Recently, the research on fairness in recommendation systems has garnered widespread attention. Moreover, numerous fair recommendation models have been developed for scenarios with limited sensitive information, thereby alleviating the issue of missing sensitive information. However, the performance of these methods still tends to decline significantly when sensitive attributes are extremely scarce. In this paper, we propose FairCL, a novel fair recommendation framework designed to perform effectively under limited sensitive attribute information. FairCL features a contrastive learning-based sensitive attribute encoder that can be integrated with existing fair recommendation algorithms. By leveraging both collaborative information and item side information, we predict unknown sensitive attributes and apply contrastive learning for sensitive attribute modeling. Furthermore, we theoretically demonstrate how FairCL can be integrated with mutual information-based and adversarial learning-based fairness algorithms. Extensive experiments on three real-world datasets show that FairCL significantly enhances fairness, even when only a small portion of users' sensitive attributes are known. The code and data are at: https://anonymous.4open.science/r/CL-for-FairRec-000A/.

Edge-assisted Hierarchical Federated Learning (EHFL) accelerates global model training across mobile devices by hierarchically aggregating models. However, EHFL encounters critical challenges such as privacy risks for local and edge-level models, vulnerability to collusive Byzantine attacks, and issues with model diversity and heterogeneity due to Non-Independent and Identically Distributed (Non-IID) data. In this paper, we propose PPBR, a novel hybrid scheme that subtly integrates Condensed Local Differential Privacy (CLDP) and Packed Linearly Homomorphic Encryption (PLHE) to achieve strong privacy protection and resilience against various Byzantine attacks in Non-IID data scenarios. Specifically, PPBR clusters the sign statistics of local models and clips the norms of edge-level momenta to filter anomalous models and mitigate Byzantine faults while retaining diverse models coming from Non-IID data. To enhance privacy protection with acceptable accuracy loss, the sign tuples of local models are perturbed with CLDP guarantees, and the momenta of edge-level models are encrypted under PLHE. Meanwhile, PPBR enhances privacy in single-edge-server and single-cloud-server aggregations by using random perturbations, secret sharing, and PLHE. In addition to safeguarding privacy with accommodating abrupt dropouts of mobile devices and edge servers, the aggregations effectively mitigate the adverse effects of Non-IID data under advanced Byzantine attacks. Theoretical analysis and comprehensive experiments validate PPBR's strong privacy guarantees and resilience to various Byzantine attacks under Non-IID data.

Multi-dimensional range queries performed on the mobile user data records become increasingly important and popular in the fields of e-commerce, social media, transportation logistics, etc. Meanwhile, mobile users usually have different privacy requirements for different attributes of the records. A straightforward and effective approach is to first get low-dimensional range query outcomes by using existing LDP mechanisms at different privacy levels, and then derive high-dimensional range query results at each level, and finally aggregate the results from all levels. However, it incurs low utility of the query results, since the non-fixed privacy budgets and the correlation between dimensions (attributes) detrimentally impact the utility of LDP methods, ultimately rendering them ineffective in practice. In this paper, we propose a new Personalized LDP approach for Multi-dimensional Range queries (PLDP-MR) over mobile user data, consisting of the user grouping, data perturbing, data re-perturbing, and range query results aggregating steps. First, PLDP-MR offers flexible dual grouping based on user-selected privacy levels and relevant attributes to obtain the corresponding one-dimensional and two-dimensional grids. PLDP-MR optimizes the grid granularity to minimize errors from perturbing users' attribute data with different LDP noises at non-fixed privacy levels. Furthermore, PLDP-MR carefully re-perturbs the LDP-noisy data from mobile users at lower privacy levels (i.e., having the higher utility) to achieve LDP with higher privacy levels and supplement the data volume of the corresponding groups. Thus, the data utility is effectively improved without additional privacy losses. Finally, PLDP-MR aggregates the frequencies in all the one-dimensional and two-dimensional grids related to the multi-dimensional range query at all query intervals and all privacy levels to derive the final query result with considering the correlation between attributes. The aggregations use maximum entropy optimization and maximum likelihood methods to further enhance its utility. The privacy and utility of PLDP-MR are analyzed, and extensive experiments demonstrate its effectiveness.

Personalized federated learning (PFL) is a specialized application of the federated learning paradigm designed to support personalized use cases. Unlike traditional federated learning, which aims to train a high-quality global model, the goal of PFL is to tailor a model that best fits each individual user. Most existing PFL approaches adopt training architectures similar to those used in traditional federated learning, relying on global or partial model sharing during training. While this helps improve model personalization across clients, it also introduces a range of challenges, including risks of data leakage and increased communication overhead. To address these challenges, we propose a novel personalized federated learning (PFL) framework called SuperMPFL, which leverages supermasks to effectively tackle issues related to accuracy, privacy, and efficiency. In particular, the SuperMPFL technique utilizes masking and ranking strategies to obscure the true gradient information. By converting gradients into ranked numerical representations, this approach enhances privacy protection during the training process. Furthermore, this approach reduces communication overhead by transmitting significantly less information compared to conventional methods. In SuperMPFL, each client receives the global model and then emphasizes its personalized parameters, particularly at the model’s edges. This design not only improves accuracy but also strengthens robustness against privacy attacks. Evaluations on standard federated learning benchmarks demonstrate the superiority of our approach, which outperforms state-of-the-art methods in terms of accuracy, privacy, and efficiency.

AI-empowered edge computing has given rise to a new paradigm and effectively facilitated the promotion and development of multimedia applications. The speech assistant is one of the significant services provided by multimedia applications, which aims to offer intelligent interactive experiences between humans and machines. However, malicious attackers may exploit spoofed speeches to deceive speech assistants, posing great challenges to the security of multimedia applications. The limited resources of multimedia terminal devices hinder their ability to effectively load speech spoofing detection models. Furthermore, processing and analyzing speech in the cloud can result in poor real-time performance and potential privacy risks. Existing speech spoofing detection methods rely heavily on annotated data and exhibit poor generalization capabilities for unseen spoofed speeches. To address these challenges, this article first proposes the Coordinate Attention Network (CA2Net) that consists of coordinate attention blocks and Res2Net blocks. CA2Net can simultaneously extract temporal and spectral speech feature information and represent multi-scale speech features at a granularity level. Besides, a contrastive learning-based speech spoofing detection framework named GEMINI is proposed. GEMINI can be effectively deployed on edge nodes and autonomously learn speech features with strong generalization capabilities. GEMINI first performs data augmentation on speech signals and extracts conventional acoustic features to enhance the feature robustness. Subsequently, GEMINI utilizes the proposed CA2Net to further explore the discriminative speech features. Then, a tensor-based multi-attention comparison model is employed to maximize the consistency between speech contexts. GEMINI continuously updates CA2Net with contrastive learning, which enables CA2Net to effectively represent speech signals and accurately detect spoofed speeches. Extensive experiments on the ASVspoof2019 dataset show that GEMINI reduces the Equal Error Rate and tandem Detection Cost Function by up to 96.75% and 96.35% in the physical access scenario, and by up to 86.62% and 87.71% in the logical access scenario compared to peer methods.

The surging development of Artificial Intelligence-Generated Content (AIGC) marks a transformative era of the content creation and production. Edge servers promise attractive benefits, e.g., reduced service delay and backhaul traffic load, for hosting AIGC services compared to cloud-based solutions. However, the scarcity of available resources on the edge pose significant challenges in deploying generative AI models. In this paper, by characterizing the resource and delay demands of typical generative AI models, we find that the consumption of storage and GPU memory, as well as the model switching delay represented by I/O delay during the preloading phase, are significant and vary across models. These multidimensional coupling factors render it difficult to make efficient edge model deployment decisions. Hence, we present a collaborative edge-cloud framework aiming to properly manage generative AI model deployment on the edge. Specifically, we formulate edge model deployment problem considering heterogeneous features of models as an optimization problem, and propose a model-level decision selection algorithm to solve it. It enables pooled resource sharing and optimizes the trade-off between resource consumption and delay in edge generative AI model deployment. Simulation results validate the efficacy of the proposed algorithm compared with baselines, demonstrating its potential to reduce overall costs by providing feature-aware model deployment decisions.

In today’s interconnected world, as the scale of data sharing across diverse systems and organizations continues to grow, the secure and verifiable access to these data is of escalating importance. Nonetheless, existing verification methods for prevalent access control models, such as Role-Based Access Control (RBAC), usually face several prevalent challenges, including inadequate support for cross-domain operations, potential systemic risks stemming from potential flaws in policy design and improper implementation in large-scale complex systems, and concerns about modeling credibility. To address the issues, we propose AcVerifier, a novel solution that employs blockchain technology coupled with deterministic finite automaton (DFA) to enable verifiable enforcement of access control policies across disparate domains. AcVerifier uploads data indices and policies to a secure authorization server, which records operation logs in blockchain-connected data containers, thereby ensuring data credibility. AcVerifier employs a hybrid verification method combining static and dynamic analysis techniques, leveraging their respective strengths for preemptive auditing and real-time monitoring. Thus, AcVerifier can verify the correctness and consistency between the permissions granted by the extended access control policies and their actual execution in cross-domain data sharing and ubiquitous circulation scenarios. Here, the permissions encompass desensitization, access, modification and forwarding of personal information. Comprehensive evaluation demonstrates the correctness and efficiency of AcVerifier in addressing the challenges of cross-domain access control verification.

Artificial Intelligence of Things (AIoT) is an emerging paradigm that integrates artificial intelligence (AI) and Internet of Things (IoT) technologies to provide intelligent IoT solutions. The AIoT system acquires data in real time through IoT sensors, performs intelligent data analysis tasks anywhere in the terminal–edge–cloud continuum, and provides accurate decision-making services based on data predictions. Cloud–edge collaboration can reduce security risks for AIoT data prediction by sharing data features instead of raw data. However, sensitive user data may still be inferred by attackers through model parameter analysis, causing irreparable harm and serious consequences. Therefore, data prediction based on cloud–edge collaboration while maintaining privacy constraints remains a significant challenge. In this article, a differentially private federated tensor completion method is proposed for cloud–edge collaborative AIoT data prediction. This method embeds differential privacy (DP) mechanisms with cloud–edge collaboration. Each edge is capable of processing and analyzing data, and collaborative learning with other edges by sharing privacy-preserving model parameters. For model security, objective perturbation is applied to ensure that the tensor completion method satisfies DP. To achieve higher accuracy, parallel tensor decomposition is introduced to avoid the update conflicts problem of federated tensor completion. Through theoretical analysis, our method can provide data protection for tensor completion with high-security promise. The experiments are performed on both synthetic and real-world data sets to demonstrate the superior performance of our method in preserving data privacy.

In this paper, we propose a novel verifiable and privacy-preserving ad exchange scheme (VPAE) for smart targeted advertising that enables an ad exchange to deliver promotional advertisements to Internet users according to their specific interests, traits, and preferences. VPAE achieves the distinguished feature to preserve private information of both user profiles and advertisers, while allowing users to verify why they receive specific advertisements for adverting transparency. By utilizing homomorphic proxy re-encryption, VPAE addresses the challenge that the ad exchange has the ability to select advertisements for users, but they cannot know more than what they should. Meanwhile, VPAE integrates polynomial evaluations to achieve that the user is capable of verifying why she receives the advertisement, without learning any information of other advertisements that she does not receive. The privacy of both user profiles and advertisers are protected without sacrificing the efficiency of profile matching. We show the computational and communication overhead of VPAE through extensive analysis to demonstrate practicality and resource needs in implementation.1

E-petition has played an important role in health and politics that collects public opinions and requests a superior or an authority to take actions towards a health or political problem. However, this activity exposes the privacy of the signers who participate to express opinions. In this paper, we propose a privacy-preserving fine-grained e-petition system that supports attribute-based identity verification for signers, while protecting their privacy. By considering the target groups of signers in a specific health or political petition, an attribute policy is defined to ensure that only the signers with the attributes that satisfy the attribute policy can sign the petition. The fine-grained petition is better than the traditional e-petitions because it can improve the trustworthiness of the petition results via proactive signer selection. Moreover, the new petition system protects the identities of the signers by using the non-interactive zero-knowledge proof system, such that the signers are anonymous in signing petitions. In addition, the proposed petition system supports the tracing of double-signing, a cheating behavior that an anonymous signer can submit more than one signature in a petition without being detected. Finally, we prove that the proposed petition system achieves the desirable security properties, including anonymity, unforgeability, and traceability, and demonstrate that the system is efficient to be implemented on the mobile devices.

Detecting malicious URLs is of great significance to reduce cyber crimes and maintain Internet security. Currently, Deep Learning (DL) techniques have been widely used to improve the classical malicious URL detection models, as DL-based detection models can perform an in-depth analysis of the text information of the URL, and detect the fishing URLs of unknown cyber attack types with high accuracy. Any missed blocking of malicious URLs can potentially result in a huge loss of information and property. In this paper, we focus on the vulnerability of the existing DL-based malicious URL detection models and show that they are sensitive to adversarial samples. First, we construct URL adversarial samples based on the component-level and character-level perturbations and use them to attack mainstream DL-based detection models, resulting in obvious decreases in the detection accuracies. Meanwhile, the perturbations are under the constraints that each adversarial sample URL is hardly distinguished from the original URL with naked eyes. Furthermore, under most circumstances, the adversarial samples constructed by replacing 14 types of characters and perturbing other all components except the scheme component lead to the largest increased number of missed blocking of malicious URLs, i.e., a bigger drop in the accuracy than other constructed methods. Finally, extensive experiments demonstrate the effectiveness of our adversarial examples. Even if the adversarial training is used against our adversarial samples, the adversarial samples still work and bring oblivious decreases in their accuracy.

Federated learning (FL) enables multiple clients to collaboratively train an accurate global model while protecting clients' data privacy. However, FL is susceptible to Byzantine attacks from malicious participants. Although the problem has gained significant attention, existing defenses have several flaws: the server irrationally chooses malicious clients for aggregation even after they have been detected in previous rounds; the defenses perform ineffectively against sybil attacks or in the heterogeneous data setting. To overcome these issues, we propose MAB-RFL, a new method for robust aggregation in FL. By modelling the client selection as an extended multi-armed bandit (MAB) problem, we propose an adaptive client selection strategy to choose honest clients that are more likely to contribute high-quality updates. We then propose two approaches to identify malicious updates from sybil and non-sybil attacks, based on which rewards for each client selection decision can be accurately evaluated to discourage malicious behaviors. MAB-RFL achieves a satisfying balance between exploration and exploitation on the potential benign clients. Extensive experimental results show that MAB-RFL outperforms existing defenses in three attack scenarios under different percentages of attackers.

Due to its powerful feature learning capability and high efficiency, deep hashing has achieved great success in large-scale image retrieval. Meanwhile, extensive works have demonstrated that deep neural networks (DNNs) are susceptible to adversarial examples, and exploring adversarial attack against deep hashing has attracted many research efforts. Nevertheless, backdoor attack, another famous threat to DNNs, has not been studied for deep hashing yet. Although various backdoor attacks have been proposed in the field of image classification, existing approaches failed to realize a truly imperceptive backdoor attack that enjoys invisible triggers and clean label setting simultaneously, and they cannot meet the intrinsic demand of image retrieval backdoor.

Federated Learning (FL) has attracted significant interest, as it provides a distributed machine learning paradigm to share data resources during model training process. However, sharing the gradients or model weights uploaded by clients or the final model aggregated by the server can lead to privacy disclosures and executing correctness issues. Specifically, the original data can be easily inferred through analyzing the shared gradients, and malicious users can disrupt the model aggregation to result in a destruction of the model accuracy. To address these issues, we propose a novel FL scheme with providing both privacy protection and robust aggregation. By using the distributed differential privacy and range proof technologies, the proposed scheme resists semi-honest servers and malicious users, while protecting the global model and providing the high accuracy. Both privacy analysis and experiments are given to demonstrate the effectiveness of our scheme.

Private Set Intersection (PSI) is typically used to achieve ID alignment with protection of IDs in the preparation phase of Vertical Federated Learning (VFL). However, existing PSI approaches are limited to protecting IDs that are outside the intersection of participants, and most ignore the sensitivity of intersection for a weak party in an asymmetrical ID alignment. Since the set size of the strong party is much greater than the weak party’s in an asymmetrical federation, and the intersection usually accounts for a substantial part of the weak party set, the weak party’s sensitive sample IDs would be severely compromised through sharing the intersection. To address this issue, we propose Differentially private PSI Cardinality and PSI (DPSI-CA, DPSI) protocols, which protect the intersection cardinality and sensitive IDs inside the intersect ion for the weak party, respectively. First, DPSI-CA encodes IDs in binary notation, and combines them with the GM encryption, to perform the ID-matchmaking by executing bitwise plaintext XOR. Then, the encrypted matching results are independently perturbed using randomized responses to produce differentially private outputs for PSI-CA, and its unbiased estimate is added to remove the deviation brought by the randomization. Furthermore, DPSI fuses Pseudo-Random Function (PRF)-based zero sharing, garbled Bloom filter, and Oblivious PRF (OPRF)-based shares reconstruction, to successfully reconstruct the shares corresponding to sampled IDs in the intersection. Meanwhile, a randomized response is used to sample the inputs and perturb the outputs of the OPRF-based shares reconstruction, producing a randomly sampled intersection for the weak party and differentially private intersection for the strong party. Finally, the privacy analysis shows that our protocols provide differential privacy for the weak party’s sensitive sample IDs, and extensive experiment results illustrate the feasibility of the asymmetrical ID alignment involving millions of IDs.

In intelligent transportation systems, the key issue of the Ride-Sharing Service (RSS) is to find proper drivers for the passengers by Intelligent Matching (IM) of two or three objects, including the positions of drivers, the travel information of passengers, and the spots where passengers and drivers meet and separate. Unfortunately, the exposure of travel plans of passengers in the IM process due to inference attacks has raised concerns about the privacy violation. To resist the inference attacks, we propose a Differentially Private Tripartite IM (DPTIM) protocol for RSS. DPTIM is based on the tripartite IM process, which intelligently finds the suitable threshold to filter out the matched objects with satisfaction scores below the threshold, so as to provide the high average satisfaction score of matched passengers. Compared to existing relevant mechanisms, DPTIM is distinguished by the feature that it leverages the inference error and differential privacy techniques to prevent the prior-information-based inference attacks and constrain the posterior information leakage, while providing satisfactory matching results. Furthermore, DPTIM meets the personalized demand of location privacy by using the passenger-specific tolerance estimation on inference errors and the personalized privacy budget. Finally, we implement DPTIM on real-world datasets, and demonstrate the satisfactory performance of DPTIM in terms of the average satisfaction score of passengers, the anti-inference-attack capability, and the passenger-specific privacy requirement.

Location privacy continues to attract significant attentions from both industry and academia in recent years. However, Location Based Service (LBS) servers or some other adversaries who can monitor a particular user's historical and current status in a long-term way may likely infer user's location privacy. To solve this problem, we propose a Longterm Observation-aware Dummy Selection (LODS) algorithm to achieve k-anonymity for users in LBSs. Different from existing approaches, the LODS takes the historical anonymity sets into account, since mobile users may query LBSs at certain places such as home or office. LODS selects candidate sets containing dummy locations with less number of occurrences firstly, in order to achieve the preferred distribution. Then, LODS further filters out candidate sets with smaller entropy. Finally, we choose the anonymity set with highest Quality of Service (QoS) as the result. Extensive experiment indicates our algorithm can protect user's location privacy effectively against long-term observation, and satisfy user's QoS requirement at the same time.

Customized-bus Sharing Service (CSS) enables a centralized server to schedule comfortable bus trips for users by ride clustering based on the individual requirements. It has been increasingly popular in crowded metropolises, bringing a lot of convenience and reducing trip costs to users. Ride clustering is essential for the server to determine the stops of a customized bus, but it also leads to the exposure of users' current locations and spatio-temporal patterns. Although privacy-preserving ride clustering can generate optimal bus routes, it depends on frequent interactions between users and the server, so all the users should be always online. In this paper, we propose a privacy-preserving ride clustering scheme for CSS to support off-line users, in which fog computing is introduced to assist the server in generating bus route without the exposure of users' travel plans. Fog servers are able to perform ride clustering interacting with the server, after receiving the preferred pick-up and drop-off positions from users. Thus, the users are unnecessary to be always online. In addition, the Paillier cryptosystem and randomization technique are leveraged to protect the user's privacy without sacrificing the clustering quality. Finally, the proposed privacy-preserving ride clustering scheme is demonstrated to have the advantage of low computational and communication overhead with high security guarantees.

Mobile Social Networks (MSNs) provide a variety of social networking applications in mobile environment, where social group finds and recruits potential members easily. Unfortunately, users enjoy these conveniences at the cost of revealing their personal data. Additionally, people usually ignore a critical factor, Impact Factor (IF), which is used to quantify group members' influence on their groups, since a group member with larger IF generally has a greater influence on potential new member recommendation. In this paper, we propose IF-RG, an IF-based group recommendation scheme with privacy preservation in MSNs. First, we construct a transmission matrix and exploit PageRank algorithm to compute and update group members' IFs. The average variation of IF is formed to measure convergence speed of the iteration method of computing IF. To make sure that the larger IFs, the more influence, IF, Ochiai similarity function and weighted majority rule are integrated in the novel matching degree between stranger and group. The fuzzy matrix algorithm not only protects users' privacy, but also helps our scheme to support group recommendation when not every one in the groups is online. Finally, security and performance are analyzed and evaluated via detailed simulations.

Friend discovery has been one of the hot topics in our social activities over the past decade. Mobile users have more opportunities to discover and make new social interactions with others in vicinity to build and extend their social communities. However, the inevitable information releasing conflicts with the increasing privacy concerns. In this paper, we employ the concept of friend-of-friend and design a secure and accurate friend discovery for privacy-aware mobile users in Proximity-based Mobile Social Networks (PMSNs). We first construct a novel similarity function with fully considering the number of common attributes, the corresponding priorities and the ratio of matched attributes over all the inputs. Then we develop a secure friend recommendation phase based on a carefully combination of the commutative encryption function and the bilinear pairings. The security and performance are thoroughly analyzed and evaluated via detailed simulations.

Privacy issues in group matching problem have become one of the most important things in Mobile Social Networks (MSNs) currently. Mobile users may feel uncomfortable when releasing personal information to some irrelevant people or groups. In this paper, we propose a practical group matching scheme without employing any Trusted Third Party (TTP) for privacy-aware users in MSNs. We first propose a fuzzy matrix algorithm to generate user's authority instead of complex cryptographic computations to reduce the communication and computation overhead, and thus build a public set to store all the group members' profiles and authorities. As a result, our group matching does not need all the group members are online anymore. Moreover, we utilize the Ochiai similarity considering both of the number of common attributes and the size of each user's profile. The privacy and performance are analyzed and evaluated via detailed simulations.

Privacy issues in Location-Based Services (LBSs) have gained tremendous attentions in literature over recent years. Existing approaches always fail to provide dual privacy protection on both user's location and point of interest (POI), incur endurable system overhead, and produce high quality of services, simultaneously. To address these problems, we propose a time obfuscation-based scheme, termed TOP-privacy, which carefully generates and sends some dummy queries at leisure time to confuse adversaries with some background information. TOP-privacy employs a dummy query generation algorithm, which includes a dummy location selection module and a classified POI pool construction module. It first selects some location candidates with similar location distribution with the real user's, and then determines the optimal POI to construct the dummy query based on the similarity to the user's real POI. Security analysis and evaluation results indicate its effectiveness and efficiency.