Abstract Advanced Persistent Threats (APTs) is one of the most serious cybersecurity threats today, posing a substantial threat to enterprises and organizations due to their stealthy and targeted nature. Data provenance-based methods are widely used for APT detection but often rely on specific rules and high-quality data due to limitations in capturing complete graph structures, reducing their effectiveness in diverse detection environments. To overcome this issue, we propose APT-HERA, a model employs heterogeneous graph representation learning to learn system behavior patterns that can adapt to environments with limited data. The embedding representations of the provenance graph in APT-HERA are derived from both homophily and heterogeneity perspectives, thereby enabling a more comprehensive extraction of the rich structural information contained within the provenance graph. The performance of APT-HERA was evaluated on four public datasets. Experimental results demonstrate that APT-HERA achieves 98% precision in information-constrained detection scenarios, outperforming state-of-the-art methods including MAGIC, Flash, and ThreaTrace under such conditions.

With the booming development of blockchain, it has gradually gained wide attention in the Internet of Things (IoT), finance, and other fields. However, due to the shared nature of blockchain ledgers among multiple users, sensitive user information, such as transaction amounts and private agreements, can be easily exposed. This poses significant privacy concerns for blockchain users. To address this issue, we propose TrustChain, a high-performance smart contract model based on the Trusted Execution Environment (TEE). TrustChain aims to safeguard the privacy of smart contract codes and user data by leveraging the secure execution environment provided by the TEE. Specifically, we introduce the TEE to run the smart contract with security and privacy without introducing a heavyweight cryptographic algorithm, thus improving the performance of the system. When running smart contracts, the operate nodes equipped with TEE ensure that the Operating System (OS) of the node itself cannot access the data within the TEE. This isolation effectively separates the sensitive information of the smart contract from the external environment. Furthermore, we introduce Verifiable Random Functions (VRFs) to randomly choose the operate nodes to prevent collusion attacks, further improving the security of the model. The graph ledger, based on the Directed Acyclic Graph (DAG), is used to adapt to the high-performance characteristics of a smart contract system based on the TEE. Finally, we simulate the scheme in TrustZone and demonstrate the feasibility of TrustChain through a series of experiments and analyses. The analysis and experimental results demonstrate that our solution exhibits excellent privacy protection performance and achieves higher throughput compared to traditional smart contracts. • We have introduced TrustChain, a smart contract model based on TEE, to ensure the privacy and security of smart contracts. • VRFs are proposed to randomly select operate nodes, preventing collusion and enhancing defense against malicious attacks. • The redesigned consensus mechanism limits blockchain storage to smart contract outputs, preventing leakage of sensitive information. • We enhanced smart contract performance by integrating a DAG-based ledger with TEE's low-latency execution.

With the fast development of the Internet of Things (IoT), the growing amounts of data transmitted through edge devices tempt hackers to attack vulnerabilities. Because of data fragmentation and heterogeneous data distribution of IoT, attack detection models on edge devices are proficient at detecting only a limited set of specific attacks, causing a high false alarm rate when detecting new traffic data. Personalized Federated Learning (PFL) widely expands the range of detectable attacks and adapts local models to new traffic data by one step of gradient descent. However, it demands a part of the new traffic data (test-support set) with correct labels to realize adaptation, which is labor-consuming when detecting large amounts of traffic data. To solve this issue, our main idea is to find helpful models to pre-label the test-support set, we propose a novel self-labeled PFL called SOH-FL, including an autoencoder based on cosine similarity (CT-AE) to extract features and an aggregation method (BS-Agg) to tailor models for pre-labeling test-support sets depending on features extracted from edge devices. SOH-FL is evaluated in three heterogeneous scenarios using the CICIDS2017 dataset, and consistently outperforms the baselines across all metrics, achieving performance comparable to PFL without manual labeling. In the real-world feature heterogeneous scenarios of the IoT-23 and TON-IoT datasets, SOH-FL achieves accuracy improvements of 11.5% and 9.1% over the baseline, respectively. The experimental code is publicly available at https://github.com/deer-echo/SOH-FL.git.

In cloud Electronic Medical Records (EMRs), health-related private information such as genetics and diseases is contained. Thus, the secure ownership transfer protocol should protect users’ privacy. In certain scenarios, some users, including patients, doctors, medical and research institutions, may be offline. As a result, existing protocols cannot be directly employed. Motivated by these observations, in this paper we propose a secure and efficient ownership transfer for cloud EMRs auditing protocol. Specifically, our protocol allows the existence of offline users while ensuring users anonymity, it is achieved using different signature constructions. Additionally, a tracing mechanism is introduced to safeguard against malicious users. We rigorously prove the security of our protocol, comprehensively evaluate the performance of it, and compare our protocol with a few closely relevant protocols. According to the evaluations, our protocol significantly improves ownership transfer efficiency while achieving additional functionalities, including public verifiability, multi-ownership transferability, anonymity, and traceability.

Abstract Federated learning (FL) is a focus of research in the area of privacy protection since it does not have the privacy issues that arise from data concentration. Although its emergence has attracted widespread attention from academia and industry, existing works on FL still face security challenges. FL can be considered as a cooperative-based task to achieve global model sharing. However, the model raises issues of cooperative security, such as free-riding and poisoning attacks. Therefore, we focus on the behavior of participants with strong cooperative relationships and build a Cooperative Security-aware Federated Learning model using blockchain. In addition, we propose a credit-based economic model including profit and punishment mechanisms to ensure fairness and security among participants. Furthermore, for data privacy, we develop a participation permission strategy to protect the privacy of participants through proxy re-encryption and homomorphic encryption. Finally, the simulation results of the real datasets show that the proposed scheme achieves a good performance in security and accuracy.

Abstract In certain cloud Electronic Medical Records (EMRs) applications, the data ownership may need to be transferred. In practice, not only the data but also the auditing ability should be transferred securely and efficiently. However, we investigate and find that most of the existing data ownership transfer protocols only work well between two individuals, and they become inefficient when dealing between two communities. The proposals for transferring tags between communities are problematic as well since, they require all members get involved or a fully trusted aggregator facilitates ownership transfer, which are unrealistic in certain scenarios. To alleviate these problems, in this paper we develop a secure auditing protocol with decentralized (t, n)-threshold ownership transfer for cloud EMRs. This protocol is designed to operate efficiently without requiring the mandatory participation of every user or the involvement of any trusted third-party. It is achieved by employing the threshold signature. Rigorous security analyses and comprehensive performance evaluations illustrate the security and practicality of our protocol. Specifically, according to the evaluations and comparisons, the communication and computational consumption is independent of the file size, i.e., it is constant in our protocol for both communities.

Abstract Studies show that machine learning models trained from biased data can discriminate against groups with certain sensitive attributes. This problem can be mitigated by cleaning the original data or learning fair representations. However, collecting real data in real-life is extremely time and resource-consuming, whereas generative models (e.g., GANs) can create new data that enable more application scenarios. Therefore, utilizing fair data generated by generative models can benefit various downstream tasks. In this paper, we propose a information-minimizing generative adversarial network to improve the fairness of machine learning by generating fair data. An ANOVA-based latent factor is constructed in the input for reducing the accuracy loss, and the joint adversarial training between the generator and classifier can better solve the indirect discrimination and achieve fair classification. Extensive experiments on various environments show the effectiveness of the proposed method.

Summary Existing dummy‐based trajectory privacy protection schemes do not take into account the correlation of multiple locations and whether the generated trajectory based on dummies matches the user's movement mode, which enables the adversary to identify some dummies. Aiming at this problem, to ensure that the generated trajectories match the movement modes of users, historical query trajectories of users are selected. In this way, the generated dummies on the selected historical query trajectories are based on the location relationship of adjacent time and the background information constraint of the dummy, namely, they should meet the time reachability, the similarity of historical query probability and the maximum in‐degree. Security analysis shows that the proposed scheme effectively perturbs the spatiotemporal correlation between the real location and dummies. Furthermore, the proposed scheme is compared with the existing schemes in terms of single‐point location exposure risk and trajectory exposure risk, and the experimental results indicate that the proposal has significant improvement in location privacy protection of the user.

Proximity detection is one of the most popular location-based applications in social networks when users intend to find their nearby friends. However, the existing proximity detection has access to precise and real-time location information of users, raising serious privacy concerns for millions of users. A number of privacy-preserving models have taken shape over the past decade, but they almost universally rely on syntactic privacy models such as$k$-anonymity and location perturbation, which are proved to waver in the balance of privacy and availability requirements. To solve this problem, we introduce a novel location privacy-preserving mechanism for proximity detection to support user-defined range queries while guaranteeing a certain level of privacy. It divides the proximity detection service into two independent subservices and ensures that each subservice provider can only access part of the user’s location information, which is encoded by Geohash and divided into two parts (i.e., prefix and suffix). By adjusting the length of location encoding, we can make a good trade-off between accuracy and system overhead. The privacy requirements of users are not implicated in the quality of service, and it achieves the balance of privacy and utility. The analysis results through an extensive simulation indicate that our scheme successfully ensures that neither each server in the system nor an external attacker can obtain the real location of the user. Moreover, it demonstrates the effectiveness of the proposed scheme.

To secure communications, this article presents a key extraction scheme for smart devices using ambient sounds. Specifically, it is designed for the scenario when smart devices do not have any pre-loaded secrets. Moreover, it can be implemented when smart devices have no access to the online trusted third party or the Network Time Protocol server. In our scheme, smart devices achieve synchronization by making use of ambient sounds. Then, they calculate and obtain the pairing distance and secure distance by applying a band-pass filter. Completing these operations, two smart devices can directly extract a communication key. We analyze the security of our scheme and evaluate the performance of it by implementing the scheme using a few off-the-shelf devices. The experimental results indicate that compared with related schemes, the bit generation rate of our scheme has significant improvement (increases at least 80%), and it can reach 312 bit/s.

Summary Modern smartphone platforms offer a multitude of useful features to their users, but at the same time, they highly affect privacy, which may lead to unnecessary personal data being collected. In this paper, we proposed an evolutionary game model to study permission request strategies of the bounded rational application providers. We study the proposed model with detailed simulations. Initial results demonstrate that evolution processes are influenced by four factors: revenue increase ratio, market share, credit cost, and market attraction. It also suggests that establishing a privacy alarm mechanism not only can improve the users' privacy awareness but also decrease the market share when providers over‐request permissions and push them request permissions properly.

Summary A major concern of the deployment of location based services (LBSs) is the safeguards of the user's location data collected by service providers, since personal location data may imply sensitive private information. However, most available works proposed so far rely on syntactic privacy models such as k‐anonymity and location perturbation, which are proved to quiver in the balance with privacy and usability requirements. In this article, we provide a new ‐map mechanism to help users better understand the privacy/accuracy tradeoff process and preserve location data. In our ‐map model, the user can specify a geographic region to hide her precise location and suppress the following queries in the same area to meet her privacy requirement, while maintaining and understanding its usability. In addition, we propose a new notion of ‐privacy based on differential privacy to account for the temporal‐spatial correlation and history correlation in the case of crossing region, which is the major privacy concern of a moving user's trace. Finally, we evaluate our framework by using an online LBS with real‐world data sets. The results not only indicate that the ‐map is significantly useful for identifying the privacy and utility tradeoffs but also show the effectiveness and practicality of the proposed ‐privacy.

Summary Fingerprinting positioning is used for indoor location estimation because of low cost and open access properties, in which a radio map is built by calibrating signal‐strength values at several training locations in the offline phase. However, due to the sophisticated propagation of radio signals, the received signal strength (RSS) in wireless network change as the environment changes, and the radio map built in the offline phase may be out of date. Furthermore, the recalibration of signal‐strength values for each environment change is laborious and time consuming. In this paper, we present a novel algorithm to reconstruct a radio map using real‐time signal‐strength readings received at some reference points. We first demonstrate that different features of signal propagation are obtained in different regions of indoor environment. Then, the indoor environment is divided into several regions by clustering the path‐loss parameters of each reference point. In addition, the relationship between the RSS of reference points and calibration nodes is established with robust linear regression. Finally, the real‐time radio map is updated dynamically according to robust regression and the real‐time RSSs of the calibration nodes. The experimental results show the usefulness of the proposed method and the accuracy of the localization can be improved.

Received signal strength (RSS) in wireless networks is widely adopted for indoor positioning purpose because of its low cost and open access properties. However due to the sophisticated propagation of radio signals, the RSS shows a significant variation during pedestrian walking, which introduces critical errors in deterministic indoor positioning. To solve this problem, the authors present a novel method to improve the indoor pedestrian positioning accuracy by modelling fingerprinting and information on the movement into a hidden Markov models (HMMs). They divide the whole continuous positioning process into specified‐size sub‐processes, which could efficiently reduce the accumulative and resonance error caused by iterative estimation. They use an accelerometer sensor to provide the information on the movement distance to calculate the transition probability of the HMMs. In their experiments, they demonstrate that, compared with the deterministic pattern matching algorithm, the proposed method greatly improves the positioning accuracy and shows robust environmental adaptability.

Proximity detection is one of the critical components in Location-based Social Networks (LBSNS), which has attracted much attention recently. With the advent of LBSNS, more and more users' location information will be collected by the service providers. However, with a potentially untrusted server, such a proximity detection service may threaten the privacy of users. In this paper, aiming at achieving enhanced privacy against the untrusted service providers in LBSNS, we introduce a new architecture with dual-servers for the first time and propose a privacy-preserving proximity detection method based on Geohash. In our architecture, the location coordinates of users are converted into a bit-string by dichotomy approximation, and divided into two subsets: prefix and suffix. The social network server firstly selects the candidate neighbors only in the light of the prefix, and then a third-party server is introduced to compute the relative distance of candidate neighbors according to the suffix. Each server can only get a subset of location code, instead of the whole location information of users as the previous work. We also prove that the new construction is secure under the untrusted server model with enhanced privacy. Finally, we provide extensive experimental results to demonstrate the efficiency of our proposed construction.

In cloud computing environment, especially in big data era, adversary may use data deduplication service supported by the cloud service provider as a side channel to eavesdrop users' privacy or sensitive information. In order to tackle this serious issue, in this paper, we propose a secure data deduplication scheme based on differential privacy. The highlights of the proposed scheme lie in constructing a hybrid cloud framework, using convergent encryption algorithm to encrypt original files, and introducing differential privacy mechanism to resist against the side channel attack. Performance evaluation shows that our scheme is able to effectively save network bandwidth and disk storage space during the processes of data deduplication. Meanwhile, security analysis indicates that our scheme can resist against the side channel attack and related files attack, and prevent the disclosure of privacy information.

Received signal strength (RSS) in wireless networks is widely adopted for indoor positioning purpose because of its low cost and open access properties. However, the popular RSSs are observed to differ significantly from discrete devices' hardware even under the same wireless conditions. Signal strength difference-based approach is an efficient strategy to overcome the drawbacks of RSSs. In this paper, we present a robust location fingerprint based on SSD. Firstly, a beacon selection algorithm is proposed to choose the beacons that have the best distinguishing results and stability as reference beacons. Secondly, a robust location fingerprint based on differential signal strength and dynamic linear interpolation is suggested to reduce the high labor cost in offline phase. The evaluation results show that the proposed approaches can achieve higher accuracy and robustness. Copyright © 2016 John Wiley & Sons, Ltd.

MANET (mobile ad hoc network) is a collection of wireless mobile nodes forming a temporary communication network without the aid of any established infrastructure. Because mobile nodes are typically constrained by power and computing resources, a selfish node may not be willing to use its computing and energy resources to forward packets that are not directly beneficial to it, even though it expects others to forward packets on its behalf. This paper not only analyzes the effect of two typical kinds of selfish nodes through simulation methods, but also proposes resolving strategy respectively. For type 1 selfish node, this paper proposes CI-DSR (cooperation inspirited dynamic source routing) protocol, which introduces an objective reputation-based strategy into the DSR protocol. For type 2 selfish nodes, a self-saving energy strategy is proposed. Simulations indicate that both of two strategies can effectively tradeoff the selfishness and cooperation