Abstract To date, machine learning models have been widely applied to intrusion detection system (IDS) for improving detection accuracy, where most IDS suffer from adversarial evasion attacks that may lead to data loss and user privacy leakage. Although there have been numerous solutions proposed against adversarial evasion attacks, they often neglect the relationships between different traffic and heavily relied on data labels. Therefore, this paper proposes AEDGNN, a new approach for detecting adversarial evasion attacks using graph neural network (GNN) model. On one hand, AEDGNN employs E-GraphSAGE to capture network topology in IDS for building the relationship between different inputs. On the other hand, AEDGNN utilizes deep graph infomax (DGI) to train the GNN in a self-supervised manner for maximizing mutual information between local and global representations. In addition, to clarify the practical performance of defending against traditional adversarial attacks, we implement AEDGNN and classic machine learning models based on CIC-IDS2018 benchmark dataset. The experimental results show that AEDGNN achieves significant improvements on both normal and adversarial samples compared to classic solutions. The accuracy of AEDGNN is 0.02%–1.53% higher than that of classic solutions for normal samples, and 26.04%–59.04% higher for adversarial samples.

Functional encryption (FE) which covers the notion of attribute-based encryption (ABE), is the cryptographic tool to realize fine-grained control on the accessibility of encrypted data. The traditional FE requires a central trusted authority to issue secret keys. It depends on the full-trust model, and is vulnerable to the security issue caused by key-escrow. While the registered FE (Reg-FE) achieves the zero-trust model and addresses the security issue by removing the use of central authority. It allows users to generate secret keys themselves and join the system by registering corresponding public keys to a curator. This work introduces delegated Reg-FE, which is a primitive with a new registration paradigm. It allows the registration of certain authorities that can issue secret keys for their respective classical FE sub-systems, beyond the prior work of registering plain users. Delegated Reg-FE implements a hybrid trust model within a two-level hierarchy. By redefining key escrow as a functional mechanism rather than a security concern, this model employs a zero-trust upper level which removes key-escrow, while the subsystem of each authority is locally full-trust and retains key-escrow mechanism. We construct four delegated Reg-FE schemes for functionalities that can be described as the $$2\times 2$$ combinations of linear function and policy check. Namely, Delegated Reg-IPFE, Delegated Reg-ABE, Reg-IPFE with delegated ABE, and Reg-ABE with delegated IPFE. All concrete schemes support bounded registrations and delegations, and achieve standard adaptive security under $$\textsc {MDDH}$$ assumption on prime-order bilinear group. Furthermore, these schemes only rely on black-box techniques. Technically, these schemes relies on dual-system techniques as prior registration-based works. And we devise a new “hierarchically invoked dual-system” technique on schemes which have sub-ABE delegation systems. Furthermore, we present a generic construction of Delegated Reg-FE from the combination of Reg-FE and FE. The instantiations of this generic construction demonstrate the feasibility of delegated Reg-FE, supporting arbitrary functions as well as unbounded numbers of registrations and delegations. However, this approach requires non-black-box techniques and achieves weaker semi-adaptive security without malicious registration, where the semi-adaptive means the adversary claims the challenge after seeing common reference string but before making any query. Its security relies solely on the underlying assumptions of the Reg-FE and FE components.

Graph Neural Networks (GNNs) have been widely used for graph analysis. Federated Graph Learning (FGL) is an emerging learning framework to collaboratively train graph data from various clients. Although FGL allows client data to remain localized, a malicious server can still steal client private data information through uploaded gradient. In this paper, we for the first time propose label distribution attacks (LDAs11The term “LDA” here is different from other machine learning terms like Latent Dirichlet Allocation.) on FGL that aim to infer the label distributions of the client-side data. Firstly, we observe that the effectiveness of LDA is closely related to the variance of node embeddings in GNNs. Next, we analyze the relation between them and propose a new attack named ECLDA, which significantly improves the attack effectiveness by compressing node embeddings. Then, extensive experiments on node classification and link prediction tasks across six widely used graph datasets show that EC-LDA outperforms the SOTA LDAs. Specifically, EC-LDA can achieve the Cos-sim as high as 1.0 under almost all cases. Finally, we explore the robustness of EC-LDA under differential privacy protection and discuss the potential effective defense methods to EC-LDA. Our code is available at https://github.com/cheng-t/EC-LDA.

Outsourcing personal data to a convenient and affordable cloud platform has become a popular practice. Considering the risk of privacy leakage, users usually encrypt their data before uploading it to the cloud server. Searchable encryption (SE) allows cloud servers to manage and search data in encrypted form based on user-specified requests. However, coercion attacks are rarely considered, where users may be forced to open search records and results. Therefore, deniable SE solutions against coercion attacks are presented, but they suffer from large storage overhead or fail to consider the dual coercion situation towards both sides of data owners and data users. In this paper, we roughly combine oblivious cross-tags protocol (OXT) and deniable encryption to propose a deniable SE (deniable cross-tag, DXT) scheme, which supports boolean queries and resists dual coercion attacks. Technically, we formalize a new primitive called updatable deniable encryption, and combine it with OXT in a non-trivial manner. In addition, we give formal system model, security model, and security proof of DXT. By employing the HUAWEI cloud platform, we conduct sufficient comparative experiments between DXT and state-of-the-art solutions based on a public dataset. The experimental results demonstrate that DXT outperforms higher search efficiency while achieving better features.

Federated Learning (FL) is a distributed machine learning framework that allows multiple clients to collaboratively train an intermediate model with keeping data local, however, sensitive information may be still inferred during exchanging local models. Although homomorphic encryption and multi-party computation are applied into FL solutions to mitigate such privacy risks, they lead to costly communication overhead and long training time. As a result, functional encryption (FE) is introduced into the field of privacy-preserving FL (PPFL) for boosting efficiency and enhancing security. Nevertheless, existing FE-based PPFL frameworks that support dynamic participation either required a trusted third party that may lead to single-point failure, or require multiple rounds of interaction that inevitably incur large communication overhead. Therefore, we propose PrivLDFL, a lightweight and dynamic PPFL framework for resource-constrained devices. Technically, we formalize dynamic decentralized multi-client FE and give instantiations, then present efficiency optimizations via designing a vector compression funnel based on Chinese Remainder Theorem, and finally achieve client dropouts via a client partitioning strategy. Besides formal security analysis on PrivLDFL, we implement it and state-of-the-art solutions on Raspberry Pi to conduct extensive experiments, confirming the practical performance of PrivLDFL on best-known public datasets.

Secure cloud storage is a prevalent way to provide data retrieval services, where users’ data are encrypted before uploading to the cloud. To effectively perform keyword searches over the encrypted data, the approach of searchable encryption (SE) was introduced. However, the leakage of the keyword-pair result pattern to the cloud could be exploited to reconstruct the queried keywords. To mitigate such information leakages, numerous result pattern-hiding SE systems were proposed but rarely supported data sharing with expressive queries and even owner-enforced authorization. Therefore, we present a result pattern hiding and authorized SE system (AXT) supporting conjunctive queries for cloud-based data sharing. Technically, we construct an authorized label private set intersection protocol from a refined authorized public key encryption with an equality test and then combine it with an introduced asymmetric variant of oblivious cross-tag protocol. Moreover, we introduce the system and security model of AXT along with rigorous security proof. Furthermore, we conduct comparative experiments between state-of-the-art solutions with AXT on HUAWEI Cloud platform under the widely recognized Enron dataset, which reveal that AXT achieves practical performance with retaining authorized data sharing and result pattern hiding, specifically, the time overhead for conjunctive queries with 10 keywords is reduced by 20$\%$.

In cross-device federated learning, verifiable secure aggregation enables clients to aggregate their locally trained model parameters through a malicious server to obtain a global model. To prevent the malicious server from tampering with the results, solutions have been proposed to ensure verifiability during this aggregation process. However, previous solutions either failed to achieve non-interactivity, which is critical in the cross-device setting, or failed to achieve malicious server resistance due to their underlying approach. Thus in this paper, we propose a lightweight non-interactive multi-client verifiable computation scheme (lwMVC) and then construct the malicious-resistant non-interactive verifiable aggregation (MRNIVA) for cross-device federated learning by the newly introduced underlying approach lwMVC. By combining Half Gates and non-iterative proxy oblivious-transfer, lwMVC meets both the non-interactivity and malicious resistance properties. Then, after solving the user dropout and efficiency problem, we construct the MRNIVA by lwMVC. Since we focus on the cross-device scenario, we conduct experiments on a Single Board Computer. The experiment result demonstrates that MRNIVA remains efficient compared to previous works after achieving both non-interactive and malicious resistance properties. As far as we know, we are the first to carry out verifiable aggregation experiments using such resource-constrained devices.

Secure cloud storage offers encrypted databases outsourcing service for resource-constrained clients, containing numerous tables with certain relations. Searchable symmetric encryption enables a client to search over its encrypted database on the cloud, while rarely considering queries over joins of tables. Join Cross-Tags (JXT) protocol (ASIACRYPT 2022) is thence presented that enables conjunctive queries over joins of tables, while neglecting arbitrary Boolean queries with disjunctive and conjunctive normal forms (DNF/CNF) in TWINSSE (PETS 2023). However, trivially combining JXT and TWINSSE for arbitrary DNF/CNF boolean queries over joins of tables seems infeasible due to: (i) no support for dis/conjunctive query with the same meta-keyword; (ii) returning inaccurate search results; (iii) incurring costly storage overhead. Therefore, we introduce TNT-QJ, a practical TwiN cross-Tag protocol for arbitrary boolean Query-Join over multi-tables. The result is technically obtained from revisiting TWINSSE’s framework via using s-term (the least frequent keyword) for the relation between a keyword and its meta-keyword, and non-trivially combined with JXT’s query-join approach for introducing a connective attributed in encryption tuples. In addition, we present a semi-full multi-fork searchable tree to store keyword information and reveal keyword containment relations, where the storage consumption is reduced from$\mathcal {O}(n^{3})$to$\mathcal {O}(n^{2})$. Finally, to clarify practical performance, we conduct extensive experiments on JXT and TNT-QJ using an open database in the HUAWEI cloud. Besides enabling disjunctive queries over joins of tables, TNT-QJ also runs$1.2\times $faster for conjunctive queries than JXT (with #keywords=2), which confirms rich features and practical efficiency.

Electronic healthcare (E-health) cloud system enables electronic health records (EHRs) sharing and improves efficiency of diagnosis and treatment. In order to address EHRs confidentiality and authorized user access control in E-health cloud, attribute-based proxy re-encryption (ABPRE) has been widely employed which provides dynamic fine-grained access control over encrypted EHRs. Unfortunately, existing ABPRE schemes still have the following defects: 1) capacity of attribute-universe is defined at setup; 2) verifiable mechanism for re-encryption reveals EHRs about patients; 3) traditional access policy reveals sensitive information pertaining to patients. This paper focuses on these issues and presents large-universe, verifiable and privacy-preserving dynamic fine-grained access control scheme for E-health cloud. More details, we solve limitation of attribute-universe to large-universe, which means that attributes aren’t required to be enumerated at setup. Considering disclosure of underlying EHRs in verifiable mechanism, scheme introduces non-interactive zero-knowledge proof as verifiable mechanism that supports public validation and doesn’t leak EHRs of patients. Furthermore, partially hidden policy is employed to protect privacy of patients in policy, which divides attribute into attribute name and attribute value, displaying attribute name and hiding attribute value. Finally, experimental evaluation is given that demonstrates the more comprehensive functionality of our scheme without sacrificing significant computational overhead.

Abstract Identity‐based Matchmaking Encryption (IB‐ME) is an extension notion of matchmaking encryption (CRYPTO 2019), where a sender and a receiver can specify an access policy for the other party. In IB‐ME, data encryption is performed by not only a receiver identity but also a sender's encryption key. Nevertheless, previous IB‐ME schemes have not considered the problem of efficient revocation . Hence, the authors introduce a new notion of revocable IB‐ME (RIB‐ME) and formalise the syntax and security model of RIB‐ME. In particular, the authors give an effective and simple construction of RIB‐ME in the standard model, whose security is reduced to the hardness of decisional bilinear Diffie—Hellman problem and computational Diffie—Hellman problem. In addition, the authors show two extensions of our RIB‐ME scheme to consider chosen‐ciphertext security and forward privacy.

Federated Learning (FL), emerging as a distributed machine learning, is a popular paradigm that allows multiple users to collaboratively train a intermediate model by exchanging local models without the training data leaving each user’s domain. However, FL still suffer from privacy risk such as leaking private information from users’ uploaded local models. To address the privacy concern, several approaches have been proposed to achieve privacy-preserving FL (PPFL) based on differential privacy (DP), multi-party computation (MPC), homomorphic encryption (HE) and functional encryption (FE). Compared with DP, MPC and HE, approaches based on FE is more advantageous and thus becomes the focus of this work. Moreover, all existing PPFL schemes via FE employ a multi-user extension of FE for a specific function, i.e., multi-input FE (MIFE). In this paper, we point out that existing FE-based PPFL schemes have faced with several security issues due to the misuse of MIFE. After reconsidering the security requirements of PPFL, we propose new goals of designing PPFL using FE. To achieve our goals, we propose a new FE called dual-mode decentralized multi-client FE (2DMCFE) and give a concreate construction for 2DMCFE. With 2DMCFE, we propose a new framework of PPFL where we establish a fresh 2DMCFE instance for each subset of users. Security proof shows the strong security of our framework under the semi-honest security setting. Furthermore, experiments conducted on real dataset demonstrate that our framework achieves comparable model accuracy and training efficiency to the basic FE-based scheme while providing stronger security guarantee.

Secure cloud-assisted data publish/subscribe (Pub/Sub) service provides an asynchronous method for publishers and subscribers to non-interactively exchange encrypted messages. Besides performing conjunctive subscription policy, numerous data Pub/Sub systems have recently been proposed to provide dynamic access control enforced from the publisher side to the subscriber side. However, these solutions fail to consider the following properties: (i) bilateral access control for both publishers and subscribers; (ii) the anonymity of the publisher; (iii) high matching time cost between publication and subscription. Therefore, we present P/S-BiAC, a secure and boolean cloud-assisted data Pub/Sub system with attribute-based bilateral access control that achieves authenticity and anonymity of publishers. In particular, P/S-BiAC enables cloud-based brokers to use the subscriber’s trapdoor to match published data with sub-linear time complexity. Technically, we introduce a “BiAC-and-Hidden” technique to refine publication tuples and trapdoor in classic searchable symmetric encryption solutions. Moreover, we implement P/S-BiAC and evaluate its practical performance based on Enron dataset in real cloud environment. To deal with a conjunctive subscription policy, P/S-BiAC runs 27.8× faster for matching time cost (with s-term=10) compared to state-of-the-art solutions, which demonstrates its feasibility in practical data Pub/Sub services with strong security properties.

Abstract Binary tree encryption is an intriguing primitive that enables many practical applications to achieve an increasing important security feature, forward security. However, the public key size of existing constructions grows linearly with the depth of the underlying binary tree in the standard model. To support more secret keys associated with nodes, it is often expected that the tree has a sufficiently large depth. This places a burden on employing it implicitly or explicitly in real world. In this work, we show how to compress linear-size public key down to constant-size public key and give our construction featuring constant-size public key in the standard model. We prove that our construction achieves an improved security, adaptive security, under the matrix decision Diffie–Hellman assumption, which is a generalization of standard $k$-Lin assumption. Moreover, our key-generation, key-derivation and encryption algorithms have lower time complexity than that of the prior construction, leading to further efficiency improvements. To illustrate these improvements in practice, we give an implementation of our construction and the prior one and then evaluate the performance in the tree depth.

In cloud computing, computationally weak users are always willing to outsource costly computations to a cloud, and at the same time they need to check the correctness of the result provided by the cloud. Such activities motivate the occurrence of verifiable computation (VC). Recently, Parno, Raykova and Vaikuntanathan showed any VC protocol can be constructed from an attribute-based encryption (ABE) scheme for a same class of functions. In this paper, we propose two practical and efficient semi-adaptively secure key-policy attribute-based encryption (KP-ABE) schemes with constant-size ciphertexts. The semi-adaptive security requires that the adversary designates the challenge attribute set after it receives public parameters but before it issues any secret key query, which is stronger than selective security guarantee. Our first construction deals with small universe while the second one supports large universe. Both constructions employ the technique underlying the prime-order instantiation of nested dual system groups, which are based on the $d$-linear assumption including SXDH and DLIN assumptions. In order to evaluate the performance, we implement our ABE schemes using $\textsf{Python}$ language in Charm. Compared with previous KP-ABE schemes with constant-size ciphertexts, our constructions achieve shorter ciphertext and secret key sizes, and require low computation costs, especially under the SXDH assumption.

The paper presents an identity based encryption (IBE) under selective opening attacks (SOA) whose security is almost-tightly related to a set of computational assumptions in composite-order bilinear groups. Our result is a combination of Bellare, Waters and Yilek's method [TCC, 2011] for constructing (not tightly) SOA secure IBE and Hofheinz, Koch and Striecks’ technique [PKC, 2015] on building almost-tightly secure IBE in the multi-ciphertext setting. In the paper, we first tune Bellare et al.’s generic construction for SOA secure IBE to show that a one-bit IBE achieving ciphertext indistinguishability under chosen plaintext attack in the multi-ciphertext setting (with one-sided public openability) tightly implies a multi-bit IBE secure under the selective opening attack. Next, we almost tightly reduce such a one-bit IBE to static assumptions in the composite-order bilinear groups employing the technique of Hofheinz et al. This yields the first SOA secure IBE with almost-tight reduction.

ABSTRACT Anonymous password authenticated key exchange (APAKE) protocols allow the server to authenticate its clients without revealing their identities. In this paper, we first construct a basic protocol SAPAKE by using the homomorphic encryption scheme and an auxiliary memory device. Compared with the previous ones, SAPAKE is more suitable for those privacy‐sensitive applications (e.g., cloud computing) where reducing server payload and improving user experience are both essential. Furthermore, we refine SAPAKE by removing the use of the memory device to gain an enhanced extension SAPAKE+ without increasing the resources consumption. SAPAKE+ achieves better user‐friendliness than SAPAKE while it requires publishing more public parameters. Both of our protocols are practical due to their low (computation and communication) resources consumption and better user‐friendliness, and achieve provable security in the random oracle model. Copyright © 2012 John Wiley & Sons, Ltd.