In recent years, Cross-Modal Retrieval (CMR), which can retrieve data across types based on query semantics, has become an attractive technology due to the widespread applications of multimedia data. Outsourcing multimedia data to a cloud server is a reliable way to improve the quality of CMR services, but it will also incur potential data privacy leakage issues. Existing schemes for privacy-preserving outsourced data search services are either inapplicable to CMR or limited by efficiency and scalability. To address the above issues, we investigate the problem of Searchable Symmetric Encryption (SSE) for CMR in this paper. Firstly, we formulate the definition of SSE for CMR (namely, SSECMR) and extend the SSE leakage functions to capture the leakage in SSECMR. Then, by constructing distance-computation-free Hamming inverted multi-index, we propose a practical SSECMRconstruction. Specifically, we transform Hamming distance-based range queries into multi-key queries, thereby avoiding computationally expensive comparison operations and enabling efficient Hamming distance queries over encrypted data. Our design supports secure CMR with sub-linear complexity in a single communication roundtrip. Through rigorous security analysis, we demonstrate that our construction can provide adaptive security. Empirical evaluations on real-world datasets demonstrate that SSECMRoutperforms the state-of-the-art scheme in both efficiency and accuracy, and is comparable to plaintext applications. Our code is available at https://github.com/SSE-CMR/SSECMR.

Unmanned aerial vehicle (UAV)-aided mobile edge computing (MEC) with a blockchain consensus algorithm is a promising solution for addressing mobile offloading of computation-intensive or latency-sensitive tasks with extensive service range, while ensuring the authenticity of the offloading. Existing UAV-aided MEC studies focus on task offloading or trajectory planning of single-antenna UAVs, missing multiple-input multiple-output (MIMO) UAV systems. In addition, most blockchain-based UAV-aided edge computing studies adopt energy-consuming proof-of-work-based consensus mechanisms, which are unsuitable for energy-limited UAV scenarios. To address the above issues, in this paper, we develop a joint deadline-aware task offloading and UAV trajectory planning scheme, utilizing the effects simulated through an energy-efficient consensus protocol. In order to prevent the exhaustion of UAVs' energy and ensure the authenticity of decision-making, we propose an energy-based Raft (E-Raft) consensus algorithm, enabling dynamic leader selection through a series of decreasing energy thresholds. Subsequently, we present a computational profit maximization problem to jointly optimize deadline-aware task offloading and the UAV swarm's trajectory planning. To address this NP-hard problem, we develop an online priority-based task offloading and trajectory selection algorithm, which is performed by the selected leader of the E-Raft consensus in each round. Simulation results demonstrate that our proposed scheme achieves up to 40% performance improvement over other approaches.

Secure Location-Based Services (LBSs) in mobile cloud have gained widespread attention in the past decade. However, previous works mainly focus on spatial or spatial keyword query services and cannot support temporal filters simultaneously, which limits the service quality in practical applications. To address the above issue, we propose an efficient Keyword-associated Spatio-Temporal Structured Encryption (KSTSE) scheme that allows conjunctive queries according to spatio-temporal range and textual keywords on encrypted data in a mobile cloud. Specifically, we first transform geographic and temporal range queries into a unified encoding existence detection problem by combining S2 encoding and prefix encoding. Next, we build an efficient encrypted existence evaluation construction based on the circular shift and coalesce Bloom filter and symmetric hidden vector encryption. Finally, to support efficient queries on large-scale datasets, we design a hierarchical index tree structure, which can dynamically prune the search space according to the keyword and spatio-temporal range during the query process, reducing the query complexity toO(logN) Rigorous security analysis and performance evaluation show that the proposed KSTSE construction is adaptively secure under reasonable leakages and performs better than state-of-the-art schemes

Honeypots are commonly used tools in network security protection. However, low-interaction honeypots cannot obtain in-depth attack information, while the deployment of high-interaction honeypots is costly. This paper presents LLM-Pot, a novel high-interaction honeypot architecture powered by the Large Language Model (LLM), which explores the direction of intelligent honeypots and addresses the limitations of conventional honeypot solutions. LLM-Pot utilizes LLM to generate dynamic, context-aware responses that accurately simulate the behaviors of real operating systems. To demonstrate the effectiveness of LLM-Pot, this work used offline and online evaluations. The offline evaluation compared LLM-Pot and Cowrie by analyzing their responses to selected commands, and the results demonstrated LLM-Pot’s superior ability in handling complex operations. Online evaluation deployed honeypots in the cloud and captured extensive attack data over two weeks. The evaluation results demonstrate that our LLM-driven approach outperforms traditional honeypots across multiple key metrics, validating LLM-Pot’s superior deception capabilities.

Electronic medical records (EMRs) are the essential sensitive personal data that is shared between patients and doctors through the semi-trusted E-health cloud. In the real application, multiauthority ciphertext-policy attribute-based searchable encryption (MA-CP-ABSE) is suitable to protect the security of the EMR for it possesses fine-grained access permission, efficient key management, and retrieval function over the encrypted data. However, previously proposed MA-CP-ABSE schemes are commonly restricted by the central authority, which is required by all attribute authorities in some operations like generating users’ secret keys. To deal with this issue, we proposed a decentralized MA-CP-ABSE (DMA-CP-ABSE) scheme. In our scheme, any attribute authority can become an independent authority to generate a secret key, which is no longer controlled by the central authority. Furthermore, a single-keyword search will produce many disrelated search results. For this, we enhance the DMA-CP-ABMSE scheme by implementing a multikeyword search function to improve the search accuracy. Besides, to handle the dynamic changing of access permission, we have designed the attribute revocation methods. Finally, we process the formal security analysis to demonstrate our scheme is secure under the chosen-keyword attack (CKA) and implement experiments to show its efficiency and feasibility.

The backdoor attack has emerged as a prevalent threat that affects the effectiveness of machine learning models in intelligent vehicles. While such attacks may not impair the normal performance of the trained model, they can be exploited by malicious entities to manipulate model inferences, resulting in serious problems. In this paper, we design a dynamic gradient clipping (DGC) method aimed at rectifying backdoor models by eliminating the underlying backdoor trigger. Firstly, we construct a repair dataset fused by some clean samples and few-shot backdoor samples to amplify the backdoor behavior when we only obtain limited backdoor samples. Subsequently, we introduce sample states to characterize the backdoor behavior of the target model, determined by the model's inference outcome. Finally, we devise the DGC method to clip parameter gradients at varying degrees, effectively eliminating the backdoor trigger within the target model. Through the evaluation, the simulation results demonstrate that our DGC method exhibits robust defense capabilities against four contemporary state-of-the-art backdoor attacks, reducing the attack success rate by 95% with only$0.1\% \sim 4.8\%$model accuracy loss.

The combination of smart cards and passwords has given birth to one of the most prevalent two-factor authentication (2FA) approaches. Numerous 2FA schemes have been proposed, nevertheless, most of them either do not possess critical security properties or are not efficient for implementation on smart cards. It is generally considered that asymmetric cryptographic primitives are indispensable to achieve security goals, which are burdensome for resource-limited devices. That is, the literature is being stuck with the security-efficiency tension. In this paper, we propose a 2FA protocol only resorting to symmetric primitives. Specifically, with the puncturable pseudorandom function, the proposed protocol hits three birds: it achieves three subtle security goals, i.e., resisting offline password guessing attacks, perfect forward secrecy and anonymity. It alleviates the long-standing security-efficiency conflict that is considered intractable in the literature. The proposed protocol is provably secure within the harshest adversary model to date. Furthermore, the evaluation results demonstrate that our protocol is the optimal choice when considering both security and efficiency.

Privacy preservation are becoming increasingly significant in machine learning, with recent privacy regulations requiring the deletion of personal data and its impact on models. Although erasing data from storage is simple, removing the influence of data on models remains a challenge. Federated unlearning is an emerging paradigm that aims to forget the knowledge contributed by some specific data to the federated model. In this paper, we design a novel federated unlearning strategy, named FedWiper, which enables exact unlearning in federated learning by erasing specific data and its impact from the federated model. Specifically, based on the granularity of the dataset, we propose training multiple federated submodels to construct a federated unlearning framework, thereby narrowing the scope of the impact of wiped data. Furthermore, the proposed Uni-Adapter structure effectively mitigates the negative impact on model performance from diminishing the dataset scale, while also reducing communication cost. Rather than focusing solely on achieving indistinguishability unlearning of the model for classification task, we extend FedWiper to unlearning for multiple types of tasks and achieve the exact unlearning. Experiments demonstrate that FedWiper can not only accelerate federated unlearning, but also achieve exact unlearning across multiple types of tasks in federated learning while ensuring minimal loss of model performance. Our Code: https://github.com/grey1989/FedWiper.

Federated learning (FL) has gained popularity in the field of machine learning, which allows multiple participants to collaboratively learn a highly-accurate global model without exposing their sensitive data. However, FL is susceptible to poisoning attacks, in which malicious clients manipulate local model parameters to corrupt the global model. Existing FL frameworks based on detecting malicious clients suffer from unreasonable assumptions (e.g., clean validation datasets) or fail to balance robustness and efficiency. To address these deficiencies, we propose FedDMC, which implements robust federated learning by efficiently and precisely detecting malicious clients. Specifically, FedDMC first applies principal component analysis to reduce the dimensionality of the model parameters, which retains the primary parameter feature and reduces the computational overhead for subsequent clustering. Then, a binary tree-based clustering method with noise is designed to eliminate the effect of noisy points in the clustering process, facilitating accurate and efficient malicious client detection. Finally, we design a self-ensemble detection correction module that utilizes historical results via exponential moving averages to improve the robustness of malicious client detection. Extensive experiments conducted on three benchmark datasets demonstrate that FedDMC outperforms state-of-the-art methods in terms of detection precision, global model accuracy, and computational complexity.

The increasing use of multi-biometric authentication has raised concerns about the security of biometric templates. Many template protection methods based on convolutional neural network have been presented, but most involve a trade-off between authentication accuracy and template security. In this paper, we present a cancelable multi-biometric template protection scheme that combines deep hashing with cancelable distance-preserving encryption (CDPE), which provides high template security without degrading the authentication performance. Specifically, a deep hashing based architecture that minimizes the quantization loss is designed to map face and iris traits to binary codes. Next, CDPE is proposed to generate a protected template given the face binary code and a user-specific key obtained from the iris binary code, which preserves the distance between original templates in the protected domain to ensure authentication performance equivalent to unprotected systems. Digital lockers instead of the key are stored to further enhance the security, which can be unlocked with genuine biometric traits to get the correct key during authentication. Theoretical and experimental results on real face and iris datasets show that our scheme can achieve equal error rate of 0.23% and genuine accept rate of 97.54%, while guaranteeing irreversibility, revocability and unlinkability of protected templates.

Face and voice are two of the most popular traits used for authentication tasks in daily life, as they can be easily captured using low-cost visual and audio sensors on smartphones, laptops, tablets,etc. Many bimodal biometric authentication schemes based on these two traits have been presented to provide higher accuracy than unimodal systems. However, these schemes are inflexibility due to the requirement of submitting two traits simultaneously, and they lack template protection, which may lead to biometric data leakage. We present a cross-modal learning based bimodal biometric authentication scheme, which improves the flexibility of existing schemes while ensuring the biometric template security. We integrate cross-modal learning into the feature extraction to obtain a bimodal biometric shared representation given input face images and voice clips. In order to enhance biometric template security without sacrificing authentication accuracy, a residual network and polar codes based template protection method is proposed, which can eliminate the noise in shared representations due to intra-user variations and generate protected templates. We have evaluated the efficacy of the bimodal biometric scheme using a real video dataset containing face images and voice clips. Experimental results demonstrate that our scheme can achieve flexible authentication with high accuracy no matter the probe input is a face image, a voice clip or a combination of them. Furthermore, the security analysis demonstrates that our scheme provides irreversibility, unlinkability and revocability of protected templates.

Abstract Machine learning has become an integral part of modern intelligent systems in all aspects of life. Membership inference attacks (MIAs), as the significant model attacks, also jeopardize the privacy of the intelligent systems. Previous works on defending MIAs concentrate on the model output perturbation or tampering with the training process. However, data and model reuse are common in intelligent systems, which results in the lack of scalability of previous defending works. This paper proposes a new privacy‐preserving framework for images to transform source data into synthetic data to train models against MIAs. The synthetic data makes it easy to defend MIAs during data and model reuse to improve the scheme's scalability. The framework generates synthetic data satisfying differential privacy through the variational autoencoder model's information extraction and data generation capabilities to improve model accuracy. A noise addition mechanism with metric privacy for the latent code generated from source data is proposed, where noise is the product of Γ‐distribution and unit hyper‐sphere samples. Moreover, it is proved that the synthetic data also satisfies metric privacy. The experimental evaluations demonstrate that the framework reduces MIAs' attack accuracy to about 0.5 and maintains higher utility than DP‐SGD under the same setting.

Searchable encryption (SE) has emerged as a cryptographic primitive that allows data users to search on encrypted data. Most existing SE schemes usually delegate search operations to an intermediary such as a cloud server, which would inevitably result in single-point failure, privacy leakage, and even untrustworthy results. Several blockchain-based SE schemes have been proposed to alleviate these issues; however, they suffer from some issues, such as the support for multi-keyword multi-owner model, query privacy and data storage availability. In this paper, we propose BPMS, blockchain-based privacy-preserving multi-keyword search in multi-owner setting, which supports searching over encrypted data in trustworthy, private and efficient manners. The attribute Bloom filter has been introduced into our BPMS to build indexes, which protects query privacy and improves index generation performance. To guarantee data storage availability, our BPMS leverages the advantages of IPFS (InterPlanetary File System) to store large scale of encrypted data. Security proof and comparative analysis in theory indicate that our BPMS is more secure and efficient. A series of experiments conducted on a real-world dataset further demonstrate that our BPMS is feasible in practice.

As one of the most popular relational databases for embedded devices, SQLite is lightweight to be embedded into applications without installing a specific database management system. However, simplicity and easy-to-use are double-edged swords; while bringing convenience, they also make data processing and storage risky. For example, an attacker can obtain data from a database file or memory and tamper with it once he has gained higher privileges, threatening the database's confidentiality and integrity. To address such security issues, based on a trusted execution environment (TEE) and a trusted platform module (TPM), we have proposed Smaug, a general secure scheme to ensure the confidentiality and integrity of SQLite and similar databases. With Smaug, all the critical data is stored in ciphertext, and data integrity protection is also provided. Besides, with TEE, all the sensitive operations are isolated from the untrusted environment, which can effectively resist attacks against memory. In addition, we use TPM to provide a solid root-of-trust (RoT) for the system. Finally, we have implemented a prototype system, and the performance evaluations have clarified the dominant factors that affect the system availability, providing a reference to the design and implementation of similar systems.

Organizations tend to collaboratively train the deep learning model over their combined datasets for a common benefit (e.g., better-trained model or learning a complicated model). However, due to the consideration about privacy leakage, organizations cannot share their data directly, especially related to sensitive domains. In this paper, a privacy-preserving collaborative deep learning mechanism, namely Sigma, is designed to allow participating organizations to train a collective model without exposing their local training data to the others. Specifically, a single-server-aided private collaborative architecture is introduced to achieve the private collaborative learning, which protects organizations’ data even if$n-1$out of$n$participants colluded. We also design a practical protocol to perform the secure model training, which can resist the typical inference attack through the sharing information. After that, we propose a fair model releasing mechanism for participants and introduce differential privacy to prevent model stealing and membership inference attack. Furthermore, we prove that Sigma can ensure participants’ privacy preservation and analyze the communication overhead in theory. To evaluate the effectiveness and efficiency of Sigma, we conduct an experiment over two real-world datasets and the simulation results demonstrate that Sigma can efficiently achieve the collaborative model training and effectively resist the membership inference attack.

With the intelligentization of Maritime Transportation System (MTS), Internet of Thing (IoT) and machine learning technologies have been widely used to achieve the intelligent control and routing planning for ships. As an important branch of machine learning, federated learning is the first choice to train an accurate joint model without sharing ships' data directly. However, there are still many unsolved challenges while using federated learning in IoT-enabled MTS, such as the privacy preservation and Byzantine attacks. To surmount the above challenges, a novel mechanism, namely DisBezant, is designed to achieve the secure and Byzantine-robust federated learning in IoT-enabled MTS. Specifically, a credibility-based mechanism is proposed to resist the Byzantine attack in non-iid (not independent and identically distributed) dataset which is usually gathered from heterogeneous ships. The credibility is introduced to measure the trustworthiness of uploaded knowledge from ships and is updated based on their shared information in each epoch. Then, we design an efficient privacy-preserving gradient aggregation protocol based on a secure two-party calculation protocol. With the help of a central server, we can accurately recognise the Byzantine attackers and update the global model parameters privately. Furthermore, we theoretically discussed the privacy preservation and efficiency of DisBezant. To verify the effectiveness of our DisBezant, we evaluate it over three real datasets and the results demonstrate that DisBezant can efficiently and effectively achieve the Byzantine-robust federated learning. Although there are 40% nodes are Byzantine attackers in participants, our DisBezant can still recognise them and ensure the accurate model training.

The widespread usage of wearables to provide healthcare services prompts the need for secure group communication among multiple devices using group keys. Gait-based group key establishment schemes are either vulnerable to video attacks, or fail to offer a secure group key update mechanism when group device changes. In this paper, we present an electrocardiogram (ECG) signals based group device pairing protocol, which can strengthen the security and reduce the overhead of wearables. Specifically, we first design a robust and lightweight fuzzy extractor that supports secure and efficient group device association between wearables. Meanwhile, we propose Improved Martingale Randomness Extraction (IMRE) algorithm, which utilizes the trend of InterPulse Interval (IPI) from ECG signal to extract high-entropy keys. Then we present a membership management mechanism that enables group key dynamic update when group device changes. Finally, we simulate our protocol and evaluate the accuracy and efficiency by various experiments. The experimental results demonstrate that the proposed work is robust and efficient, and the threat model-based security analysis shows that the proposed protocol can prevent both active and passive attacks.

Cloud computing, which provides a brand-new service model, has become an important infrastructure in the information age, and has been widely used in numerous fields. The Key-Policy Attribute-Based Encryption (KP-ABE) scheme allows the encrypted data with fine-grained access control in the cloud environment. However, achieving large-scale user revocation in the application scenario of KP-ABE becomes one of the thorny problems. Furthermore, the computation and communication costs of the previous user revocation schemes were generally high, especially when a large number of users were revoked. To address these problems, an enhanced high-performance user-revocable KP-ABE scheme combined with the puncture method was proposed. In this article, the user could be revoked by the fine-grained restriction policy. When revoking the user, the cloud would run the puncture algorithm to embed the restriction policy defined by the data owner into the ciphertext. This method could effectively omit the re-encryption and key updating processes, by which the computation and communication overhead of the user revocation are efficiently reduced, and the user revocation becomes more flexible and efficient. Moreover, the Chosen-Plaintext Attack (CPA) security proof and extensive simulation results demonstrate the reliability and efficiency of the proposed scheme for user revocation in a cloud environment.

Collaborative learning has been used to train a joint model on geographically diverse data through periodically sharing knowledge. Although participants keep the data locally in collaborative learning, the adversary can still launch inference attacks through participants’ shared information. In this article, we focus on the property inference attack during model training and design a novel defense mechanism, namely, NOSnoop, to defend such an attack. We propose a collaborative meta-learning architecture to learn the common knowledge over all participants and utilize the natural advantage of meta-learning to hide the sensitive property data. We consider both irrelevant property and relevant property preservation in NOSnoop. For irrelevant property preservation, we utilize the inherent advantage of meta-learning to hide the sensitive property data in meta-training support data set. Thus, the adversary cannot capture the key information related to the sensitive properties and cannot infer victim’s private property successfully. For relevant property preservation, an adversarial game is further proposed to reduce the inference success rate of the adversary. We conduct comprehensive experiments to evaluate the effectiveness of NOSnoop. When hiding the sensitive property data in meta-training support data set, NOSnoop achieves an inference AUC score as low as 0.4984 for irrelevant property preservation, meaning the adversary cannot distinguish whether the training batch has the sensitive property data or not. When preserving the relevant property, NOSnoop is able to achieve an inference AUC score of 0.5091 without compromising model utility.

Industrial Internet of Things (IIoT) devices are widely used for monitoring and controlling the process of automated manufacturing. Owing to the limited computing capacity of the IIoT sensors in the production line, the scheduling task in the production line needs to be offloaded to the edge computing server (ECS). To obtain the desired quality of service (QoS) during offloading scheduling tasks, the precise interaction information between the production line and ECSs has to be uploaded to the cloud platform, which poses privacy issues. The existing works mostly assume that all the interaction information, i.e., the offloading decision for the subtask in a scheduling task, has same privacy level, which cannot meet the various privacy requirements of the offloading decision for the subtask. Hence, we propose a local-differential-privacy-based deep reinforcement learning (LDP-DRL) approach in the edge-cloud-assisted IIoT to provide personalized privacy guarantee. The LDP mechanism can generate different levels of noise to satisfy the various privacy requirements of the offloading decision for the subtask. The prioritized experience replay is integrated in DRL to reduce the impact of noise on the QoS performance of task offloading. The formal analysis of LDP-DRL is provided in terms of privacy level and convergence. Finally, extensive experiments are conducted to evaluate the effectiveness, the capacity of privacy protection, the impact of discount factor on the convergence, and the cost efficiency of the LDP-DRL approach.

Because of the powerful computing and storage capability in cloud computing, machine learning as a service (MLaaS) has recently been valued by the organizations for machine learning training over some related representative datasets. When these datasets are collected from different organizations and have different distributions, multi-task learning (MTL) is usually used to improve the generalization performance by scheduling the related training tasks into the virtual machines in MLaaS and transferring the related knowledge between those tasks. However, because of concerns about privacy breaches (e.g., property inference attack and model inverse attack), organizations cannot directly outsource their training data to MLaaS or share their extracted knowledge in plaintext, especially the organizations in sensitive domains. In this article, we propose a novel privacy-preserving mechanism for distributed MTL, namely NOInfer, to allow several task nodes to train the model locally and transfer their shared knowledge privately. Specifically, we construct a single-server architecture to achieve the private MTL, which protects task nodes’ local data even if n-1 out of n nodes colluded. Then, a new protocol for the Alternating Direction Method of Multipliers (ADMM) is designed to perform the privacy-preserving model training, which resists the inference attack through the intermediate results and ensures that the training efficiency is independent of the number of training samples. When releasing the trained model, we also design a differentially private model releasing mechanism to resist the membership inference attack. Furthermore, we analyze the privacy preservation and efficiency of NOInfer in theory. Finally, we evaluate our NOInfer over two testing datasets and evaluation results demonstrate that NOInfer efficiently and effectively achieves the distributed MTL.

Deep convolutional neural networks (CNNs) for Super-Resolution (SR) from low-resolution (LR) images have achieved remarkable reconstruction performance with the utilization of residual networks and visual attention mechanism. However, the existing single image super-resolution (SISR) methods with deeper or wider network architectures encounter module representation bottleneck and neglect module efficiency in real-world applications. To solve these issues, in this paper, we design channel hourglass residual structure (CHRS) consisted of several nested residual modules for reducing parameters and extracting more representational features. Furthermore, we integrate channel attention (CA) mechanism into CHRS to generate channel hourglass residual block (CHRB) which can be easily extended to other methods for improving performance. We also propose channel hourglass residual network (CHRN) which not only pays attention to network learning efficiency but also learns more discriminative expressions. Extensive experiments demonstrate the effectiveness of our CHRN and the generalization ability of our CHRB.

Nonnegative matrix factorization seeks to find a basic matrix and a weight matrix to approximate the nonnegative matrix. It has proven to be a powerful low-rank decomposition technique for nonnegative multivariate data. However, its performance largely depends on the assumption of a fixed number of features. In this work, we propose a new probabilistic nonnegative matrix factorization which factorizes a nonnegative matrix into a low-rank factor matrix with {0,1} constraints and a nonnegative weight matrix. In order to automatically learn the potential binary features and feature number. A deterministic Indian buffet process variational inference is introduced to obtain the binary factor matrix. And the weight matrix is set to satisfy the exponential prior. In order to obtain the real posterior distribution of the two factor matrices, a variational Bayesian exponential Gaussian inference model is established. The comparative experiments on both the synthetic and real-world data sets show the efficacy of the proposed method.

Deep learning has aroused a lot of attention and has been used successfully in many domains, such as accurate image recognition and medical diagnosis. Generally, the training of models requires large, representative datasets, which may be collected from a large number of users and contain sensitive information (e.g., users' photos and medical information). The collected data would be stored and computed by service providers (SPs) or delegated to an untrusted cloud. The users can neither control how it will be used, nor realize what will be learned from it, which make the privacy issues prominent and severe. To solve the privacy issues, one of the most popular approaches is to encrypt users' data with their public keys. However, this technique inevitably leads to another challenge that how to train the model based on multi-key encrypted data. In this paper, we propose a novel privacy-preserving deep learning model, namely PDLM, to apply deep learning over the encrypted data under multiple keys. In PDLM, lots of users contribute their encrypted data to SP to learn a specific model. We adopt an effective privacy-preserving calculation toolkit to achieve the training process based on stochastic gradient descent (SGD) in a privacy-preserving manner. We also prove that our PDLM can achieve users' privacy preservation and analyze the efficiency of PDLM in theory. Finally, we conduct an experiment to evaluate PDLM over two real-world datasets and empirical results demonstrate that our PDLM can effectively and efficiently train the model in a privacy-preserving way.

With the booming penetration of wrist worn smart devices in daily lives, a wide range of applications have been enabled, such as exchanging social information, sharing sports data, and sending messages. Securing data exchange between these devices has become a challenging issue, considering the high security requirements and low computation capabilities of these wrist worn devices. In this paper, we propose a secure wrist worn smart device pairing scheme by exploiting the motion signal of the devices generated by the handshake to negotiate a reliable key between users. To ensure the security of key negotiation, a novel fuzzy cryptography algorithm is further developed. Compared with existing algorithms, the proposed algorithm avoids complicated error correction algorithms and has low requirements for data coincidence on the premise of individual differentiation. At the same time, the security is guaranteed by feature reordering and protection of auxiliary data. Extensive experimental results are provided, which demonstrate that the proposed handshake acceleration-based pairing scheme is robust, secure, and efficient.

Since the dummy-based method can provide precise query results without any requirement for a third party or key sharing, it has been widely used to protect the user's location privacy in location-based services. However, the neighboring location sets submitted in consecutive requests always include a close spatiotemporal correlation, which enables the adversary to identify some dummies. Therefore, the existing dummy-based schemes cannot protect the user's location privacy completely. To solve this problem, based on the dummies generated by the existing schemes, this paper filters out the dummies that can be identified by taking into account of the spatiotemporal correlation from three aspects, namely time reachability, direction similarity and in-degree/out-degree. In this way, the rest dummies can satisfy the user's personalized privacy protection requirement. Security analysis shows that the proposed scheme successfully perturbs the spatiotemporal correlation between neighboring location sets, therefore, it is infeasible for the adversary to distinguish the user's real location from the dummies. Furthermore, extensive experiments indicate that the proposal is able to protect the user's location privacy effectively and efficiently.