Echo of Neighbors: Privacy Amplification for Personalized Private Federated Learning with Shuffle Model.

Federated Learning, as a popular paradigm for collaborative training, is vulnerable against privacy attacks. Different privacy levels regarding users' attitudes need to be satisfied locally, while a strict privacy guarantee for the global model is also required centrally. Personalized Local Differential Privacy (PLDP) is suitable for preserving users' varying local privacy, yet only provides a central privacy guarantee equivalent to the worst-case local privacy level. Thus, achieving strong central privacy as well as personalized local privacy with a utility-promising model is a challenging problem. In this work, a general framework (APES) is built up to strengthen model privacy under personalized local privacy by leveraging the privacy amplification effect of the shuffle model. To tighten the privacy bound, we quantify the heterogeneous contributions to the central privacy user by user. The contributions are characterized by the ability of generating "echos" from the perturbation of each user, which is carefully measured by proposed methods Neighbor Divergence and Clip-Laplace Mechanism. Furthermore, we propose a refined framework (S-APES) with the post-sparsification technique to reduce privacy loss in high-dimension scenarios. To the best of our knowledge, the impact of shuffling on personalized local privacy is considered for the first time. We provide a strong privacy amplification effect, and the bound is tighter than the baseline result based on existing methods for uniform local privacy. Experiments demonstrate that our frameworks ensure comparable or higher accuracy for the global model.

Contrastive Active Learning Under Class Distribution Mismatch.

Active learning(AL) has been successful based on the premise that labeled and unlabeled data come from the same class distribution. However, its performance undergoes a severe deterioration under class distribution mismatch, wherein the unlabeled data contain numerous instances out of the class distribution of labeled data. In this article, we solve this practical yet rarely studied problem by minimizing the AL error, which is formally defined and decomposed as the valid query error and invalid query error. Specifically, the invalid query error is associated with the queries from unknown categories, and the valid query error is attributed to less informative queries from target categories. In light of this discovery, we propose a contrastive AL framework, named ConAL, to simultaneously learn the semantics and distinctiveness of the instances by contrastive techniques, thereby reducing the invalid query error and valid query error, respectively. Theoretically, we prove that the AL error of ConAL has a tight upper bound. Experimentally, ConAL achieves superior performance on two benchmark datasets, CIFAR10 and CIFAR100, and a cross-dataset with class distribution across multi-datasets. Furthermore, we validate that the ConAL technique performs admirably even on the realistic dataset. To the best of our knowledge, ConAL is the first AL work for class distribution mismatch.

Inaccurate-Supervised Learning With Generative Adversarial Nets.

Inaccurate-supervised learning (ISL) is a weakly supervised learning framework for imprecise annotation, which is derived from some specific popular learning frameworks, mainly including partial label learning (PLL), partial multilabel learning (PML), and multiview PML (MVPML). While PLL, PML, and MVPML are each solved as independent models through different methods and no general framework can currently be applied to these frameworks, most existing methods for solving them were designed based on traditional machine-learning techniques, such as logistic regression, KNN, SVM, decision tree. Prior to this study, there was no single general framework that used adversarial networks to solve ISL problems. To narrow this gap, this study proposed an adversarial network structure to solve ISL problems, called ISL with generative adversarial nets (ISL-GANs). In ISL-GAN, fake samples, which are quite similar to real samples, gradually promote the Discriminator to disambiguate the noise labels of real samples. We also provide theoretical analyses for ISL-GAN in effectively handling ISL data. In this article, we propose a general framework to solve PLL, PML, and MVPML, while in the published conference version, we adopt the specific framework, which is a special case of the general one, to solve the PLL problem. Finally, the effectiveness is demonstrated through extensive experiments on various imprecise annotation learning tasks, including PLL, PML, and MVPML.

Relative Fuzzy Rough Approximations for Feature Selection and Classification.

Fuzzy rough set (FRS) theory is generally used to measure the uncertainty of data. However, this theory cannot work well when the class density of a data distribution differs greatly. In this work, a relative distance measure is first proposed to fit the mentioned data distribution. Based on the measure, a relative FRS model is introduced to remedy the mentioned imperfection of classical FRSs. Then, the positive region, negative region, and boundary region are defined to measure the uncertainty of data with the relative FRSs. Besides, a relative fuzzy dependency is defined to evaluate the importance of features to decision. With the proposed feature evaluation, we propose a feature selection algorithm and design a classifier based on the maximal positive region. The classification principle is that an unlabeled sample will be classified into the class corresponding to the maximal degree of the positive region. Experimental results show the relative fuzzy dependency is an effective and efficient measure for evaluating features, and the proposed feature selection algorithm presents better performance than some classical algorithms. Besides, it also shows the proposed classifier can achieve slightly better performance than the KNN classifier, which demonstrates that the maximal positive region-based classifier is effective and feasible.

The experimental optimization and comprehensive environmental risk assessment of heavy metals during the enhancement of sewage sludge dewaterability with ethanol and Fe(â ¢)-rice husk.

The optimal concentrations of ethanol, Fe3+ and rice husk (RH) to enhance sludge dewaterability were determined by response surface methodology (RSM). Results showed the optimal concentrations of ethanol, Fe3+ and RH were 22.2 g/g DS, 239.9 mg/g DS and 348.9 mg/g DS, respectively, and the CST reduction efficiency reached 72.3%. The transformation behavior and mechanism of the heavy metals (HMs) during conditioning process were determined in terms of total HMs content, leaching tests, and fraction distribution. The environmental risk of HMs was quantitatively evaluated after conditioning in terms of bioavailability and ecotoxicity, potential ecological risks, and pollution levels. Results showed that the high ecological risk of HMs in raw sludge cake is primarily dominated by Cd and the use of Fe3+ alone negatively affected the immobilization of HMs and reduction of leaching toxicity. However, after repeated conditioning with Fe3+ and ethanol, the total HMs content reduction values in sludge cake were 75%, 93%, 100%, 91%, and 74% for Pb, Cr, Cd, Zn, and Cu, respectively. The potential ecological risk index (PERI) and geoaccumulation indicated low or no overall environmental risk after repeated conditioning. Particularly, the risk of Cd was reduced from high risk to low risk after repeated conditioning according to the PERI. Ethanol/Fe3+-RH can effectively reduce HMs risk from the sludge cake in the dewatering tests.

Comparative study on the mobility and speciation of heavy metals in ashes from co-combustion of sewage sludge/dredged sludge and rice husk.

The co-combustion of sludge (sewage and dredged sludge) with rice husk is expected to become a trend because of its economic and environmental benefits. However, the massive residues from the co-combustion process and the mobility of heavy metals (HMs) warrant special attention. The basic performance and environmental properties of the trace elements (Cr, Cu, Fe, Mn, Ba and Zn) from the co-combustion ashes were studied to promote the further utilization of these materials. These ashes have a shell particle shape, high specific area, high amorphous content and low crystalline phase content. The investigation mainly focused on the environmental properties of these ashes to evaluate the risk of these by-products to the environment. Results show Cu, Mn, and Zn have cumulative leaching concentrations of 1.033, 23.32, and 3.363 mg/L for W, by contrast, Cr, Cu, Fe, Mn, Ba, and Zn have cumulative leaching concentrations of 0.488, 0.296, 8.069, 10.44, 2.568, and 2.691 mg/L for H, which are much greater than the Chinese ground water standard (GB/T14848-93). Meanwhile Mn, Zn, Ba, Cr, and Fe all pose a very high risk for H, while Cu only poses a medium risk, and all HMs in W exhibit much lower contamination levels than those in H by the method of risk assessment code (RAC). It indicates that these ashes have undesirably high levels of HMs that demonstrate high mobility and pose environmental risks according to their leachability and chemical speciation. And the HMs in W show lower mobility and environmental hazards than those in H.

Collusion-aware privacy-preserving range query in tiered wireless sensor networks.

Wireless sensor networks (WSNs) are indispensable building blocks for the Internet of Things (IoT). With the development of WSNs, privacy issues have drawn more attention. Existing work on the privacy-preserving range query mainly focuses on privacy preservation and integrity verification in two-tiered WSNs in the case of compromisedmaster nodes, but neglects the damage of node collusion. In this paper, we propose a series of collusion-aware privacy-preserving range query protocols in two-tiered WSNs. To the best of our knowledge, this paper is the first to consider collusion attacks for a range query in tiered WSNs while fulfilling the preservation of privacy and integrity. To preserve the privacy of data and queries, we propose a novel encoding scheme to conceal sensitive information. To preserve the integrity of the results, we present a verification scheme using the correlation among data. In addition, two schemes are further presented to improve result accuracy and reduce communication cost. Finally, theoretical analysis and experimental results confirm the efficiency, accuracy and privacy of our proposals.