms respectively for securely computing the statistics (i.e., max/min, variance, mean, and count) over the encrypted data in smart grids are crafted. Security analysis shows that the proposed EPPSQ scheme can effectively guarantee data privacy while outsourcing the encrypted data and query services in smart grids. Further, extensive experiments demonstrate that the proposed EPPSQ scheme outperforms existing CKKS-based or BFV-based studies in terms of computational efficiency.

Towards Efficient and Privacy-Preserving User-Defined Skyline Query Over Single Cloud

Skyline queries, especially those variants that allow users to define their own query criteria, are very promising and practical techniques in multi-criteria decision making applications. Meanwhile, the growing data volume drives the service providers to outsource their data to the cloud for reaping economic benefits. However, privacy concerns compel the outsourced data to be encrypted and require to perform the skyline queries over encrypted data. To achieve the privacy-preserving skyline queries, many schemes were proposed in the literature. However, those existing solutions cannot fully support the user-defined query criteria in skyline queries, and most of them employ a two-server model to support skyline queries over ciphertexts, which needs multi-round communications between the deployed two servers. In this article, we propose a privacy-preserving user-defined skyline query scheme in a single-server model, which eliminates extra communications. Specifically, we first formally define the user-defined skyline query. Then, based on the idea of converting order relations into computing the inner products of two multi-dimensional points, we design three predicate encryption schemes. Finally, we adopt these predicate encryption schemes to construct our proposed scheme. Detailed security analysis shows that these predicate encryption schemes are selectively secure, and the proposed user-defined skyline query scheme is privacy-preserving. In addition, extensive experiments are conducted, and the results show that our proposed scheme outperforms the alternative scheme by up to an order of magnitude in terms of computational costs when performing user-defined skyline queries.

LEGO: A hybrid toolkit for efficient 2PC-based privacy-preserving machine learning

Recently, privacy-preserving machine learning (PPML) has received a lot of research attention, due to the increasing demand for multiple data owners in training machine learning models. Thus, many works have been proposed for privacy-preserving machine learning, in which many breakthroughs have been made. However, However, there is still much room for improvement in terms of efficiency and accuracy. In this paper, we design a toolkit called LEGO that efficiently combines garbled circuits, secret sharing, oblivious transfer, and homomorphic encryption to reduce time and improve accuracy. Our work focuses on the two-server model where the two servers train the models on the different owners’ private data. For fully connected layers and convolutional layers, we utilize the associated multiplication triplets to reduce the communication. For the offline phase, we design a novel protocol based on oblivious transfer and improved the previous protocol based on homomorphic encryption. We use the MPC-friendly activation function to improve the performance. Besides, we accelerate the local computation by adding GPU support. Our work is implemented in C++, and it can securely train machine learning models, including linear regression, logistic regression, neural networks, and convolutional neural networks. The experimental results show that our work is significantly faster than the state-of-the-art works.

Toward Privacy-Preserving Aggregate Reverse Skyline Query With Strong Security

It has been witnessed that Aggregate Reverse Skyline (ARS) query has recently received a wide range of practical applications due to its marvelous property of identifying the influence of query requests. Nevertheless, the query users may hesitate to participate in such query services as the query requests and query results may leak sensitive personal data or valuable business data assets to the service providers. To tackle the concerns, a promising solution is to encrypt the query requests, conduct the ARS queries over encrypted query requests without decrypting, and return the encrypted query results. Unfortunately, many existing solutions are either deployed over a two-server model or unable to fully preserve query privacy. In this paper, we propose a novel privacy-preserving aggregate reverse skyline query (PPARS) scheme on a single server model while ensuring full query privacy. Specifically, we first transform the problem of ARS query into a combination of set membership test and logical expressions. Then, by employing the prefix encoding technique, bloom filter technique, and fully homomorphic encryption, we run the transformed logical expressions to obtain the encrypted aggregate values without leaking query requests, query results, and access patterns. Furthermore, we propose an interpolation-based packing technique to improve the communication efficiency of PPARS. Detailed and formal security analysis demonstrates that our proposed schemes can guarantee strong security. In addition, extensive experiments are conducted, and the results validate the efficiency of our proposed schemes.

Multi-key FHE without ciphertext-expansion in two-server model

Multi-key FHE without ciphertext-expansion in two-server model

Location Privacy-Preserving Task Recommendation With Geometric Range Query in Mobile Crowdsensing

In mobile crowdsensing, location-based task recommendation requires each data requester to submit a task-related geometric range to crowdsensing service providers such that they can match suitable workers within this range. Generally, a trusted server (i.e., database owner) should be deployed to protect location privacy during the process, which is not desirable in practice. In this paper, we propose the location privacy-preserving task recommendation (PPTR) schemes with geometric range query in mobile crowdsensing without the trusted database owner. Specifically, we first propose a PPTR scheme with linear search complexity, named PPTR-L, based on a two-server model. By leveraging techniques of polynomial fitting and randomizable matrix multiplication, PPTR-L enables the service provider to find the workers located in the data requester’s arbitrary geometric query range without disclosing the sensitive location privacy. To further improve query efficiency, we design a novel data structure for task recommendation and propose PPTR-F to achieve faster-than-linear search complexity. Through security analysis, it is shown that our schemes can protect the confidentiality of workers’ locations and data requesters’ queries. Extensive experiments are performed to demonstrate that our schemes can achieve high computational efficiency in terms of geometric range query.

Enabling Secure and Versatile Packet Inspection With Probable Cause Privacy for Outsourced Middlebox

Middlebox is an intermediary network equipment which can be outsourced to remote cloud servers for low-cost and customizable network services, such as load balancer and intrusion detection. A fundamental function of the middlebox is packet inspection, where both the packet header and payload are extracted and analyzed based on inspection rules. However, as the packet may contain sensitive individual or organizational information, it may raise severe privacy concerns without proper countermeasures. In this article, we propose a secure and versatile packet inspection scheme for outsourced middlebox. The proposed scheme builds upon two non-collusion cloud servers, where the first server conducts the inspection task over the ciphertext domain and the second reveal the inspection results. By doing so, the proposed scheme achieves versatile inspection functionalities: range-query-based header inspection and token-based payload inspection, while preserving the privacy of packet header, payload, and inspection rules. Moreover, we identify and address two challenging issues in the state-of-the-art literatures. First, we tailor the design of mis-operation resistant searchable homomorphic encryption (MR-SHE) and somewhat homomorphic encryption in the two-server model, to resist <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">offline dictionary attack on payload headers</i> . Second, we propose a key management mechanism with compelled access for the middlebox, to achieve <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">fine-grained probable cause privacy</i> . We also conduct extensive experiments and compare the results with existing schemes to demonstrate the feasibility of the proposed scheme.

Efficient distributed privacy-preserving collaborative outlier detection

As a common way to identify anomalous data, outlier detection is widely applicable for intrusions detection, adverse reactions analysis, financial fraud prevention, etc. The accuracy of outlier detection depends crucially on the number of data involved in the test, i.e., the more data participate in detection, the higher accuracy we get. For this reason, cross-dataset collaborative outlier detection is introduced to conquer the lack of data in a single-dataset setting. However, privacy concerns seriously prevent the application of collaborative outlier detection, since most organization are unwilling to share their data with others directly in practice. In this paper, we present efficient protocols for privacy preserving collaborative outlier detection from arbitrarily partitioned data using Local Distance-based Outlier Factor (LDOF). Our protocols fall in the two-server model where data owners distribute their private data among two non-colluding servers who detect outlier on the joint data by secure two-party computation. In particular, we perform arithmetic operations which takes place inside LDOF on arithmetic circuits instead of boolean circuits, and perform sorting operations on boolean circuits. Such a design enables standard operations are performed with suitable circuits, and thus our scheme is more efficient. In addition, to further improve protocol efficiency, local sensitive hash (LSH) is utilized to filter out data which do not need secure computation to reduce the the amount of shared data. We implement our system in C++ on real data. The security analysis and experiments show the security and efficiency of the proposed scheme. Our protocols are more faster than the state of previous methods.

Privacy-Aware and Efficient Mobile Crowdsensing with Truth Discovery

Truth discovery in mobile crowdsensing has recently received wide attention. It refers to the procedure for estimating the unknown user reliability from collected sensory data and inferring truthful information via reliability-aware data aggregation. Though widely studied in the plaintext domain, truth discovery remains largely under-explored in privacy-aware mobile crowdsensing. Existing works either do not consider user reliability issue or fall short of achieving practical cost efficiency, due to iterative transmission and computation over large ciphertexts from homomorphic cryptosystem. In this paper, we propose two new privacy-aware crowdsensing designs with truth discovery that significantly improve the bandwidth and computation performance on individual users. Our insight is to identify the core atomic operation in the iterative truth discovery procedure, and carefully craft security designs accordingly to enable efficient truth discovery in the ciphertext domain. Our first design is highly customized for the single-server setting, while our second design under the two-server model further shifts most of user workloads to the cloud server side. Both our designs protect individual sensory data and reliability degrees throughout the truth discovery procedure. Experiments show that compared with the prior result, our designs gain at least $30 \times$ 30 × and $10 \times$ 10 × savings on user communication and computation, respectively.

Achieving Secure and Dynamic Range Queries Over Encrypted Cloud Data

Cloud computing is motivating data owners to outsource their databases to the cloud. However, for privacy concerns, the sensitive data has to be encrypted before outsourcing, which inevitably posts a challenging task for effective data utilization. Existing work either focuses on keyword searches, or suffers from inadequate security guarantees or inefficiency. In this paper, we concentrate on multi-dimensional range queries over dynamic encrypted cloud data. We first propose a tree-based private range query scheme over dynamic encrypted cloud data (TRQED), which supports faster-than-linear range queries and protects single-dimensional privacy. Then, we discuss the defects of TRQED in terms of privacy-preservation. We modify the framework of the system by adopting a two-server model and put forward a safer range query scheme, called TRQED $^+$ . By newly designed secure node query (SNQ) and secure point query (SPQ), we propose the perturbation-based oblivious R-tree traversal (ORT) operation to preserve both path pattern and stronger single-dimensional privacy. Finally, we conduct comprehensive experiments on real-world datasets and perform comparisons with existing works to evaluate the performance of the proposed schemes. Experimental results show that our TRQED and TRQED $^+$ surpass the state-of-the-art methods in privacy-preservation level and efficiency.

Outsourced Biometric Identification With Privacy

Biometric identification typically scans a large-scale database of biometric records for finding a close enough match of an individual. This paper investigates how to outsource this computationally expensive scanning while protecting the privacy of both the database and the computation. Exploiting the inherent structures of biometric data and the properties of identification operations, we first present a privacy-preserving biometric identification scheme which uses a single server. We then consider its extensions in the two-server model. It achieves a higher level of privacy than our single-server solution assuming two servers are not colluding. Apart from somewhat homomorphic encryption, our second scheme uses batched protocols for secure shuffling and minimum selection. Our experiments on both synthetic and real data sets show that our solutions outperform existing schemes while preserving privacy.

A Two-Server Queue with Mutual Consultations

This paper analyses a two-server queueing model with consultation between the servers. Each of the servers provides consultation to the other server with a preemptive priority over customers. After consultation, the services in progress will be resumed. The arrivals to the system follows a Poisson process and requirement of consultations follow mutually independent Poisson processes; the duration of consultations follow mutually independent and identical exponential distributions. The service times of the customers at the servers are assumed to follow mutually independent phase type distributions. An explicit expression for the system stability is obtained. System performance indices are computed and numerical illustrations are provided. Further we consider two particular cases of the model described. Finally, we compare these through numerical investigations.

Privacy-preserving top-k keyword similarity search over outsourced cloud data

In this paper, we study the problem of privacy-preserving top-k keyword similarity search over outsourced cloud data. Taking edit distance as a measure of similarity, we first build up the similarity keyword sets for all the keywords in the data collection. We then calculate the relevance scores of the elements in the similarity keyword sets by the widely used tf-idf theory. Leveraging both the similarity keyword sets and the relevance scores, we present a new secure and efficient tree-based index structure for privacy-preserving top-k keyword similarity search. To prevent potential statistical attacks, we also introduce a two-server model to separate the association between the index structure and the data collection in cloud servers. Thorough analysis is given on the validity of search functionality and formal security proofs are presented for the privacy guarantee of our solution. Experimental results on real-world data sets further demonstrate the availability and efficiency of our solution.

On the optimal allocation of users to servers in an emergency service

This paper considers the optimal allocation of users to servers in an emergency medical service, in which two or three ambulances cooperate to serve calls from a region of rectangular shape. Queueing models are often used to approach emergency health services. We analyse three different dispatch policies to determine the best available unit to be sent and introduce a classification of districts in four different classes. Properties of the optimal boundaries, i.e., the ones that minimize the average response time between users and servers and which define districts, given the home locations of servers, are investigated. A two-server model is reviewed, and results are obtained for a three-server model, enlightening a non-monotone behaviour of the shift of the optimal boundaries, with respect to the equidistance curve, in the different states of the system, depending on the utilization ratio, on the policy assumed and on the locations of the servers. This way, it is not possible to give an a priori' general rule on the properties of the boundaries of the districts only on the basis of the number of busy servers at the specific time.

Heuristics for allocation of reconfigurable resources in a serial line with reliability considerations

We consider the allocation of reconfigurable resources in a serial line with machine failures. Each station is equipped with non-idling dedicated servers while the whole system is equipped with a finite number of reconfigurable servers. The reconfigurable servers are available to be assigned to any station and all servers are allowed to collaborate on a single job. We provide conditions for a policy to achieve throughput optimality. We also show in the two-station case that transition monotone optimal policies exist. We discuss heuristics based on the two-server model that reduce average holding costs significantly. These heuristics are compared to several heuristics from the literature via a detailed numerical study.

A two-queue and two-server model with a threshold-based control service policy

This paper investigates a queueing system consisting of two-parallel queues and two servers. The service policy is a threshold-based control one such that two thresholds v (⩾0) and N (⩾v) called control level are set up in one of the two queues, say, the second queue. At each epoch of service completion, the server decides which queue is to be served next according to the control level the number of customers in the second queue reaches. For both queues, the arrival processes are Poisson, and the service times are exponentially distributed with different means. We derive the generating functions of the stationary joint queue-length distribution, and then obtain the mean queue length and the mean waiting time for each queue.