The widespread and inconsistent compression applied by Online Social Networks severely degrades the performance of synthetic image detectors. We attribute this degradation to two main issues: 1) the model confuses forgery artifacts with compression artifacts, and 2) compression erodes crucial discriminative high-frequency details. Existing methods suppress compression features during training but overlook the overlap between compression features and forgery-related features, leading to the unintended removal of forgery traces. To address artifact confusion, we introduce a Decision-Driven Orthogonal Constraint, which defines a classification decision axis pointing from the real class centroid to the forged class centroid. This constraint enforces compression artifacts to be orthogonal to the decision axis, mitigating their interference with forgery detection without entirely removing them, thus preventing the suppression of forgery-related features. To mitigate the erosion of high-frequency details, we propose to mine complementary forgery cues from both low-frequency information and compressed high-frequency components. A bidirectional update strategy and an adaptive global-local modulator are proposed to facilitate the utilization of forgery cues. Extensive experiments demonstrate that our method achieves state-of-the-art generalization performance in challenging open-world detection scenarios.

Due to the inherent imbalance in real-world datasets, naïve Empirical Risk Minimization (ERM) tends to bias the learning process towards the majority classes, hindering generalization to minority classes. To rebalance the learning process, one straightforward yet effective approach is to modify the loss function via class-dependent terms, such as re-weighting and logit-adjustment. However, existing analysis of these loss-oriented methods remains coarse-grained and fragmented, failing to explain some empirical results. After reviewing prior work, we find that the properties used through their analysis are typically global, i.e., defined over the whole dataset. Hence, these properties fail to effectively capture how class-dependent terms influence the learning process. To bridge this gap, we turn to explore the localized versions of such properties i.e., defined within each class. Specifically, we employ localized calibration to provide consistency validation across a broader range of losses and localized Lipschitz continuity to provide a fine-grained generalization bound. In this way, we reach a unified perspective for improving and adjusting loss-oriented methods. Finally, a principled learning algorithm is developed based on these insights. Empirical results on both traditional ResNets and foundation models validate our theoretical analyses and demonstrate the effectiveness of the proposed method.

The Area Under the Receiver Operating Characteristics Curve (AUC) is a widely used metric for evaluating model performance across all possible decision thresholds. Existing methods for AUC optimization typically assume a predefined parametric distribution of thresholds. However, the optimal decision threshold depends on the misclassification costs, which follow a non-parametric distribution.This motivates us to introduce a variant of AUC, termed Cost-aware AUC (CAUC), where the thresholds are conditioned on an empirically determined cost distribution. Unfortunately, as a bilevel problem, it is challenging to directly optimize the CAUC: 1) The inner problem of finding the optimal thresholds is non-convex, leading to potential issues with convergence; 2) The outer problem involves the derivative of False Positive Rate (FPR) w.r.t. the threshold, which is unavailable without an explicit formulation of threshold distribution. To address challenge 1), we utilize the convex relaxation technique to reshape the inner problem into a convex one. Facing challenge 2), we propose an adaptive kernel density estimation framework. Specifically, the derivative of FPR is considered an aggregation of various kernel functions. To avoid manually crafting the aggregation function, we propose a finite-difference-based stochastic algorithm to optimize the model without explicit aggregation function. Theoretically, the proposed algorithm enjoys a convergence rate of $\mathcal {O}(\epsilon ^{-4})$O(ε-4). Empirical studies across various datasets and cost distributions speak to the effectiveness and soundness of our framework.

Visual object tracking is essentially crucial for unmanned aerial vehicles (UAVs). Despite the substantial progress, most of the existing UAV trackers are designed for well-conditioned daytime data, while for the scenarios in challenging weather condition, e.g. foggy or nighttime environment, the tremendous domain gap leads to significant performance degradation. To address this issue, in this paper, we propose a novel robust UAV tracker termed LVPTrack, which conducts high quality label-aligned visual prompt tuning to adapt to various challenging weather conditions. Specifically, we first synthesize the sequential foggy and nighttime video frames to assist the model training. A domain adaptive teacher-student network is utilized to distill the hierarchical visual semantic of the target objects in cross-domain scenarios. Then we propose a target-aware pseudo-label voting (PLV) strategy to alleviate the target-level misalignment in the dual domains. Furthermore, we propose a dynamic aggregated prompt (DAP) module to facilitate the appearance variation adaptation of the target object in challenging scenarios. Extensive experiments demonstrate that our tracker achieves superior performance over existing state-of-the-art UAV trackers.

Real-world datasets often follow a long-tailed distribution, making generalization to tail classes difficult. Recent methods resorted to long-tail variants of Sharpness-Aware Minimization (SAM), such as ImbSAM and CC-SAM, to improve generalization by flattening the loss landscape. However, these attempts face a trade-off between computational efficiency and control over the loss landscape. On the one hand, ImbSAM is efficient but offers only coarse control as it excludes head classes from the SAM process. On the other hand, CC-SAM provides fine-grained control through class-dependent perturbations but at the cost of efficiency due to multiple backpropagations. Seeing this dilemma, we introduce Focal-SAM, which assigns different penalties to class-wise sharpness, achieving fine-grained control without extra backpropagations, thus maintaining efficiency. Furthermore, we theoretically analyze Focal-SAM's generalization ability and derive a sharper generalization bound. Extensive experiments on both traditional and foundation models validate the effectiveness of Focal-SAM.

Existing synthetic image detection approaches can be categorized into three paradigms: spatial, frequency, and fingerprint-based methods. Our analysis reveals a fundamental commonality across these paradigms: a significant reliance on high-frequency image components. This observation highlights the discriminative power of high-frequency information for this task and provides a strong rationale for learning generalized artifact representations based on multi-modal fusion strategies. Building on this insight, we introduce a multi-modal high-frequency interactive detection framework for general synthetic image detection. This framework explicitly integrates high-frequency information from both the spatial and frequency domains. Specifically, its spatial processing branch incorporates a novel high-frequency self-enhancement module to bolster local high-frequency representations. Concurrently, the frequency processing branch utilizes a multi-scale frequency information enhancement module to capture diverse contextual cues. At the feature fusion stage, we propose a pooling-guided cross-modal high-frequency interaction module, which dynamically weights cross-modal information to further reinforce salient high-frequency representations. Extensive experiments on public datasets demonstrate that our proposed framework achieves state-of-the-art performance in real-world detection scenarios.

ABSTRACT The transaction processing capacity of blockchain systems remains a critical barrier to adoption in real‐time applications. Recent studies have explored different optimization techniques, including sharding, off‐chain processing, and hybrid consensus algorithms. However, most of those techniques change the original architecture or process of the blockchain and may raise compatibility issues. Resolving these challenges calls for creative methods that can effectively balance transaction throughput with latency without compromising blockchains' core infrastructures. This paper proposes a learning to prediction framework combining a Kalman filter and artificial neural network for transaction throughput forecasting, integrated with a fuzzy logic controller embedded in smart contracts. The approach can dynamically optimize transaction traffic flow based on the predicted throughput and the observed transaction latency, thus improving blockchain performance in real‐time. Deployed on a hyperledger fabric healthcare testbed and evaluated through a series of ablation experiments, the results demonstrate a significant improvement over the baseline and therefore illustrate the potential of the proposed approach in improving blockchain performance for practical applications.

With the rapid development of Internet of Things (IoT) technology, interconnectivity between devices has become increasingly widespread. However, traditional IoT security measures struggle to cope with increasingly complex security threats and cannot fully exploit the advantages of interconnectivity due to the limited computational resources of the devices. To address this, we propose a blockchain-based IoT security framework comprising wallet component, smart contract component, multilayer security component, and common component. This framework, designed for resource-constrained environments and embedded into cellular communication modules, enables multiend offloading of computational tasks and secure transmission for IoT devices. Experimental results show that the data processing capability of the decentralized network architecture based on this framework is improved by 115.06% compared to traditional methods, enhances the security and autonomy of IoT devices, and significantly strengthens the degree of IoT decentralization. This provides a valuable reference for designing next-generation IoT security architectures.

Millimeter wave (MmWave) has become the most promising candidate to enable multi-Gbps transmission and high-precision detection due to the large available bandwidth. Because of the serious propagation attenuation, it is necessary for a vehicle-to-infrastructure (V2I) system communicating in the mmWave band to overcome severe path loss by utilizing beamforming technology. However, swift and accurate beam alignment at the transceivers is challenging when considering user mobility and fast-varying wireless environment. In this paper, we propose an Adaptive Discounted Thompson Sampling (ADTS) based beam alignment algorithm without any prior information such as coarse user location information, which ignores historical observations made beyond the past time periods and avoids misleading for the current state. Besides, we conduct performance analysis to determine the achievable performance bound of the proposed algorithm. Using simulation results, we show that our proposed algorithm achieves good performance in terms of the average effective achievable rate and the beam alignment accuracy without prior knowledge.

Abstract Patient enrollment, data sharing, and data privacy are enormous medical challenges for clinical trial studies. In recent years, blockchain technology has drawn the attention of various researchers and institutes. As a new and innovative distributed ledger technology, blockchain can be critical to addressing these challenges, thus making clinical research transparent and building public trust fairly and openly. However, the existing literature lacks a comprehensive survey on the adoption of blockchain in clinical trials. To fill the research void, this paper presents a punctilious taxonomy of blockchain technology in clinical trials according to the literature. This taxonomy comprises decentralized scenarios, decentralized practices, blockchain types, deployment methods, and consensus algorithms. The results show that blockchain technology can cover all aspects of the clinical trial study in a decentralized, secure, transparent manner. Besides, some open research challenges of blockchain are categorized into three groups: technical challenges, security challenges, and organizational challenges. Moreover, some recent blockchain projects, micro applications in clinical trials, and several research areas or technologies for future research and development are discussed.

When numerous RFID readers are placed in the same area, they may interfere with each other due to the reader collision problem. In recent years, many studies have been presented to address the reader collision problem. However, there is no consonance on the interference model to use in the analysis of the protocols. The main adopted models are the single interference model, which is simple and fast, but only considers the readers within a threshold distance, and the additive interference model, which sums the interferences of all the concurrent interrogations. Recent studies have shown that the single interference model cannot detect a relevant part of the possible collisions detected by the additive one. This paper analyzes and compares the network performance of an RFID system by applying both the models. Considering two proposed scenarios, the performance of the two models are evaluated and presented.

The path of the information flow in Wireless Sensor Networks (WSNs) depends on the application and the particular requirements of the network. While tree-based routing algorithms are able to cope with most of the demands of periodic monitoring, a flooding protocol is necessary when the information should reach every node of the network. Opportunistic flooding is a novel protocol to flood information in WSNs which considers unreliable wireless links. It achieves good performance concerning flooding delay and energy consumption while exploiting low-duty-cycling protocols. In this work, a deep evaluation and characterization of the opportunistic flooding protocol is presented considering reliable and unreliable transmission schemes. Performance analysis is evaluated for the two mechanisms based on acknowledgments and probabilistic retransmissions. Simulation results under multiple scenarios show the behavior of the two mechanisms in different environments while providing a frame of reference to choose one scheme based on the application requirements.