Transfer learning enhances model performance in financial time series by leveraging data from related domains. The selection of appropriate source domains is crucial to avoid negative transfer. We propose using Gramian Angular Field (GAF) transformations to improve time series similarity functions for better domain alignment. Extensive experiments with DNN and LSTM models show that GAF-based similarity functions, specifically Coral (GAF) for DNN and CMD (GAF) for LSTM, significantly reduce prediction errors, demonstrating their effectiveness in complex financial environments.

The advancement of image editing tools has enabled malicious manipulation of sensitive document images, underscoring the need for robust document image forgery detection.Though forgery detectors for natural images have been extensively studied, they struggle with document images, as the tampered regions can be seamlessly blended into the uniform document background (BG) and structured text. On the other hand, existing document-specific methods lack sufficient robustness against various degradations, which limits their practical deployment. This paper presents ADCD-Net, a robust document forgery localization model that adaptively leverages the RGB/DCT forensic traces and integrates key characteristics of document images. Specifically, to address the DCT traces' sensitivity to block misalignment, we adaptively modulate the DCT feature contribution based on a predicted alignment score, resulting in much improved resilience to various distortions, including resizing and cropping. Also, a hierarchical content disentanglement approach is proposed to boost the localization performance via mitigating the text-BG disparities. Furthermore, noticing the predominantly pristine nature of BG regions, we construct a pristine prototype capturing traces of untampered regions, and eventually enhance both the localization accuracy and robustness. Our proposed ADCD-Net demonstrates superior forgery localization performance, consistently outperforming state-of-the-art methods by 20.79\% averaged over 5 types of distortions. The code is available at https://github.com/KAHIMWONG/ACDC-Net.

The rise of LLMs has increased concerns over source tracing and copyright protection for AIGC, highlighting the need for advanced detection technologies. Passive detection methods usually face high false positives, while active watermarking techniques using logits or sampling manipulation offer more effective protection. Existing LLM watermarking methods, though effective on unaltered content, suffer significant performance drops when the text is modified and could introduce biases that degrade LLM performance in downstream tasks. These methods fail to achieve an optimal tradeoff between text quality and robustness, particularly due to the lack of end-to-end optimization of the encoder and decoder. In this paper, we introduce a novel end-to-end logits perturbation method for watermarking LLM-generated text. By joint optimization, our approach achieves a better balance between quality and robustness. To address non-differentiable operations in the end-to-end training pipeline, we introduce an online-prompting technique that leverages the on-the-fly LLM as a differentiable surrogate. Our method achieves superior robustness, outperforming distortion-free methods by 37–39% under paraphrasing and 17.2% on average, while maintaining text quality on par with the distortion-free methods in terms of text perplexity and downstream tasks. Our method can be easily generalized to different LLMs. Code is available at https://github.com/KAHIMWONG/E2E_LLM_WM.

In financial analysis, time series modeling is often hampered by data scarcity, limiting neural network models’ ability to generalize. Transfer learning mitigates this by leveraging data from similar domains, but selecting appropriate source domains is crucial to avoid negative transfer. This study enhances source domain selection in transfer learning by introducing Gramian Angular Field (GAF) transformations to improve time series similarity functions. We evaluate a comprehensive range of baseline similarity functions, including both basic and state-of-the-art (SOTA) functions, and perform extensive experiments with Deep Neural Networks (DNN) and Long Short-Term Memory (LSTM) networks. The results demonstrate that GAF-based similarity functions significantly reduce prediction errors. Notably, Coral (GAF) for DNN and CMD (GAF) for LSTM consistently deliver superior performance, highlighting their effectiveness in complex financial environments.

With the rapid advancement of intelligent devices and cloud services, a novel edge-cloud computing paradigm is emerging, finding widespread adoption in numerous advanced applications. Despite its considerable convenience and benefits, edge-cloud computing raises security and privacy concerns. Although many cryptographic solutions have been proposed for the Internet of Things and cloud services, ensuring diverse access control in an untrusted edge-cloud environment and realizing flexible revocation and efficient outsourcing remain challenging. In this article, we propose a certificateless attribute-based matchmaking encryption scheme (CRO-ABME) that supports fine-grained bilateral access control, attribute and identity revocation, and cryptographic workload outsourcing. Leveraging CRO-ABME, we design an edge-cloud data sharing system that ensures secure data uploading with privacy protection between end-users, such that only authorized matchers can access the data in edge-cloud computing. Furthermore, rigorous security proofs for CRO-ABME are provided, and experimental analyses demonstrate the efficiency and flexibility of our proposed scheme.

Computing power has evolved into a foundational and indispensable resource in the area of deep learning, particularly in tasks such as Face Recognition (FR) model training on large-scale datasets, where multiple GPUs are often a necessity. Recognizing this challenge, some FR methods have started exploring ways to compress the fully-connected layer in FR models. Unlike other approaches, our observations reveal that without prompt scheduling of the learning rate (LR) during FR model training, the loss curve tends to exhibit numerous stationary subsequences. To address this issue, we introduce a novel LR scheduler leveraging Exponential Moving Average (EMA) and Haar Convolutional Kernel (HCK) to eliminate stationary subsequences, resulting in a significant reduction in converging time. However, the proposed scheduler incurs a considerable computational overhead due to its time complexity. To overcome this limitation, we propose FastFace, a fast-converging scheduler with negligible time complexity, i.e.O(1) per iteration, during training. In practice, FastFace is able to accelerate FR model training to a quarter of its original time without sacrificing more than 1% accuracy, making large-scale FR training feasible even with just one single GPU in terms of both time and space complexity. Extensive experiments validate the efficiency and effectiveness of FastFace. The code is publicly available at: https://github.com/amoonfana/FastFace.

Decentralized Finance (DeFi), propelled by Blockchain technology, has revolutionized traditional financial systems, improving transparency, reducing costs, and fostering financial inclusion. However, transaction activities i n these systems fluctuate significantly and the throughput can be effected. To address this issue, we propose a Dynamic Mining Interval (DMI) mechanism that adjusts mining intervals in response to block size and trading volume to enhance the transaction throughput of Blockchain platforms. Besides, in the context of public Blockchains such as Bitcoin, Ethereum, and Litecoin, a shift towards transaction fees dominance over coin-based rewards is projected in near future. As a result, the ecosystem continues to face threats from deviant mining activities such as Undercutting Attacks, Selfish Mining, and Pool Hopping, among others. In recent years, Dynamic Transaction Storage (DTS) strategies were proposed to allocate transactions dynamically based on fees thereby stabilizing block incentives. However, DTS’ utilization of Merkle tree leaf nodes can reduce system throughput. To alleviate this problem, in this paper, we propose an approach for combining DMI and DTS. Besides, we also discuss the DMI selection mechanism for adjusting mining intervals based on various factors.

Blockchain-enabled smart contracts have revolutionized the insurance industry due to their potential to streamline backend operations, mitigate fraudulent claims, and enhance data security and transparency. Guided by the design science methodology, the authors propose two specific smart contract frameworks to enhance insurance claims processing related to vehicle damage claims and personal injury claims. These proposed frameworks can improve the overall efficiency and effectiveness of insurance claims processing by automating claims submission, review, analysis, and payment, while reducing fraud and data leakage, by merging various data sources and disintermediation. Furthermore, the authors design a smart contract template supported by eight operational algorithms to facilitate the processing of insurance claims with the help of smart contracts. This template provides practitioners with a standardized prototype for the development of secure and efficient insurance applications.

Complex non-convex ad hoc networks (CNCAH) contain intersecting polygons and edges. In many instances, the layouts of these networks are not entirely convex in shape. In this article, we propose a Kamada-Kawai-based algorithm called W-KK-MS for boundary node detection problems, which is capable of aligning node positions while achieving high sensitivity, specificity, and accuracy in producing a visual drawing from the input network topology. The algorithm put forward in this article selects and assigns weights to top- k nodes in each iteration to speed up the updating process of nodes. We also propose a novel approach to detect and unfold stacked regions in CNCAH networks. Experimental results show that the proposed algorithms can achieve fast convergence on boundary node detection in CNCAH networks and are able to successfully unfold stacked regions. The design and implementation of a prototype system called ELnet for analyzing CNCAH networks is also described in this article. The ELnet system is capable of generating synthetic networks for testing, integrating with force-directed algorithms, and visualizing and analyzing algorithms’ outcomes.

Force-directed algorithms such as the Kamada-Kawai algorithm have shown promising results for solving the boundary detection problem in a mobile ad hoc network. However, the classical Kamada-Kawai algorithm does not scale well when it is used in networks with large numbers of nodes. It also produces poor results in non-convex networks. To address these problems, this article proposes an improved version of the Kamada-Kawai algorithm. The proposed extension includes novel heuristics and algorithms that achieve a faster energy level reduction. Our experimental results show that the improved algorithm can significantly shorten the processing time and detect boundary nodes with an acceptable level of accuracy.

Crime specific event patterns are crucial in detecting potential relationships among suspects in criminal networks. However, current link analysis tools commonly used in detection do not utilize such patterns for detecting various types of crimes. These analysis tools usually provide generic functions for all types of crimes and heavily rely on the user's expertise on the domain knowledge of the crime for successful detection. As a result, they are less effective in detecting patterns in certain crimes. In addition, substantial effort is also required for analyzing vast amount of crime data and visualizing the structural views of the entire criminal network. In order to alleviate these problems, an event-based approach to money laundering data analysis and visualization is proposed in this paper. The effectiveness of the proposed method is demonstrated on a money laundering case from Taiwan.

Archive is a collection of historical records which have been accumulated over the course of an individual and institution's lifetime. Archive is considered as an invaluable asset which provides us with an insight into the lives and circumstances of past generations. In this paper, we describe the workflow model of an archival management system adopted by Macau Historical Archives. To further analyze the strength and the weaknesses of the current system, the conceptual workflow model from Macau Historical Archives is mapped with OAIS functional model. As the project is still ongoing, the paper mainly studies the modeling requirements, discusses the difficulties have to overcome, and concludes with our future work.

Departmental workflows within a digital business ecosystem are often executed concurrently and required to share limited number of resources. However, unexpected events from the business environment and delay in activities can cause temporal exceptions in these workflows. Predicting temporal exceptions is a complex task since a workflow can be implemented with various types of control flow patterns. In this paper, we describe a critical path based approach for predicting temporal exceptions in concurrent workflows which are required to share limited resources. Our approach allows predicting temporal exceptions in multiple attempts while workflows are being executed.

Infectious diseases that are caused by pathogenic microorganisms can spread fast and far, from one person to another, directly or indirectly. Prompt quarantining of the infected from the rest, coupled with contact tracing, has been an effective measure to encounter outbreaks. However, urban life and international travel make containment difficult. Furthermore, the length of incubation periods of some contagious diseases like SARS enable infected passengers to elude health screenings before first symptoms appear and thus to carry the disease further. Detecting and visualizing contact–tracing networks, and immediately identifying the routes of infection, are thus important. We apply information visualization and hidden cluster detection for finding cliques of potentially infected people during incubation. Preemptive control and early quarantine are hence possible by our method. Our prototype Infectious Disease Detection and Quarantine Management System (IDDQMS), which can identify and trace clusters of infection by mining patients’ history, is introduced in this paper.

Recent technological advances in electronic commerce have fuelled the need for designing effective strategies for supply chain management. These strategies are essential in guiding various activities within a supply chain such as component acquisitio

Elevators are considered as important transportation systems for urban communities. Elevators are installed with onboard controllers (circuit boards) and these controllers can generate a large volume of signals and events. In this paper, we describe an event-driven system to test, control, and monitor a large number of on board elevator controllers. The integrated system consists of a virtual controller, control and monitoring terminals, a central server, a playback function with animation, a genetic algorithm based maintenance scheduling module, and a data warehouse for managing massive real-time elevator signals. Based on the event-driven architecture, the proposed system is capable of facilitating faster deployment of new types of elevators. The system also provides engineers with playback functions for troubleshooting any hardware or software errors. In order to reduce overhead cost, the proposed system is designed to optimize resource allocation in maintenance scheduling. By deploying data warehouse technology, the proposed system allows significant reduction of storage requirement for managing real-time signals.