Federated Learning (FL) is vulnerable to backdoor attacks by design since it cannot inspect clients’ local data to protect their privacy. This privacy-preserving feature creates an opportunity for malicious clients to introduce backdoors. However, existing backdoor attacks face two main limitations. First, brute amplification (i.e., uniformly scaling up malicious parameters) can be easily detected, hence compromising attack stealthiness. Second, evasion strategies employed to prevent their backdoors from being overwritten by benign updates are frequently ineffective, reducing the overall attack stability upon model deployment. To address these limitations, we propose an adaptive proactive boosting strategy to enhance both the stealthiness and durability of backdoor attacks in FL. As a concrete example,ReBAintroduces a durable importance metric based on stability degrees of parameters as an update mask for malicious attackers, assigning higher weights to backdoor-related parameters during the update process. To ensure stealthiness,ReBAformulates an optimization problem regarding amplification factor by minimizing the distance between malicious and clean updates, thereby correcting malicious updates within a benign distance space. Extensive evaluations on 3 datasets and across 14 defenses demonstrate the efficacy ofReBA, outperforming over 12 baseline backdoor attacks. Our code is available at https://anonymous.4open.science/r/ReBA-D82F.

Video-based visible-infrared person re-identification (VVI-ReID) task focuses on cross-modality retrieval of pedestrian videos, which are captured in visible and infrared modalities by non-overlapping cameras across diverse scenes, and holds significant value for security surveillance scenarios. The challenges of this task mainly stem from three issues: the difficulty of capturing comprehensive spatio-temporal cues, intra-class variations within video sequences, and inter-modality discrepancies between visible and infrared data. Existing methods mainly try to address the modality gap or focus on one of the other aspects, but rarely do they jointly consider these key factors. Motivated by these core challenges, we propose the M3-ReID (Multi-View & Granularity & Modality) method, a unified framework that simultaneously enhances spatio-temporal feature extraction, intra-class discrimination, and cross-modality consistency. Specifically, to capture diverse spatio-temporal patterns, we design a Multi-View Learning module that leverages different spatial and temporal-spatial perspectives to adaptively emphasize diverse key regions and motion cues. To enhance intra-class modeling of each identity, we introduce a Multi-Granularity Representation strategy that optimizes features across both fine-grained frame level and coarse-grained video level by minimizing mutual information among redundant frames while enhancing identity representations. Furthermore, to bridge the visible-infrared gap, we propose a Multi-Modality Alignment mechanism that explicitly aligns metric learning and cross-modality matching goals, transforming features into a unified embedding space with modality consistency and class discrimination. Extensive experiments on benchmark VVI-ReID datasets demonstrate the superiority of our proposed M3-ReID framework against existing methods.

With the recent development of AIoT (AI+IoT), crowdsourcing has emerged as a promising paradigm for distributed problem solving and business practice. Crowdsourcing entails posting tasks on a dedicated Web platform, enabling networked workers to choose preferred tasks on a first-come, first-served basis, typically of the same type to ensure high assignment accuracy. However, existing crowdsourcing task assignment methods do not take into account the potential fatigue of workers for similar tasks. In this article, we propose a task assignment architecture using a (TA-GAE), which comprehensively considers the relationship between the occupation and skills of workers and potential tasks, facilitating an accurate assignment of a wide variety of tasks to workers. The proposed architecture consists of three modules, The Graph Creation module analyzes the potential connections between tasks based on worker evaluations and constructs an initial task graph that represents these connections. The gravity-based graph autoencoder module is inspired by Newton’s law of universal gravitation. We analogize the tasks on the crowdsourcing platform to masses in the universe and calculate the mutual attractive force between two tasks to quantify their correlation. The Hybrid Task Assignment module recommends task lists to workers by combining traditional collaborative filtering and content-based task assignment strategies. The experimental results demonstrate that the proposed architecture outperforms several state-of-the-art methods and achieves a diversity rate of over 40% across four data sets: 1) fliggy trip; 2) MovieLens 1M; 3) library; and 4) survey.

Transformer-based Large Models (TLM), such as generative pre-trained models (GPT), have become increasingly popular for practical applications through Deep Learning as a Service (DLaaS). They have been extensively used in natural language processing and computer vision. However, concerns regarding potential private data leakage arise with this type of inference service. While some private inference techniques can protect privacy, they often introduce high latency and approximate replacements in the design protocols, resulting in changes to the model structure and decreased accuracy. In this research, we present SecureTLM, a private inference method based on secure multi-party computation (MPC) that does not require modifications to the underlying model structure. SecureTLM offers protocols for crucial computations in TLM, such as Multiplication, Softmax, GeLU, and LayerNorm, without altering the model structure. Experimental results demonstrate that SecureTLM ensures data privacy, maintains correctness, and achieves efficiency in private inference tasks.

In recent years, remote depression detection is becoming a new favorite of the Internet of Things because of its promise. However, deployable and sensitive information security has not been properly addressed. Thus, a novel depression detection framework is proposed to successfully solve the above problems. This article concentrates solely on the development and execution of our proposed core algorithm, LKCT. Further details concerning the system’s construction will be disclosed separately as future work. The contribution of this article can be summarized as follows: 1) proposing a depression detection framework that is easy to promote and can effectively solve the security of sensitive information; 2) proposing a novel two-stream deep network LKCT based on the global–local concept; 3) designing a novel network called purified vision transformer (PIT) to enhance the model’s ability to capture local details in images; and 4) introducing decoupled knowledge distillation to distill LKCT into the lightweight network MobileNet for deployability. To evaluate the true performance of our proposed method, we conducted sufficient experiments on the Chinese Academy of Sciences Institute of Automation (CASIA), facial expression recognition challenge (FERC), AVEC2014, and private depression data sets. LKCT achieved an accuracy of 93.4% on CASIA, demonstrating its strong ability to distinguish facial features. The accuracy on FERC reached 99%, proving its ability to accurately distinguish simple emotions. On AVEC2014, the model achieved state-of-the-art results with an RMSE of 7.41 and MAE of 5.48, validating its ability to identify depression. The distilled model obtained an average accuracy of 83% on the private data set, indicating good generalization ability.

Abstract Deep learning techniques have been widely used in natural language processing (NLP) tasks and have made remarkable progress. However, training the deep learning model relies on a large amount of data which may involve sensitive information like electronic medical records. The attacker can infer sensitive information from the model, which leads to privacy leakage. To solve this problem, we propose a Differentially Private Recurrent Variational AutoEncoder (DP-RVAE) that can generate simulated data in place of the sensitive dataset to preserve privacy. To generate high utility synthetic text, a part of sensitive text data is employed as the conditional input of the model and uses a dropout and noise perturbing mechanism to preserve differential privacy. In addition, we expand the proposed DP-RVAE to a federated learning setting and design a novel training paradigm for NLP tasks. Specifically, DP-RVAE is deployed to the client-side to train and generate personalized text. These DP-RVAE models would be aggregated and updated through the Federated Optimisation (FedOPT) algorithm so that personal information can be well preserved. We evaluate our proposed DP-RVAE through a text classification task on the Tweets depression sentiment and IMDB reviews datasets. Our DP-RVAE achieves a higher average test accuracy by 5.90% and 3.94% compared to the typical centralized training and federated learning approach, respectively. We also perform the keywords inference attack experiment on the medical description dataset collected from the real world. Compared to the typical differentially private preserving approach, the DP-RVAE decreases by 15.2% in average attack accuracy. The experimental results demonstrate that DP-RVAE can be applied to the NLP models to leverage accuracy while preserving sensitive privacy.

Cover Caption: The cover image is based on the Research Article Active forgetting via influence estimation for neural networks by Xianjia Meng et al., https://doi.org/10.1002/int.22981.

The rapidly exploding of user data, especially applications of neural networks, involves analyzing data collected from individuals, which brings convenience to life. Meanwhile, privacy leakage in the applications as a potential threat needs to be addressed urgently. However, removing private information from models is difficult once the user's sensitive data enters machine learning models, particularly neural networks. Most of the previous amnestic methods based on retraining require full access to the training set of the target model and have limited improvements in computational resources and time improvement. In this paper, we propose Scrubber, which removes sensitive data from the original model via influence estimation to produce an unlearning model that is approximately indistinguishable from the retrained model. S crubber builds on the essential concept of influence function and reformulates the influence estimation as a closed-form update of forgetting. For learned models with strictly convex loss functions, our approach theoretically guarantees the effectiveness of forgetting while empirically demonstrating forgetting performance. For models with non-convex losses, we relax strictly convex assumptions by applying a damping term that allows us to make approximate estimates with negligible errors from the original assumption. Furthermore, experiments show that S crubber only causes less than 1% and 3% accuracy drop with more than 80% forgetting rate on average for logistic regression models and convolutional neural networks. The accuracy drop is reduced by 2%–3% compared to most state-of-the-art methods.

Using an unmanned aerial vehicle (UAV) to collect data from wireless sensor networks deployed in the field, one of the key tasks is to plan the path for the collection so as to minimize the energy consumption of the UAV. At present, most of the existing methods generally take the shortest flight distance as the optimal objective to plan the optimal path. They simply believe that the shortest path means the least energy consumption of the UAV and ignore the fact that changing direction (heading) can also consume the UAV’s energy in its flight. If the path can be planned based on the UAV’s energy consumption closer to the real situation, the energy consumption of the UAV can be really reduced and its working energy efficiency can be improved. Therefore, this paper proposes a path planning method for UAV-assisted data collection, which can plan an energy-efficient flight path. Firstly, by analyzing the experiment data, we, respectively, model the relationship between the angle of heading change and the energy consumption of the UAV and the relationship between the distance of straight flight and the energy consumption of the UAV. Then, an energy consumption estimation model based on distance and the angle of heading change (ECEMBDA) is put up. By using this model, we can estimate or predict the energy consumption of a UAV to fly from one point (or node) to another (including the start point). Finally, the greedy algorithm is used to plan the path for UAV-assisted data collection according to the above estimated energy consumption. Through simulation and experiments, we compare our proposed method with the conventional method based on pure distance index and greedy algorithm. The results show that this method can obtain data collection path with lower energy consumption and smoother path trajectory, which is more suitable for actual flight.

Video surveillance is an effective way to record current events. In view of the difficulty of efficient transmission of massive surveillance video and the risk of leakage in the transmission process, a new data encryption and fast transmission algorithm is proposed in this paper. From the perspective of events, the constraints of time and space dimension is broken. First, a background and moving object extraction model is built based on video composition. Then, a strong correlation data encryption and fast transmission model is constructed to achieve efficient data compression. Finally, a data mapping mechanism is established to realize the decoding of surveillance video. Our experimental results show that the compression ratio of the proposed algorithm is more than 60% under the premise of image confidentiality.

Wireless sensor networks as the base support for the Internet of things have been a large number of popularity and application. Such as intelligent agriculture, we have to use the sensor network to obtain the growing environment data of crops and others. However, the difficulty of power supply of wireless nodes has seriously hindered the application and development of Internet of things. In order to solve this problem, people use low-power sleep scheduling and other energy-saving methods on the nodes. Although these methods can prolong the working time of nodes, they will eventually become invalid because of the exhaustion of energy. The use of solar energy, wind energy, and wireless signals in the environment to obtain energy is another way to solve the energy problem of nodes. However, these methods are affected by weather, environment, and other factors, and they are unstable. Thus, the discontinuity work of the node is caused. In recent years, the development of wireless power transfer (WPT) has brought another solution to this problem. In this paper, a three-layer framework is proposed for mobile station data collection in rechargeable wireless sensor networks to keep the node running forever, named TLFW which includes the sensor layer, cluster head layer, and mobile station layer. And the framework can minimize the total energy consumption of the system. The simulation results show that the scheme can reduce the energy consumption of the entire system, compared with a Mobile Station in a Rechargeable Sensor Network (MSiRSN).

Traffic estimation is a popular approach to acquire traffic conditions in urban areas. At present, using the traffic data to realize the low-cost traffic estimation has already been widely favored. Although those data include various sensitive element, people ignore the harm caused by information leakage while the data are used. Additionally, the transmission of vehicle data also requires a very large communication bandwidth. To address those problems, we focus on the privacy-preserving vehicle data and reducing the amount of ciphertext data to achieve a city-scale traffic estimation. Meanwhile, we present a novel framework that integrates compressive sensing (CS) technology into privacy- preserving vehicle data. Furthermore, outsourcing vehicle data to the cloud is adopted to overcome the limitations of the in-vehicle sensors. In particular, we present a feasible computational scheme for traffic estimation, further improve the capacity of privacy- preserving and decrease system energy consumption. Finally, we validate the effectiveness of the scheme proposed through the real-world dataset.

Wireless localization technology is a vital component in many long-term monitoring applications, such as activity monitoring and real-time tracking. Most existing localization methods however require the target to carry communicationcapable devices to send or receive messages, which may not hold for wildlife monitoring or intrusion detection. Prior proposals are based on device-free localization techniques, such as Channel State Information (CSI). However, they cost huge human effort in fingerprint collection when locate the target in different scenarios with different area size. This paper proposes a robust and accurate at low-cost devicefree localization system named LiReT. To reduce the time cost and human effort in fingerprint collection when the monitoring environment changed, we represent a LiReT algorithm based on a multivariable linear regression model to transfer the CSI measurements (fingerprint) at distance L to L'. Thus, LiReT can locate the target accurately at low-cost. Result from experiments demonstrate that our system can improve the localization accuracy by up to 51.68%, which is competitive with existing solutions.

Network coding is emerging as a promising technique that can provide significant improvements in the throughput of Internet of Things (IoT). Previous network coding schemes focus on several nodes, regardless of the topology and communication range in the whole network. Consequently, these schemes are greedy. Namely, all opportunities of combinations of packets in these nodes are exploited. We demonstrate that there is still room for whole network throughput improvement for these greedy design principles. Thus, in this paper, we propose a novel network coding scheme, ECS (Efficient Coding Scheme), which is designed to achieve a higher throughput improvement with lower computational complexity and buffer occupancy compared to current greedy schemes for wireless mesh networks. ECS utilizes the knowledge of the topologies to minimize interference and obtain more throughput. We also prove that the widely used expected transmission count metric (ETX) in opportunistic listening has an inherent error ratio that would lead to decoding failure. ECS therefore exploits a more reliable broadcast protocol to decrease the impact of this inherent error ratio in ETX. Simulation results show that ECS can greatly improve the performance of network coding and decrease buffer occupancy.