Federated learning (FL) has been widely applied in medical field, which allows clients to collaboratively train global models without sharing local data. Nevertheless, the diversity and scarcity of samples from rare diseases may result in a decline in the performance of local models on client-side due to using a singular global model. Moreover, direct transmission of local models or parameters will likely lead to user privacy violations. To solve these problems, we propose a Grouped Federated Meta-Learning (GrFML) method to improve the performance of local personalization models while protecting data privacy. Specifically, we first utilize a self-attention mechanism to extract partial features from the client’s local data, which are uploaded to the server (medical data is susceptible to perturbation and data integrity, thus this process does not expose the private data). The server groups clients with similar features based on these extracted features. Then, multiple meta-models are trained on these groups and distributed back to the clients to enhance the performance of the client’s local models. Furthermore, during the FL process, we introduce dynamic perturbation to the uploaded gradients based on the model’s test accuracy to protect their privacy. Typically, the perturbation magnitude is directly proportional to the model’s test accuracy. Extensive experiments shown that the GrFML model significantly improves client personalization model accuracy and achieves a good privacy-utility trade-off.

Decentralized Federated Learning (DFL) is extensively applied in various areas, e.g., healthcare, finance, and Internet of Things (loT), offering practical solutions for distributed intelligent applications and data collaboration. In DFL systems, participants, e.g., edge devices, organizations, or nodes, collaborate in the training of a shared global model by aggregating local models from various participants. During this process, participants need to communicate frequently with a central authority/node/server to share model parameters. Such communication is vulnerable to malicious attacks or tampering, posing a significant threat to the integrity of model training. The integrity verification method can provide an integrity guarantee for the global model of DFL. However, most of the existing integrity verification schemes are centralized and not suitable for resource-constrained DFL scenarios. Therefore, how to verify the integrity of the global model becomes an important issue in DFL. To address it, we devise a global model integrity verification method for DFL. Specifically, we generate a digital signature for each global model parameter as proof of integrity, while improving the efficiency of integrity verification by electing delegates to conduct the verification process. A series of experiments is conducted to validate the performance of the proposed method. The experimental results demonstrate that our approach not only effectively ensures the integrity of the global model but also functions well under limited resources.

Federated Meta-Learning (FML) has achieved privacy protection and intelligent improvement in many fields, such as healthcare and finance. However, the small dataset characteristic of FML may satisfy the strong assumptions of Deep Leakage from Gradients (DLG) attacks, making it susceptible to such attacks. Previous research has considered the strong assumptions of DLG to be impractical in real-world scenarios, resulting in a gap in the defense against DLG attacks in FML scenarios. In this paper, we propose a method called Prevent Deep Leakage Gradients (PDLG), PDLG utilizes condensation of training dataset to prevent DLG in FML, addressing the strong assumption problem of DLG in real-world scenarios. Specifically, PDLG first condenses the local raw training data through the client. Second use condensation data instead of the raw training data to train the model, and evaluate the changes in model classification performance in FML. Third DLG is used for the condensed dataset. As the Mean Square Error (MSE) between the real and the dummy gradients decreased, DLG could not restore the original images. The experimental results demonstrated that the classification performance of the models trained on condensation data remained within 10%. PDLG achieving a balance between model performance and privacy.

Wearable Internet of Things (IoT) devices are easily compromised by malware due to their security vulnerabilities. The bots infected by malware may continue to infect healthy neighbor devices through wireless communication technology in mobile wearable IoT networks (WIoT). All bots form a botnet which eventually leads to a series of malicious attacks. Therefore, it is necessary to predict the dynamic malware propagation path between wearable devices, which can help provide target immunization measures on devices to prevent the formation of botnets. In this article, we capture the local interaction and spatial–temporal propagation behavior of malware utilizing the individual-based cellular automata (CA) model. First, taking into account the mobility of walking users carrying wearable devices in the actual WIoT, we present a human mobility model called Gauss–Markov truncated Levy walk (GM-TLW) to describe the movement patterns of mobile users. Second, based on the moving coordinates of all wearable devices obtained from the GM-TLW mobility model, we leverage the improved CA propagation model to study the time evolution of the number of bots and the spreading spatial distribution of malware. We compare our propagation model with the differential equation model and traditional CA model, and analyze the impact of various parameters on the dynamics of botnet formation using numerical simulations. Finally, detailed simulation results show that the GM-TLW model is more suitable for realistic human mobility scenarios. In addition, the proposed CA-based model is more precise than the differential equation model to modeling the malware propagation and provides a basis for defenders to adopt the optimal malware control strategies.

Model inversion attacks can reconstruct the training samples of victim deep learning models. The existing efforts heavily rely on auxiliary information of the target samples (prior target information) to achieve their adversarial goals. However, prior target information is hard to obtain in practice. In this paper, we explore the effect of class information in model inversion attacks to reduce the reliance of prior target information. Our contributions on class information exploitation are two-fold. Firstly, we propose a supervised inversion model, Supervised Model Inversion (SMI). The proposed inversion model learns pixel-level features and data-to-class features from the rounded-outputs of the victim model and labeled auxiliary dataset. Secondly, we leverage victim model's rounded-outputs to guide the optimization of reconstructing inversion samples after trained inversion model. Our experimental results show that inversion samples reconstructed by SMI are more visually plausible with more details, comparing to the three representative model inversion attacks. We further perform an extensive study on various auxiliary dataset settings. It is found that the class combination in the auxiliary dataset rather than the number of classes that determines the quality of inversion samples. The ground-truth labels can improve the qualities of inversion samples but not essential to inversion attacks.

With the growing proliferation of intelligent end devices and data analytics techniques, real momentum towards the development of smart healthcare networks (SHN) has already been evident. Multiple parties in SHNs continuously exchange medical data in order to achieve a precise diagnosis and process optimization. Privacy issue emerges since medical data are susceptible, while the combination of a series of medical data may lead to further privacy leakage. Adversaries launch unceasingly launch poisoning attacks, a dominant attack to maliciously manipulate data, severely impact the authenticity of the data transmitting over the SHNs, leading to misdiagnosing or even physical damage. In this paper, we propose a personalized differential privacy model built upon blockchain, in which the community density is exploited to customize the degree of privacy protection and inject corresponding noise data. Besides using blockchain as the underlying network architecture to defeat poisoning attacks. The proposed model can guarantee the authentication of the differentially private data, traceability of data, and single-point failure avoidance in SHN. Evaluation and extensive results using real-world data sets demonstrate the superiority of the proposed model.

Federated learning (FL) is widely adopted in traffic forecasting tasks involving large-scale IoT-enabled sensor data since its decentralization nature enables data providers’ privacy to be preserved. When employingstate-of-the-artdeep learning-based traffic predictors in FL systems, the existing FL frameworks confront overlarge communication overhead when transmitting these models’ parameter updates since the modeling depth and breadth renders them incorporating an enormous number of parameters. In this article, we propose a practical FL scheme, namely, Clustering-based hierarchical and Two-step-optimized FL (CTFed), to tackle this issue. The proposed scheme follows adivide et imperastrategy that clusters the clients into multiple groups based on the similarity between their local models’ parameters. We integrate the particle swarm optimization algorithm and devises a two-step approach for local model optimization. This scheme enables only one but representative local model update from each cluster to be uploaded to the central server, thus reduces the communication overhead of the model updates transmission in FL. CTFed is orthogonal to the gradient compression- or sparsification-based approaches so that they can orchestrate to optimize the communication overhead. Extensive case studies on three real-world data sets and threestate-of-the-artmodels demonstrate the outstanding training efficiency, accurate prediction performance, and robustness to unstable network environments of the proposed scheme.

A Covert Electricity-Theft Cyberattack Against Machine Learning-Based Detection Models

Internet of Things (IoT) anomaly detection is significant due to its fundamental roles of securing modern critical infrastructures, such as falsified data injection detection and transmission line faults diagnostic in smart grids. Researchers have proposed various detection methods fostered by machine learning (ML) techniques. Federated learning (FL), as a promising distributed ML paradigm, has been employed recently to improve detection performance due to its advantages of privacy-preserving and lower latency. However, existing FL-based methods still suffer from efficiency, robustness, and security challenges. To address these problems, in this article, we initially introduce a blockchain-empowered decentralized and asynchronous FL framework for anomaly detection in IoT systems, which ensures data integrity and prevents single-point failure while improving the efficiency. Further, we design an improved differentially private FL based on generative adversarial nets, aiming to optimize data utility throughout the training process. To the best of our knowledge, it is the first system to employ a decentralized FL approach with privacy-preserving for IoT anomaly detection. Simulation results on the real-world dataset demonstrate the superior performance from aspects of robustness, accuracy, and fast convergence while maintaining high level of privacy and security protection.

Due to the dynamic changes of network data over time, static intrusion detection systems cannot adapt well to the behavioral characteristics of the input network data, resulting in reduced detection accuracy. In addition, continuous input data streams will bring huge challenges to resource storage and computing costs. Therefore, we propose an intrusion detection method of dynamic ensemble incremental learning (DEIL-RVM), and realize a dynamically adjusted ensemble intrusion detection model. In which a new overall misclassification probability weight value (OMPW) based on incremental set or data chunk is designed as the basis for updating the ensemble model, and it can be used to prune and replace the poor base component in the ensemble model. We presented a probabilistic decision function taking into account the posterior probability of each base RVM model dividing the sample into each category. The RVM with high sparsity is used as the base component to obtain the good balance between the accuracy, robustness and resource consumption, which can sacrifice less time and storage cost in ensemble incremental learning while achieving higher detection accuracy and stability in network data streams.

Machine learning as a service (MLaaS) brings many benefits to people's daily life. However, the service mode of MLaaS will increase the risk of users' privacy leakage. Existing works focusing on privacy-preserving based on encryption, differential privacy, and distributed framework require high computing resources or cannot be applied in MLaaS. In this paper, we propose feature dilution (FD), a noise-based desensitization algorithm to remove sensitive information in raw data. In particular, FD continuously adds raw data features to the random noise until it meets the minimum amount for an effective query, and we call this noise weak-feature noise (WFN). By fine-tuning the MLaaS architecture, we have realized that users can utilize WFN to get normal services without exposing their local private data. Meanwhile, noise addition technology is introduced by us to reduce the risk of privacy leakage caused by “weak features”. Extensive experiments have demonstrated that users can use FD to obtain effective services without exposing their private data. Finally, we conducted practical tests on weak-feature noises and found that these noises are difficult to use by malicious service providers.

Next point-of-interest (POI) recommendation has been an important task for location-based intelligent services. However, the application of such promising technique is still limited due to the following three challenges: 1) the difficulty of capturing complicated spatiotemporal patterns of user movements; 2) the hardness of modeling fine-grained long-term preferences of users; and 3) the effective learning of interaction between long- and short-term preferences. Motivated by this, we propose a memory augmented hierarchical attention network (MAHAN), which considers both short-term check-in sequences and long-term memories. To capture the complicated interest tendencies of users within a short-term period, we design a spatiotemporal self-attention network (ST-SAN). For long-term preferences modeling, we employ a memory network to maintain fine-grained preferences of users and dynamically operate them based on users' constantly updated check-ins. Moreover, we first employ a coattention network/mechanism to integrate the proposed ST-SAN and memory network, which can fully learn the dynamic interaction between long- and short-term preferences. Our extensive experiments on two publicly available data sets demonstrate the effectiveness of MAHAN.

The demand for data center bandwidth has exploded due to the continuous development of cloud computing, causing the use of network resources close to saturation. Optical network has become an encouraging technology for many burgeoning networks and parallel/distributed computing applications because of its huge bandwidth. This article focuses on efficient embedding of data centers into optical networks, which aims to reduce complexity of the network topology by using the parallel transmission characteristics of optical fiber. We first present a novel virtual network embedding (VNE) mathematical model used for optical data center networks. Then we derive a priority of location VNE algorithm according to node proximity sensing and path comprehensive evaluation. Furthermore, we propose routing and wavelength assignment for DCNs into optical networks, and identify the lower bound of the required number of wavelengths. Extensive evaluations show that the proposed embedding algorithm can reduce the average waiting time of virtual network requests by 20 percent, increase the request acceptance rate and revenue-overhead ratio by 13 percent, as compared to the latest VNE algorithm.

Non technical losses (NTL) detection plays a crucial role in protecting the security of smart grids. Employing massive energy consumption data and advanced artificial intelligence (AI) techniques for NTL detection are helpful. However, there are concerns regarding the effectiveness of existing AI-based detectors against covert attack methods. In particular, the tampered metering data with normal consumption patterns may result in low detection rate. Motivated by this, we propose a hybrid data-driven detection framework. In particular, we introduce a wide & deep convolutional neural networks (CNN) model to capture the global and periodic features of consumption data. We also leverage the maximal information coefficient algorithm to analysis and detect those covert abnormal measurements. Our extensive experiments under different attack scenarios demonstrate the effectiveness of the proposed method.

In the age of big data, the rapid development of social networking applications has become an improtant data source, while the massive collection of personal data leads to significant privacy concerns. Differential privacy emerged as an effective tool to get access to useful information while provide strong privacy guarantees. However, most the current proposed solutions suppose that all individuals across the network require a uniform level of privacy protection, which rules out of individuals' personalized requirements. Aiming at solving this problem, in this paper, we propose a trust-grained personalized differential privacy mechanism, called TGDP, by combining the notion of trust. Specifically, whenever a user wants to get another user's personal information, the proposed mechanism returns a corresponding private response in which the privacy level selected for each individual depend on the trust value between them in the network. Compared with traditional methods, the scheme can provide a fine-grained differential privacy protection method, while guarantee the utility of social networks. Finally, the scheme is evaluated analytically, and demonstrated experimentally on the real- world data, which reflects its effectiveness and utility.

Due to dramatically increasing information published in social networks, privacy issues have given rise to public concerns. Although the presence of differential privacy provides privacy protection with theoretical foundations, the trade-off between privacy and data utility still demands further improvement. However, most existing works do not consider the impact of the adversary in the measurement of data utility. In this paper, we firstly propose a personalized differential privacy based on social distance. Then, we analyze the maximum data utility when users and adversaries are blind to the strategy sets of each other. We formulize all the payoff functions in the differential privacy sense, which is followed by the establishment of a Static Bayesian Game. The trade-off is calculated by deriving the Bayesian Nash Equilibrium. In addition, the in-place trade-off can maximize the user' data utility if the action sets of the user and the adversary are public while the strategy sets are unrevealed. Our extensive experiments on the real-world dataset prove the proposed model is effective and feasible.