Federated Learning (FL) faces significant challenges arising from both data and system heterogeneity. While Clustered Federated Learning (CFL) mitigates data heterogeneity by grouping clients with similar data distributions, it remains vulnerable to system heterogeneity, which can slow convergence due to performance disparities among clients. Moreover, data drift may degrade clustering accuracy and training efficiency over time. In this work, we propose a Model Structure-aware Clustered Federated Learning (MSCFL) framework that simultaneously addresses the issues of data heterogeneity, system heterogeneity, and data drift. MSCFL incorporates model pruning (MP) into the CFL framework to enhance training efficiency under system heterogeneity. To enable this integration, we address the key challenge of performing effective clustering based on heterogeneous, pruned local models with varying structures. To this end, we design a model structure-based similarity computation algorithm to integrate CFL with MP. To effectively address data drift, we propose a dynamic cluster migration strategy that efficiently monitors model structures via Hamming Distance and triggers re-clustering only when necessary. Extensive experimental results show that MSCFL improves the accuracy and convergence speed of cluster models, outperforming traditional CFL in various settings.

Friends query is one of the critical components in online social networks (OSNs). However, with the advent of OSNs, more and more private information of individuals (e.g., personal sensitive profiles, relationships and locations) will be collected by the service providers during the query. Some existing privacy-preserving solutions either can provide only approximate queries or need trust assumption on the service providers. To overcome these limitations, we propose PPFQ-a Privacy-Preserving Friends Query scheme over OSNs. To the best of our knowledge, this is the first work that aims at preventing attack launched by the service provider and attackers, but still allow users to make friends query based on profile and geo-distance. In PPFQ, we design a profile-based friends query protocol where user's profiles can be encrypted and searched by his friends privately with different keys. In addition, for the first time, we design a k-nearest friends query protocol to achieve location-based search on encrypted domain. By deploying the two-server secure distance computation and comparison mechanism with Batcher's sorting network, this protocol achieves accurate nearby friends retrieving while preventing the privacy of geo-location and the distance order from revealing them to the servers. We also give a comprehensive analysis of the correctness and security of PPFQ with mathematical proofs. Finally, we provide extensive experimental results to demonstrate the efficiency of our proposed construction.

Abstract Vector commitment (VC) schemes allow committing to an ordered sequence of ${q}$ values ${(m_1,\cdots ,m_q)}$ in such a way that one can later open the commitment at specific positions. However, the existing VC schemes suffer from two substantial shortcomings that limit their use: (i) the commitments cannot be opened except at some specific positions, and (ii) their security only captures position-binding but offers no privacy: the client may learn additional information about the committed sequence through the proofs and the commitments. To resolve these problems, we first extend VC to a more expressive primitive called VC with sum binding (VCS), in which the commitment can also be opened to the sum of all elements in the committed sequence. VCS additionally satisfies the security of sum binding, which guarantees that the commitment cannot be opened to different sums. To enhance its privacy, we extend VCS to zero-knowledge VCS (ZKVCS), in which commitments and proofs constructed during the protocol execution leak nothing about the committed sequence. We formalize this new property by a standard real/ideal experiment. Meanwhile, the detailed performance analyses and simulations show that our proposed schemes are more practical. Finally, we introduce a novel notion of (zero-knowledge) verifiable database supporting sum and show how to construct it from our (ZK)VCS scheme.

Summary Since cloud computing and cloud storage have become a common practice in our daily life, people are paying more attention on protecting the privacy of their sensitive data. In this paper, we consider the problem of graph encryption in such a way that one can perform all‐path queries over an encrypted graph in a privacy‐preserving manner. A solution for such problem has many potential applications in network virtualization since graph data structures are commonly used to represent the topology of substrate and virtual networks. We propose a searchable symmetric encryption scheme for graph data that support all‐path queries. The scheme allows a client to encrypt a sensitive graph and outsource it to an untrusted server for storage. After that, the client can perform arbitrary all‐path queries between a source and a destination. For each query, the server computes and returns all the paths that satisfied the query. We prove that our scheme is secure against adaptive chosen‐query attacks in the semi‐honest setting.

Along with the fast development of wireless technologies, smart devices have become an integral part of our daily life. Authentication is one of the most common and effective methods for these smart devices to prevent unauthorized access. Moreover, smart devices tend to have limited computing power, and they may possess sensitive data. In this paper, we investigate performing graph operations in a privacy-preserving manner, which can be used for anonymous authentication for smart devices. We propose two protocols that allow two parties to jointly compute the intersection and union of their private graphs. Our protocols utilize homomorphic encryption to prevent information leakage during the process, and we provide security proofs of the protocols in the semihonest setting. At last, we implement and evaluate the efficiency of our protocols through experiments on real-world graph data.

Private set intersection (PSI) allows two parties to compute the intersection of their private sets while revealing nothing except the intersection. With the development of fog computing, the need has arisen to delegate PSI on outsourced datasets to the fog. However, the existing PSI schemes are based on either fully homomorphic encryption (FHE) or pairing computation. To the best of our knowledge, FHE and pairing operations consume a huge amount of computational resource. It is therefore an untenable scenario for resource-limited clients to carry out these operations. Furthermore, these PSI schemes cannot be applied to fog computing due to some inherent problems such as unacceptable latency and lack of mobility support. To resolve this problem, we first propose a novel primitive called “faster fog-aided private set intersection with integrity preserving”, where the fog conducts delegated intersection operations over encrypted data without the decryption capacity. One of our technical highlights is to reduce the computation cost greatly by eliminating the FHE and pairing computation. Then we present a concrete construction and prove its security required under some cryptographic assumptions. Finally, we make a detailed theoretical analysis and simulation, and compare the results with those of the state-of-the-art schemes in two respects: communication overhead and computation overhead. The theoretical analysis and simulation show that our scheme is more efficient and practical.

Summary The intersection operation on multisets has many applications in different scenarios, such as data mining and pattern matching. This motivates us to study the problem that when users outsource their private encrypted sets and delegate the set‐intersection operation on the corresponding plaintexts to the cloud, the cloud can manage to conduct the operation and make the result verifiable without the decryption capability. We formally introduced a model of Verifiable Intersection Operations on Encrypted Data in Cloud and presented the definitions of the correctness and security properties. We also proposed a concrete scheme basing on the bilinear‐map accumulator and re‐encryption. The scheme was proved secure under some cryptographic assumptions. And the experimental results of the algorithm implementation show the scheme is partly practical for real scenarios. Copyright © 2016 John Wiley & Sons, Ltd.

In 2015, Wang proposed the first identity-based provable data possession scheme named ID-DPDP (IEEE Transactions on Services Computing, vol. 8, no. 2, pp. 328-340, Mar./Apr. 2015) to verify outsourced data publicly without the implementation of PKI. Unfortunately, in this letter, we demonstrate that this scheme is insecure in the sense that cloud servers can generate valid proofs without possessing the original data blocks. We also show another security issue in this scheme which leads to some data blocks can never be verified unless all the data blocks are challenged. Meanwhile, we provide solutions to these problems while preserving the security features of the original scheme.