The explosive growth of multimodal web data demands communication that transmits meaning rather than raw bits. Existing semantic-communication systems often fail under noise, missing modalities, and distribution shifts because they optimize surface features instead of modality-invariant knowledge. We present Grasp, a knowledge-centric framework for cross-modal communication. Grasp segments streams into semantic blocks and builds a graph over them; a lightweight Graph Neural Networks (GNN) produces schedulable, importance-weighted representations. At its core is knowledge purification : we minimize a conditional mutual information upper bound to perform a three-way disentanglement—strongly related, weakly related, and task-irrelevant components—so that only essential semantics are transmitted while non-essential factors are suppressed. To maintain synchrony, we introduce one-to-two temporal contrastive learning to achieve triple alignment of video, audio, and text despite sampling asynchrony. For efficient transmission, Grasp uses a cross-modal shared vector-quantization codebook—a discrete knowledge codebook —updated by multimodal attention. At the receiver, a soft-recovery mechanism leverages this shared knowledge to robustly reconstruct semantics under low signal-to-noise ratio (SNR) or missing modalities, yielding graceful degradation. Across web tasks—including cross-modal retrieval and missing-modality inference—Grasp improves knowledge consistency, semantic fidelity, and downstream performance over strong baselines while maintaining low latency. These results show that communication structured around purified knowledge is key to building robust, semantic-aware systems for the modern web.

Recommender systems (RSs), as crucial components of online services, can help users efficiently obtain information they may like. In reality, RSs face long-term threats. Attackers manipulate recommendation results by injecting malicious data in order to obtain benefits. At present, research on the security of RSs lacks a comprehensive understanding of attack capabilities. Moreover, existing defense strategies have not yet been systematically associated with attack characteristics. More importantly, existing defense methods rarely focus on real unlabeled data in practical application scenarios for anomaly detection and forensics. Therefore, this survey systematically analyzes the security of RSs and provides new insights. Specifically, we first categorize attack models from an attack perspective into: attack strategies based on targets, attack strategies against security and privacy, attack strategies based on prior knowledge, and attack strategies against other RSs. From a perspective of defense, existing detection models, second, can be divided into: behavioral representation based on statistics, detection based on hidden features, detection against privacy attacks, anomaly discovery based on association mining, and abnormality forensics for real-world data. Finally, we propose several potential research directions aimed at providing guidance for the security research of RSs. Additionally, to facilitate experimental reproducibility and comparative research, this survey also provides a repository of resources for attacks and defenses (https://github.com/xiaofengbbb/RS-Papers).

With the increasing availability of large-scale spatiotemporal data from location-based services, trajectory publishing has become essential for data-driven analysis and intelligent applications. However, insufficient protection of trajectory location data may result in the disclosure of user privacy and social relationship information. To address this issue, we propose a semantic-aware privacy-preserving trajectory data publishing scheme (SAPP). First, a sliding-window algorithm is employed to extract stay points as key semantic locations and to generate a uniformly sampled set of candidate obfuscation points. Then, a semantic-aware scoring function is designed to probabilistically select candidate points that preserve semantics while avoiding sensitive regions. Furthermore, SAPP computes the sensitivity of each location based on semantic frequency and dynamically allocates the privacy budget. Finally, random noise is added to candidate trajectories using the Laplace mechanism. Through a dual-perturbation mechanism, spatial correlations in sensitive regions are weakened. Security analysis and experimental results further demonstrate that, compared with existing approaches, SAPP reduces TPPS and SFRR by up to 18% and 14%, respectively, indicating stronger resistance against trajectory inference and semantic leakage attacks while maintaining high data utility and time efficiency.

Truth discovery has emerged as an effective tool to mitigate data inconsistency in crowdsensing by prioritizing data from high-quality responders. While local differential privacy (LDP) has emerged as a crucial privacy-preserving paradigm, existing studies under LDP rarely explore a worker's participation in specific tasks for sparse scenarios, which may also reveal sensitive information such as individual preferences and behaviors. Existing LDP mechanisms, when applied to truth discovery in sparse settings, may create undesirable dense distributions, provide insufficient privacy protection, and introduce excessive noise, compromising the efficacy of subsequent non-private truth discovery. Additionally, the interplay between noise injection and truth discovery remains insufficiently explored in the current literature. To address these issues, we propose a lOcally differentially private truth diSCovery approach for spArse cRowdsensing, namely OSCAR. The main idea is to use advanced optimization techniques to reconstruct the sparse data distribution and re-formalize truth discovery by considering the statistical characteristics of injected Laplacian noise while protecting the privacy of both the tasks being completed and the corresponding sensory data. Specifically, to address the data density concerns while alleviating noise, we design a randomized response based Bernoulli matrix factorization method BerRR. To recover the sparse structures from densified, perturbed data, we formalize a 0-1 integer programming problem and develop a sparse recovery solving method SpaIE based on implicit enumeration. We further devise a Laplacian-sensitive truth discovery method LapCRH that leverages maximum likelihood estimation to re-formalize truth discovery by measuring differences between noisy values and truths based on the statistical characteristic of Laplacian noise. Our comprehensive theoretical analysis establishes OSCAR's privacy guarantees, utility bounds, and computational complexity. Experimental results show that OSCAR surpasses the state-of-the-arts by at least 30% in accuracy improvement.

With the wide application of deep learning (DL) across various fields, deep joint source–channel coding (DeepJSCC) schemes have emerged as a new coding approach for image transmission. Compared with traditional separated source and CC (SSCC) schemes, DeepJSCC is more robust to the channel environment. To address the limited sensing capability of individual devices, distributed cooperative transmission is implemented among edge devices. However, this approach significantly increases communication overhead. In addition, existing distributed DeepJSCC schemes primarily focus on specific tasks, such as classification or data recovery. In this paper, we explore the wireless semantic image collaborative nonorthogonal transmission for distributed edge networks, where edge devices distributed across the network extract features of the same target image from different viewpoints and transmit these features to an edge server. A two‐view distributed cooperative DeepJSCC (two‐view‐DC‐DeepJSCC) with or without information disentanglement scheme is proposed. In particular, the two‐view‐DC‐DeepJSCC with information disentanglement (two‐view‐DC‐DeepJSCC‐D) is proposed for achieving balancing performance between multitasking of image semantic communication; while the two‐view‐DC‐DeepJSCC without information disentanglement only pursues outstanding data recovery performance. Through curriculum learning (CL), the proposed two‐view‐DC‐DeepJSCC‐D effectively captures both common and private information from two‐view data. The edge server uses the received information to accomplish tasks such as image recovery, classification, and clustering. The experimental results demonstrate that our proposed two‐view‐DC‐DeepJSCC‐D scheme is capable of simultaneously performing image recovery, classification, and clustering tasks. In addition, the proposed two‐view‐DC‐DeepJSCC has better recovery performance compared to the existing schemes, while the proposed two‐view‐DC‐DeepJSCC‐D not only maintains a competitive advantage in image recovery but also has a significant improvement in classification and clustering accuracy. However, the proposed two‐view‐DC‐DeepJSCC‐D will sacrifice some image recovery performance to balance multiple tasks. Furthermore, two‐view‐DC‐DeepJSCC‐D exhibits stronger robustness across various signal‐to‐noise ratios.

With the rapid development of geographic location technology and the explosive growth of data, a large amount of spatial data is outsourced to the cloud server for reducing the local high storage and computing burdens, but at the same time causes security issues. Thus, extensive privacy-preserving spatial data query schemes have been proposed. Most of the existing schemes use Asymmetric Scalar-Product-Preserving Encryption (ASPE) to encrypt data, but ASPE has proven to be insecure against known plaintext attack. And the existing schemes require users to provide more information about query range and thus generate a large amount of ciphertexts, which causes high storage and computational burdens. To solve these issues, based on enhanced ASPE designed in our conference version, we first propose a basic Privacy-preserving Spatial Data Query (PSDQ) scheme by using a new unified index structure, which only requires users to provide less information about query range. Then, we propose an enhanced PSDQ scheme (PSDQ$^+$) by using Geohash-based$R$-tree structure (called$GR$-tree) and efficient pruning strategy, which greatly reduces the query time. Formal security analysis proves that our schemes achieve Indistinguishability under Chosen Plaintext Attack (IND-CPA), and extensive experiments demonstrate that our schemes are efficient in practice.