Abstract. Multislice ptychography (msPtycho) extends conventional two-dimensional ptychographic phase retrieval to thick or strongly scattering specimens by modeling the object as a stack of slices and accounting for multiple scattering. While this improves physical fidelity, existing reconstruction algorithms lack rigorous theoretical guarantees and often suffer from either slow convergence or high computational complexity. In this work, we propose a fast and theoretically grounded framework for msPtycho. We formulate a constrained optimization model that exploits the successive dependence of adjacent exit waves, reducing the multilinear coupling into a chain of bilinear relations. Building on this structure, we develop an alternating minimization of msPtycho (AM[Formula: see text]SP), which admits closed-form updates for all subproblems and is proven to converge globally to stationary points. To further enhance the performance for reconstructing deeply layered specimens, we introduce a forward-propagating correction to suppress error accumulation across slices and an extrapolation strategy to accelerate convergence. Together, these yield the accelerated AM[Formula: see text]SP-FX algorithm. Extensive numerical experiments demonstrate that the proposed methods achieve faster convergence and better reconstructions. Notably, AM[Formula: see text]SP-FX achieves superior reconstruction quality and consistently faster per-iteration runtimes than both layer-wise optimization and 3D Ptychographical Iterative Engine (3PIE); in particular, it requires an order of magnitude less time per iteration than 3PIE. Overall, the proposed framework offers a practical and theoretically sound solution for high-fidelity multislice ptychographic imaging.

A key tenet of 6G is a network of networks, upon which sophisticated services can be composed in real-time; this service composition can be realized through decentralized 6G marketplaces of services. These services may include high fidelity real-time holograms, as well as sensitive medical applications, and hence users will have stringent privacy expectations of these services and on the trustworthiness of the constituent service providers and the marketplaces themselves. Rather than specify the details of the required trust in each constituent service provider/marketplace, users will want to specify their requirements through high-level intents on the functionality, price, and trust of their desired composite services. Cross-chain technologies, together with reputation frameworks, provide accountability as an underpinning of trust in such decentralized marketplaces. In this paper, we present an intent-based architecture for cross-chain marketplaces of 6G services, where trust is a first-class factor. Building on our earlier work, we describe several design and implementation approaches for such intent-based service composition integrated with Hyperledger Fabric and Hyperledger Cacti blockchain/cross-chain technologies, and evaluate the comparative benefits of these approaches with respect to key aspects of intent fulfillment and intent assurance. Our work is in the context of UNEXT™, an intelligent networking platform being created at Nokia Bell Labs.

Machine unlearning allows data owners to erase the impact of their specified data from trained models. Unfortunately, recent studies have shown that adversaries can recover the erased data, posing serious threats to user privacy. An effective unlearning method removes the information of the specified data from the trained model, resulting in different outputs for the same input before and after unlearning. Adversaries can exploit these output differences to conduct privacy leakage attacks, such as reconstruction and membership inference attacks. However, directly applying traditional defenses to unlearning leads to significant model utility degradation. In this article, we introduce a Compressive Representation Forgetting Unlearning scheme (CRFU), designed to safeguard against privacy leakage on unlearning. CRFU achieves data erasure by minimizing the mutual information between the trained compressive representation (learned through information bottleneck theory) and the erased data, thereby maximizing the distortion of data. This ensures that the model's output contains less information that adversaries can exploit. Furthermore, we introduce a remembering constraint and an unlearning rate to balance the forgetting of erased data with the preservation of previously learned knowledge, thereby reducing accuracy degradation. Theoretical analysis demonstrates that CRFU can effectively defend against privacy leakage attacks. Our experimental results show that CRFU significantly increases the reconstruction mean square error (MSE), achieving a defense effect improvement of approximately 200% against privacy reconstruction attacks with only 1.5% accuracy degradation on MNIST.

Manifold learning is essential for analyzing high-dimensional data, but it suffers from high time complexity. To address this, researchers proposed using anchors and constructing a similarity matrix to expedite eigen decomposition and reduce sparse consumption. However, randomly selected anchors fail to represent the data well, and using K-means for anchor generation is time-consuming. In this paper, we introduce Granular-ball (GB) into unsupervised manifold learning, presenting GB-USC and GB-USEC. By employing a coarse-to-fine approach, GB-USC generates high-quality anchors aligned with the data distribution. A bipartite graph is constructed between data points and anchors, enabling low-dimensional manifold embedding using transfer cut. GB-USEC combines multiple GB-USC clusters, generating consistent low-dimensional embeddings across dimensions and determining clustering results through voting. The experimental results show that compared with the state-of-the-art algorithm U-SPEC, GB-USC achieves the similar performance with the average running time of GB-USC is 33.96% less than that of U-SPEC for several million-level datasets. Additionally, our ensemble algorithm improves the clustering efficiency by an average of 29.19% compared with U-SENC.

Federated learning (FL) requires clients to train constituted models based on their local datasets. Clients usually directly train local models using their entire datasets without distinguishing which information of data is task-relevant or irrelevant. Task-irrelevant information does not contribute to the learning task but exposes additional privacy information to adversaries. Studies have shown that unintended information leakage from gradients during FL iterations threatens clients' privacy. Researchers applied differential privacy (DP) to protect clients' gradients, but it does not help to reduce task-irrelevant information from the gradients. In this paper, we propose a compressive privacy federated learning (CP-FL) scheme to protect the task-irrelevant information from gradient leakage attacks. In CP-FL, clients train a local compressive model according to the global task. The local compressive model constructs a new representation, which extracts task-relevant and removes task-irrelevant information from clients' data. Since the global model is updated based on the compressed representation that eliminates the task-irrelevant information, it can effectively prevent adversaries from inferring those property values from the uploaded gradients. Moreover, with the help of a powerful local compressive model that sanitizes the challenging data into a low-dimension space representation, CP-FL can use a small global model instead of a sizeable one, significantly reducing communication. Both theoretical analysis and extensive experimental results demonstrate that CP-FL can effectively defend against gradient leakage attacks while maintaining practical utility.

Federated learning (FL) has become a hot research domain due to its privacy protection for model collaboratively training in edge computing systems. However, recent studies indicated that most FL algorithms have desperately suffered from backdoor attacks. Although many backdoor defence FL algorithms were proposed, their effects were highly related to the ratio of malicious clients (RMC) of all participated edge nodes. To be more specific, most of them only set RMC around 10% to 30% in their experiments, and their results also showed that the rate of successful backdoor defence seriously drops when RMC increases. In the paper, we propose a novel federated learning scheme with mode connectivity (FedMC) to defend against backdoor attacks, mitigating the sharp defence effect degradation as RMC increases. Conventional mode connectivity mainly focuses on training a connecting curve between two end models, which is inapplicable in distributed multiple clients FL situations. We extend the two-ends mode connectivity to multi-ends by introducing a scalable regularization term consisting of the edge clients' models to involve their knowledge in the connective model training. In each communication round, the FL-Server aggregates and absorbs the contribution of clients by training a connective model based on a small set of clean samples, which builds a pathway to accurately connect all edge clients' models and mitigates the backdoor triggers of models. Extensive experiments and results demonstrate that FedMC can effectively defend against backdoor attacks while maintaining the accuracy on untampered test data.

Federated learning (FL) is an emerging solution for machine learning model training in edge/fog computing systems. Unlike traditional systems that collect and train models on clouds, FL allows multiple edge/fog nodes to train a global model collaboratively without revealing their local data to clouds. Compared with traditional systems, it is inherited with better privacy protection ability. Although the basic privacy protection is inherited in FL, the privacy leakage from shard models is still unsolved. Existing solutions attempt to enhance the privacy of shared model parameters by adding differential privacy (DP) noise. However, these solutions all suffer from accuracy loss and convergence problems owing to the injected noise. In this paper, we propose a novel federated learning protocol to solve the above problem. The model trained on a carefully selected sampling subset can achieve the same level privacy protection as DP while preserving the model accuracy. Experimentally, we proved that our protocol achieves better model accuracy in the same privacy guarantee compared with noise injecting DP methods.

The fifth-generation (5G) cellular network is expected to provide submeter positioning accuracy without draining the battery of user equipment (UE). As a solution, ultradense network (UDN) deployment and network-based positioning were proposed. However, the openness of UDN and the vulnerability of network devices [e.g., access nodes (ANs)] make it easy for attackers to poison such a positioning system. However, no existing work explores how to overcome this issue. This article concentrates on jamming and collusion attacks in the network-based positioning system. Specifically, we design a novel scheme that contains three functional modules to erase the influence of these attacks. A truth discovery module applies a clustering-based method aiming to generate the most approximate position value and find out suspicious signals. Based on neural network models, we further develop an attack detection module and an attack tracing module to perceive attacked UE and locate malicious or attacked ANs. Through simulation, we conduct extensive experiments to illustrate the effectiveness of our scheme. The result shows high detection and tracing accuracy with very simple neural network models, which also implies the potential of our proposed scheme in practical deployment.

High-accuracy positioning has drawn huge attention with the potential in enhancing location-aware communications, intelligent transportation, and so on. The emergency of the fifth-generation (5G) technologies, such as device-to-device (D2D) communications, vehicle-to-vehicle (V2V) communications, and crowdsourcing networks is expected to help achieve highly accurate positioning. By employing nearby mobile terminals to estimate position cooperatively, these technologies can improve positioning accuracy effectively, especially in indoor and urban areas. Despite the benefit, the potential information disclosure in these positioning systems threatens the engagement of public participants (also known as reference points). The location of the reference points and their distances to a target point is quite sensitive since they can be easily used to locate the reference points once exposed. Though existing solutions based on Paillier homomorphic encryption have been proposed to preserve the privacy of distance information. The sensitivity of reference points’ locations is ignored. Additionally, the adoption of Paillier introduces a high computation cost, which is impractical in reality. To address the above problems, this article proposes two efficient protocols, named Pub-pos and Pri-pos. By leveraging matrix concatenation and multiplication, these two protocols can disguise the original sensitive data, including both distance and location information, into a random matrix while keeping a positioning result intact. We analyze the security strength, complexity, and optimal variable selection of the proposed protocols. Numerous experiments verify that our proposed protocols have significant efficiency improvement in both system and individual levels compared with a Paillier-based solution.

As the sensitivity of position, the privacy protection in both 5G positioning and its further application in location-based services (LBSs) has been paid special attention and studied. Solutions based on k-anonymity, homomorphic encryption, and secure multi-party computation have been proposed. However, these solutions either require a trusted third party or incur heavy overheads. Besides, there still lacks an integrated solution that can protect privacy for both positioning and LBS provision. Based on Intel SGX, this article proposes a novel light-weight scheme that can protect privacy in both 5G positioning and its further applications in LBS provision in an integrated way. Through secret sharing, the proposed scheme can also support multiple location-based service providers without frequent key exchange. We seriously analyze the security of our scheme. Based on scheme implementation, its efficiency is proved through the performance evaluation conducted over a real-world database.

Device-to-Device (D2D) cooperative location verification allows a device to verify its location with the help of neighbouring devices. It is especially handy in location-based services where location verification is essential. However, the exposure of device location during verification rises a big privacy risk for participants since they have to send their real-time locations to unknown verifiers holding anonymous identities. Thus, a privacy-preserving solution is urgently needed to provide verification without location disclosure. Traditional solutions based on Paillier and garbled circuits can solve the problem but also introduce high cost. Based on order-preserving encryption, we propose an efficient protocol with high-security guarantee to address this issue. Apart from rigorous security proof, complexity analysis and extensive experiments are also conducted to evaluate the proposed solution. The results compared with related work show that order-preserving encryption based mechanism achieves the best balance with regards to privacy, utility and performance requirements.

Edge computing has been widely adopted in many systems, thanks for its advantages to offer low latency and alleviate heavy request loads from end users. Its integration with indoor positioning is one of promising research topics. Different from a traditional positioning system where a user normally query remotely deployed positioning services provided by a Location Information Service Provider (LIS), LIS will outsource its service to an edge device, and the user can obtain the service by directly accessing the edge device in an edge computing-based system. Though the benefits from edge computing, there is still some open issues for service outsourcing. One of them is how to ensure that the outsourced service is executed honestly by the edge device. However, the current literature has not yet seriously studied this issue with a feasible solution. In this paper, we design a verification scheme to solve this open problem for indoor positioning based on edge computing. By injecting some specially designed dataset into a trained machine learning based positioning model, the functionality of outsourced model on edge devices can be verified through this dataset with regard to its prediction accuracy from outsourced model. The verification is successful only when the prediction accuracy can pass a threshold. In experiments, we provide extensive empirical evidence using state-of-the-art positioning models based on real-world datasets to prove the effectiveness of our proposed scheme and meanwhile investigate the effects caused by different factors.

Abstract The unprecedented proliferation of mobile smart devices has propelled a promising computing paradigm, Mobile Crowd Sensing (MCS), where people share surrounding insight or personal data with others. As a fast, easy, and cost-effective way to address large-scale societal problems, MCS is widely applied into many fields, e.g., environment monitoring, map construction, public safety, etc. Despite the popularity, the risk of sensitive information disclosure in MCS poses a serious threat to the participants and limits its further development in privacy-sensitive fields. Thus, the research on privacy protection in MCS becomes important and urgent. This paper targets the privacy issues of MCS and conducts a comprehensive literature research on it by providing a thorough survey. We first introduce a typical system structure of MCS, summarize its characteristics, propose essential requirements on privacy on the basis of a threat model. Then, we survey existing solutions on privacy protection and evaluate their performances by employing the proposed requirements. In essence, we classify the privacy protection schemes into four categories with regard to identity privacy, data privacy, attribute privacy, and task privacy. Besides, we review the achievements on privacy-preserving incentives in MCS from four viewpoints of incentive measures: credit incentive, auction incentive, currency incentive, and reputation incentive. Finally, we point out some open issues and propose future research directions based on the findings from our survey.