Single Round Of Communication Research Articles

Transfer learning via random forests: A one-shot federated approach

A one-shot federated transfer learning method using random forests (FTRF) is developed to improve the prediction accuracy at a target data site by leveraging information from auxiliary sites. Both theoretical and numerical results show that the proposed federated transfer learning approach is at least as accurate as the model trained on the target data alone regardless of possible data heterogeneity, which includes imbalanced and non-IID data distributions across sites and model mis-specification. FTRF has the ability to evaluate the similarity between the target and auxiliary sites, enabling the target site to autonomously select more similar site information to enhance its predictive performance. To ensure communication efficiency, FTRF adopts the model averaging idea that requires a single round of communication between the target and the auxiliary sites. Only fitted models from auxiliary sites are sent to the target site. Unlike traditional model averaging, FTRF incorporates predicted outcomes from other sites and the original variables when estimating model averaging weights, resulting in a variable-dependent weighting to better utilize models from auxiliary sites to improve prediction. Five real-world data examples show that FTRF reduces the prediction error by 2-40% compared to methods not utilizing auxiliary information.

Jointly Improving the Sample and Communication Complexities in Decentralized Stochastic Minimax Optimization

We propose a novel single-loop decentralized algorithm, DGDA-VR, for solving the stochastic nonconvex strongly-concave minimax problems over a connected network of agents, which are equipped with stochastic first-order oracles to estimate their local gradients. DGDA-VR, incorporating variance reduction, achieves O(ε^−3) oracle complexity and O(ε^−2) communication complexity without resorting to multi-communication rounds – both are optimal, i.e., matching the lower bounds for this class of problems. Since DGDA-VR does not require multiple communication rounds, it is applicable to a broader range of decentralized computational environments. To the best of our knowledge, this is the first distributed method using a single communication round in each iteration to jointly optimize the oracle and communication complexities for the problem considered here.

Federated fusion learning with attention mechanism for multi-client medical image analysis

Federated Learning (FL) has gained significant attention because of its potential for privacy-preserving distributed learning. However, statistical heterogeneity and label scarcity remain major issues in multi-client scenarios. Furthermore, existing FL algorithms do not consider the unique data distribution of each client, which is essential for enhancing the overall performance of a model. In this study, a new FL algorithm called federated fusion learning (FFL) is proposed to address the challenges of statistical heterogeneity and label scarcity in multiclient data fusion. In FFL, each device calculates its data distribution parameters and sends them to the server in a single communication round. The server then uses the received distribution parameters to generate synthetic data for each client and combines them to create a larger dataset. FFL comprises three modules that identify the latent space of each client, intelligently fuse various modalities at the server, and construct global weights. The study also introduces a novel refinement mechanism that embeds training images into the W space, thereby fusing the latent spaces of different clients using a custom multiclient latent space fusion (MCLSF) module. The attention mechanism used in FFL enables effective multiclient fusion by leveraging the inherent correlations between various data modalities. Ten datasets sourced from MedMNIST encompassing diverse imaging modalities such as X-rays, computed tomography (CT), and optical coherence tomography (OCT) were considered. In addition, detailed ablation studies were conducted on multiple client configurations and non-independent and identical (non-IID) setups to ensure the generalizability of the proposed approach. The results demonstrate the enhanced generalizability and overall performance of the global model compared to existing state-of-the-art methods, confirming its broader applicability and effectiveness.

Communication-Efficient Pseudo Value-Based Random Forests for Federated Survival Analysis

Federated Survival Analysis (FSA) is an emerging technique for analyzing decentralized survival data while preserving data privacy and providing more generalized survival predictions. Existing FSA methods often rely on deep learning models, which can be computationally expensive and require substantial data and communication rounds. Recent research has demonstrated that ensemble-based approaches like random survival forests can achieve comparable performance to deep learning models with a single communication round within a federated learning (FL) framework, especially when dealing with small, decentralized datasets. However, these approaches have yet to address the challenges specific to FSA, such as data heterogeneity, non-uniform censoring, and competing risks. To address these challenges, we propose FedPRF, an FL framework for survival analysis based on Federated Pseudo Values (FPV) based random forests model, we call it RFpseudo. FedPRF introduces FPV to handle issues related to censoring and effectively address the unique challenges of FSA. FedPRF is computationally efficient, requiring only two communication rounds across clients: (i) computing FPV and (ii) aggregating a subset of trees trained locally at clients. Extensive experiments on distributed survival data with a single event and multiple competing events demonstrate that FedPRF achieves performance close to the gold-standard centralized training setting and outperforms deep learning-based FSA approaches. Importantly, FedPRF maintains the interpretability of centrally trained survival models. Furthermore, it is scalable to large-scale data and highly distributed settings with numerous clients.

Fast and private multi-dimensional range search over encrypted data

For businesses looking to outsource their data to remote servers, cloud-based data storage is a popular choice. It is popular due to its flexibility, cost-effectiveness, and widespread availability. However, ensuring the confidentiality of data is a critical challenge that must be addressed. As a response to this issue, searchable encryption techniques have been developed. These techniques enable search queries to be performed on encrypted data while still keeping the plaintext confidential. While most existing symmetric searchable encryption schemes are designed for one-dimensional data records or document-keyword inverted indices, this paper introduces MDRSSE, a novel symmetric searchable encryption scheme specifically tailored for multi-dimensional range search. MDRSSE stands out as one of the pioneering SSE schemes to support multi-dimensional range search efficiently, without incurring undetermined additional communication or computation costs. By employing a single round of communication between the client and server, MDRSSE enables an honest-but-curious server to respond to multi-dimensional range queries without gaining knowledge of the data records or revealing the search query. Notably, MDRSSE boasts the lowest overall search complexity compared to existing state-of-the-art symmetric searchable encryption schemes designed for multi-dimensional range search. Extensive experimental tests were conducted to validate the robustness and practicality of our proposed scheme. The results demonstrate that, for a dataset consisting of 100K records with 12 dimensions (with each leaf node holding 500 records), it takes only 2.2 seconds to generate the encrypted dataset, and the overall setup phase completes within 2.5 seconds. Furthermore, for a range query encompassing 50 nodes, the search time is less than 2 ms and 3 ms for the client and server, respectively. MDRSSE achieves semantic security under the IND-CPA assumption, all without requiring additional storage size at the server.

Scaling survival analysis in healthcare with federated survival forests: A comparative study on heart failure and breast cancer genomics

Survival analysis is a fundamental tool in medicine, modeling the time until an event of interest occurs in a population. However, in real-world applications, survival data are often incomplete, censored, distributed, and confidential, especially in healthcare settings where privacy is critical. The scarcity of data can severely limit the scalability of survival models to distributed applications that rely on large data pools. Federated learning is a promising technique that enables machine learning models to be trained on multiple datasets without compromising user privacy, making it particularly well-suited for addressing the challenges of survival data and large-scale survival applications. Despite significant developments in federated learning for classification and regression, many directions remain unexplored in the context of survival analysis. In this work, we propose an extension of the Federated Survival Forest algorithm, called FedSurF++. This federated ensemble method constructs random survival forests in heterogeneous federations. Specifically, we investigate several new tree sampling methods from client forests and compare the results with state-of-the-art survival models based on neural networks. The key advantage of FedSurF++ is its ability to achieve comparable performance to existing methods while requiring only a single communication round to complete. The extensive empirical investigation results in a significant improvement from the algorithmic and privacy preservation perspectives, making the original FedSurF algorithm more efficient, robust, and private. We also present results on two real-world datasets – a heart failure dataset from the Lombardy HFData project and Fed-TCGA-BRCA from the Falmby suite – demonstrating the success of FedSurF++ in real-world healthcare studies. Our results underscore the potential of FedSurF++ to improve the scalability and effectiveness of survival analysis in distributed settings while preserving user privacy.

Compressing Distributed Network Sketches With Traffic-Aware Summaries

Network measurements are important for identifying congestion, DDoS attacks, and more. To support real-time analytics, stream ingestion is performed jointly by multiple nodes, each observing part of the traffic, periodically reporting its measurements to a single centralized server that aggregates them. To avoid communication congestion, each node reports a compressed version of its collected measurements. Traditionally, nodes symmetrically report summaries of the same size computed on their data. We explain that to maximize the accuracy of the joint measurement, nodes should imply various compression ratios on their measurements based on the amount of traffic observed by each node. We illustrate the approach for three common sketches: The Count-Min sketch (CM), which estimates flow sizes as well as for the K-minimum-values (KMV) sketch and the HyperLogLog (HLL), which both estimate the number of distinct flows. For each sketch, we compute node compression ratios based on the traffic distribution. In general, this is done with a single round of communication with the central server, after which the compression ratio for each node can be computed. We perform extensive simulations for the sketches and analytically show that, under real-world scenarios, our sketches send smaller summaries than traditional ones while retaining similar error bounds.

An Efficient Threshold Access-Structure for RLWE-Based Multiparty Homomorphic Encryption

We propose and implement a multiparty homomorphic encryption (MHE) scheme with a t-out-of-N-threshold access-structure that is efficient and does not require a trusted dealer in the common random string model. We construct this scheme from the ring-learning-with-error assumptions and as an extension of the MHE scheme of Mouchet et al. (PETS 21). By means of a specially adapted share re-sharing procedure, this extension can be used to relax the N-out-of-N-threshold access-structure of the original scheme into a t-out-of-N-threshold one. This procedure introduces only a single round of communication during the setup phase, after which any set of at least t parties can compute a t-out-of-t additive sharing of the secret-key with no interaction; this new sharing can be used directly in the scheme of Mouchet et al. We show that, by performing Shamir re-sharing over the MHE ciphertext-space ring with a carefully chosen exceptional set, this reconstruction procedure can be made secure and has negligible overhead. Moreover, it only requires the parties to store a constant-size state after its setup phase. Hence, in addition to fault tolerance, lowering the corruption threshold also yields considerable efficiency benefits, by enabling the distribution of batched secret-key operations among the online parties. We implemented and open-sourced our scheme in the Lattigo library.

Enumerating all bilocal Clifford distillation protocols through symmetry reduction

Entanglement distillation is an essential building block in quantum communication protocols. Here, we study the class of near-term implementable distillation protocols that use bilocal Clifford operations followed by a single round of communication. We introduce tools to enumerate and optimise over all protocols for up to n=5 (not necessarily equal) Bell-diagonal states using a commodity desktop computer. Furthermore, by exploiting the symmetries of the input states, we find all protocols for up to n=8 copies of a Werner state. For the latter case, we present circuits that achieve the highest fidelity with perfect operations and no decoherence. These circuits have modest depth and number of two-qubit gates. Our results are based on a correspondence between distillation protocols and double cosets of the symplectic group, and improve on previously known protocols.

A Cooperative Hierarchical Multi-Agent System for EV Charging Scheduling in Presence of Multiple Charging Stations

The increasing penetration of plug-in electric vehicles (EVs) to the electrical grid raises concerns over secure and economic operation of the system. A coordination mechanism between system operator and EV aggregators is necessary to ensure that the system is operated within the security limits, and to reduce the charging costs while satisfying EV users’ energy needs. In this work, we present a cooperative hierarchical multi-agent system and propose an EV charging scheduling strategy in order to minimize the demand and energy charges while meeting the EV users’ energy requirements and satisfying the system security constraints. Within the designed framework, the higher-level agents calculate a set of proposed control signals by solving the designated optimization problems, and send them to the lower-level agents to facilitate an optimal scheduling in line with the aforementioned objectives. Through this hierarchically distributed approach, it is possible to effectively coordinate multiple EV charging stations without the need of direct communication or any prior information related to EV arrivals. The computational complexity of the problem is reduced by distributing the work among agents, and the privacy of sensitive data, such as system topology, load profiles, and EV parameters, is preserved. Moreover, unlike the traditional distributed solution methods that converge iteratively, the proposed approach calculates the optimal charging schedule after a single round of communication. The efficacy of the proposed methodology is demonstrated by a series of case studies on 33-bus and 118-bus distribution test feeders.

Towards real-time privacy-preserving video surveillance

Video surveillance on a massive scale can be a vital tool for law enforcement agencies. To mitigate the serious privacy concerns of wide-scale video surveillance, researchers have designed secure and privacy-preserving protocols that obliviously match live feeds against a suspects’ database. However, existing approaches are very expensive in terms of computation and communication costs and, as a result, they do not scale well for ubiquitous deployment. To this end, we propose a general framework for privacy-preserving identification that operates by storing an encrypted version of the suspects’ database at the video cameras. We show that this approach (i) reduces the protocol to a single round of communication between the camera and the server and (ii) speeds up the computation times significantly through the use of input-independent precomputations. We apply our framework to two practical use-cases, namely, face and license plate number recognition. In addition to the identification result, our face recognition protocol discloses some trivial information to the database server; however, this information is not sufficient for the server to infer any meaningful characteristics about the underlying individuals. On the other hand, the license plate recognition protocol is provably secure and can also handle minor character recognition errors that often occur in such systems. We implemented working prototypes of both surveillance systems and our experimental results are very promising. In the case of face recognition, and for a database of 100 suspects, the online computation time at the camera and the server is 155 ms and 34 ms, respectively, while the online communication cost is only 12 KB. Similarly, for a database of 3000 entries, license plate recognition requires only 232 ms and 75 ms at the camera and the server, respectively, while the online communication cost is 375 KB.

Facilitating Secure and Efficient Spatial Query Processing on the Cloud

Database outsourcing is a common cloud computing paradigm that allows data owners to take advantage of its on-demand storage and computational resources. The main challenge is maintaining data confidentiality with respect to untrusted parties i.e., cloud service provider, as well as providing relevant query results in real-time to authenticated users. Existing approaches either compromise confidentiality of the data or suffer from high communication cost between the server and the user. To overcome this problem, we propose a dual transformation and encryption scheme for spatial data, where encrypted queries are executed entirely at the service provider on the encrypted database and encrypted results are returned to the user. The user issues encrypted spatial range queries to the service provider and then uses the encryption key to decrypt the query response returned. This allows a balance between the security of data and efficient query response as the queries are processed on encrypted data at the cloud server. Moreover, we compare with existing approaches on large datasets and show that this approach reduces the average query communication cost between the authorized user and service provider, as only a single round of communication is required by the proposed approach.

Privacy-Preserving Outsourced Clinical Decision Support System in the Cloud

In this paper, we propose a privacy-preserving clinical decision support system using Naive Bayesian (NB) classifier, hereafter referred to as Peneus, designed for the outsourced cloud computing environment. Peneus allows one to use patient health information to train the NB classifier privately, which can then be used to predict a patient's (undiagnosed) disease based on his/her symptoms in a single communication round. Specifically, we design secure Single Instruction Multiple Data (SIMD) integer circuits using the fully homomorphic encryption scheme, which can greatly increase the performance compared with the original secure integer circuit. Then, we present a privacy-preserving historical Personal Health Information (PHI) aggregation protocol to allow different PHI sources to be securely aggregated without the risk of compromising the privacy of individual data owner. Also, secure NB classifier is constructed to achieve secure disease prediction in the cloud without the help of an additional non-colluding computation server. We then demonstrate that Peneus achieves the goal of patient health status monitoring without privacy leakage to unauthorized parties, as well as the utility and the efficiency of Peneus using simulations and analysis.

Computing large independent sets in a single round

Independent sets play a central role in distributed algorithmics. We examine here the minimal requirements for computing non-trivial independent sets. In particular, we focus on algorithms that operate in a single communication round. A classic result of Linial shows that a constant number of rounds does not suffice to compute a maximal independent set. We are therefore interested in the size of the solution that can be computed, especially in comparison to the optimal. Our main result is a randomized one-round algorithm that achieves poly-logarithmic approximation on graphs of polynomially bounded-independence. Specifically, we show that the algorithm achieves the Caro-Wei bound (an extension of the Turán bound for independent sets) in general graphs up to a constant factor, and that the Caro-Wei bound yields a poly-logarithmic approximation on bounded-independence graphs. The algorithm uses only a single bit message and operates in a beeping model, where a node receives only the disjunction of the bits transmitted by its neighbors. We give limitation results that show that these are the minimal requirements for obtaining non-trivial solutions. In particular, a sublinear approximation cannot be obtained in a single round on general graphs, nor when nodes cannot both transmit and receive messages. We also show that our analysis of the Caro-Wei bound on polynomially bounded-independence graphs is tight, and that the poly-logarithmic approximation factor does not extend to mathrm {O}(1)-claw free graphs.

Outsourcing secure two‐party computation as a black box

AbstractSecure multiparty computation (SMC) offers a technique to preserve functionality and data privacy in mobile applications. Current protocols that make this costly cryptographic construction feasible on mobile devices securely outsource the bulk of the computation to a cloud provider. However, these outsourcing techniques are built on specific secure computation assumptions and tools, and applying new SMC ideas to the outsourced setting requires the protocols to be completely rebuilt and proven secure. In this work, we develop a generic technique for lifting any secure two‐party computation protocol into an outsourced two‐party SMC protocol. By augmenting the function being evaluated with auxiliary consistency checks and input values, we can create an outsourced protocol with low overhead cost. Our implementation and evaluation show that in the best case our outsourcing additions execute within the confidence intervals of two servers running the same computation and consume approximately the same bandwidth. In addition, the mobile device itself uses minimal bandwidth over a single round of communication. This work demonstrates that efficient outsourcing is possible with any underlying SMC scheme and provides an outsourcing protocol that is efficient and directly applicable to current and future SMC techniques. Copyright © 2016 John Wiley & Sons, Ltd.

Communication-efficient algorithms for statistical optimization

We analyze two communication-efficient algorithms for distributed optimization in statistical settings involving large-scale data sets. The first algorithm is a standard averaging method that distributes the N data samples evenly to m machines, performs separate minimization on each subset, and then averages the estimates. We provide a sharp analysis of this average mixture algorithm, showing that under a reasonable set of conditions, the combined parameter achieves mean-squared error (MSE) that decays as O(N-1 +(N/m)-2). Whenever m ≤ √N, this guarantee matches the best possible rate achievable by a centralized algorithm having access to all N samples. The second algorithm is a novel method, based on an appropriate form of bootstrap subsampling. Requiring only a single round of communication, it has mean-squared error that decays as O(N-1 + (N/m)-3), and so is more robust to the amount of parallelization. In addition, we show that a stochastic gradient-based method attains mean-squared error decaying as O(N-1 + (N/m)-3/2), easing computation at the expense of a potentially slower MSE rate. We also provide an experimental evaluation of our methods, investigating their performance both on simulated data and on a large-scale regression problem from the internet search domain. In particular, we show that our methods can be used to efficiently solve an advertisement prediction problem from the Chinese SoSo Search Engine, which involves logistic regression with N ≈ 2.4×108 samples and d ≈ 740,000 covariates.

Improved Communication Schedules with Buffers

We consider the multimessage multicasting over the n processor complete (or fully connected) static network when there are l incoming (message) buffers on every processor. We present an efficient algorithm to route the messages for every degree d problem instance in d2/l + l - 1 total communication rounds, where d is the maximum number of messages that each processor may send (or receive). Our algorithm takes linear time with respect to the input length, i.e. O(n + q) where q is the total number of messages that all processors must receive. For l = d we present a lower bound for the total communication time. The lower bound matches the upper bound for the schedules generated by our algorithm. For convenience we assume that the network is completely connected. However, it is important to note that each communication round can be automatically translated into one communication round for processors interconnected via a replication network followed by a permutation network (e.g., two adjacent Benes networks), because in these networks all possible one-to-many communications can be performed in a single communication round.

Consistent and automatic replica regeneration

Reducing management costs and improving the availability of large-scale distributed systems require automatic replica regeneration , that is, creating new replicas in response to replica failures. A major challenge to regeneration is maintaining consistency when the replica group changes. Doing so is particularly difficult across the wide area where failure detection is complicated by network congestion and node overload.In this context, this article presents Om, the first read/write peer-to-peer, wide-area storage system that achieves high availability and manageability through online automatic regeneration while still preserving consistency guarantees. We achieve these properties through the following techniques. First, by utilizing the limited view divergence property in today's Internet and by adopting the witness model , Om is able to regenerate from any single replica, rather than requiring a majority quorum, at the cost of a small (10 −6 in our experiments) probability of violating consistency during each regeneration. As a result, Om can deliver high availability with a small number of replicas, while traditional designs would significantly increase the number of replicas. Next, we distinguish failure-free reconfigurations from failure-induced ones, enabling common reconfigurations to proceed with a single round of communication. Finally, we use a lease graph among the replicas and a two-phase write protocol to optimize for reads, so that reads in Om can be processed by any single replica. Experiments on PlanetLab show that consistent regeneration in Om completes in approximately 20 seconds.

Moshe

We present Moshe, a novel scalable group membership algorithm built specifically for use in wide area networks (WANs), which can suffer partitions. Moshe is designed with three new significant features that are important in this setting: it avoids delivering views that reflect out-of-date memberships; it requires a single round of messages in the common case; and it employs a client-server design for scalability. Furthermore, Moshe's interface supplies the hooks needed to provide clients with full virtual synchrony semantics. We have implemented Moshe on top of a network event mechanism also designed specifically for use in a WAN. In addition to specifying the properties of the algorithm and proving that this specification is met, we provide empirical results of an implementation of Moshe running over the Internet. The empirical results justify the assumptions made by our design and exhibit good performance. In particular, Moshe terminates within a single communication round over 98% of the time. The experimental results also lead to interesting observations regarding the performance of membership algorithms over the Internet.

A Virtually Synchronous Group Multicast Algorithm for WANs: Formal Approach

This paper presents a formal design for a novel group communication service targeted for wide-area networks (WANs). The service provides virtual synchrony semantics. Such semantics facilitate the design of fault tolerant distributed applications. The presented design is more suitable for WANs than previously suggested ones. In particular, it features the first algorithm to achieve virtual synchrony semantics in a single communication round. The design also employs a scalable WAN-oriented architecture: it effectively decouples the main two components of virtually synchronous group communication---group membership and reliable group multicast. The design is carried out formally and rigorously. This paper includes formal specifications of both safety and liveness properties. The algorithm is formally modeled and assertionally verified.