Privacy-preserving localization in crowdsourcing has drawn much attention recently. Under the classical nonadjacent subtraction-based localization (NSL) model, existing solutions based on homomorphic encryption techniques are of high computational and communication overheads. In this paper, an adjacent subtraction-based localization (ASL) model is first proposed. Then, an efficient privacy-preserving localization (EPPL) algorithm is developed under ASL without using any homomorphic encryption technique. In terms of the correctness, privacy, and efficiency, a comprehensive analysis is presented to investigate EPPL's performance. Furthermore, the statistical equivalence between ASL and NSL is proved through the fact that the difference between their average location estimation results converges toward zero. The lower and upper bounds of the localization error are also derived for ASL under a bounded noise model. Extensive simulations are conducted to illustrate the equivalence between ASL and NSL, and the performance of EPPL regarding the correctness, privacy, and efficiency.

This paper investigates the problem of participant selection considering colluding attacks for crowdsourcing. Compared with existing work, a practical vulnerability-induced colluding attack model is considered, where the malicious participants with different vulnerability levels collude with each other to perform attacks, which makes the participant selection problem more challenging. To address this problem, according to the structural characteristics of the colluding attack model, we derive the necessary condition and sufficient condition of achieving the colluding possibility minimization of the selected participants. A novel resilient participant selection algorithm is developed based on the necessary condition and sufficient condition. The proposed algorithm can select participants to complete tasks with the time complexity $O(mn)$ , where $m, n$ are the numbers of participants and tasks, respectively. We analyze that the proposed algorithm achieves the colluding possibility minimization to defeat the vulnerability-induced colluding attack. It is also analyzed that the social cost of the proposed algorithm is smaller than existing algorithms in terms of the task-performing cost and the potential damage of the colluding attacks. Extensive real-world trace-based simulations are conducted to demonstrate the effectiveness of the proposed algorithm and the correctness of the theoretical results.

Model quantization has drawn much attention for federated learning (FL) over the Internet of Things (IoT) since it is an effective way to address the critical bottleneck of communication efficiency. State-of-the-art studies have generally assumed homogeneous model quantization, where all clients’ updates are quantized using the same number of bits and aggregated with the same weight at the server. However, in practical IoT scenarios, various IoT devices may apply heterogeneous quantization bits due to their different hardware capabilities, which leads to heterogeneous model quantization accuracy. This article addresses the problem of heterogeneous quantization bit allocation and aggregation for FL. The clients may be allocated with a different number of quantization bits, subject to a total quantization bit constraint. The server may employ different aggregation weights to each IoT device. In particular, we propose FedHBAA—FL with Heterogeneous Bit Allocation and Aggregation, a novel joint quantization bit allocation and server aggregation algorithm. In FedHBAA, we first develop an optimal server aggregation scheme under any given bit allocation among clients. By minimizing the drift term in the convergence rate analysis, a closed-form aggregation weight solution as a function of the allocated bits is obtained for both the strongly convex and nonconvex loss functions. Then, by solving the derived optimal aggregation weights, the optimal bit allocation scheme is obtained. Numerical experiments demonstrate that FedHBAA outperforms the traditional FedAVG algorithm with equal quantization bits.

Federated Learning (FL) has drawn much attention for distributed system over the Internet of Things (IoT), since it enables collaborative machine learning on heterogeneous devices while resolves concerns about privacy leakage. Due to the catastrophic forgetting phenomenon of optimization methods, existing federated learning approaches are restricted to single task learning and typically assume that data from all nodes are simultaneously available during training. However, in practical IoT scenarios, the data preparation from nodes may be asynchronous, and different tasks require incremental training. To address the issues, we propose a continual federated learning (CFL) framework with parameter-level importance aggregation (PI-Fed), which supports collaborative task-incremental learning with privacy preservation. Specifically, PI-Fed evaluates the importance of each parameter in the global model to all history tasks, which is computed locally and aggregated at the center server. Then the server performs soft-masking on the averaged gradient collected from local clients based on the parameter importance. By minimizing the change on important parameters, PI-Fed effectively overcomes catastrophic forgetting and also achieves high efficiency without experience replay. Extensive experiments on 4 benchmarks with at most 20 sequential tasks demonstrate that our proposed PI-Fed significantly outperforms traditional FL baselines (FedAvg, FedNova and SCAFFOLD).