Blind document image inpainting (DII) aims to restore degraded document scans without prior knowledge of noise locations, yet existing methods either produce over-smoothed text or introduce pseudo-characters. We propose a novel high-fidelity, text-guided restoration framework based on a ControlNet-guided Latent Diffusion Model (CGLDM). First, we extract and refine noisy OCR outputs using a two-stage Vision Language Model (VLM) and Large Language Model (LLM) pipeline, leveraging both global document context and local text cues to deliver near ground-truth textual fidelity. Next, these refined text features, together with an initial visually restored image, condition a latent diffusion process that progressively denoises and reconstructs clean image latents. To suppress patch-wise background inconsistencies inherent in high-resolution processing, we introduce an explicit feature-alignment loss of diffusion model training that enforces agreement between the predicted features and the VAE-encoded features of the ground-truth image. Extensive experiments on the FUNSD-ZH dataset demonstrate that our approach outperforms state-of-the-art methods in OCR legibility and maintains competitive image-level fidelity.

Wearables are increasingly popular and are being accepted in healthcare applications, such as monitoring activity and heart rate. Like other wireless sensors, time synchronization is an important issue to ensure the credibility of data. While studies in this area are not uncommon, most methods try to synchronize the clock on individual devices and require access to the hardware and modification to the firmware. We consider a case for consumer wearables, where there is zero access to both firmware and hardware of the devices, except listening to the streaming data passively, and all synchronization can only be done on the receiving end. We present a two-part approach to synchronize streaming data received from wearable sensors. First, the fixed time offset due to device processing time is corrected by a one-time calibration. Second, the random time offset due to wireless communication is corrected by finetuning the sampling rate. Instead of adjusting the time on separate devices, the method attempts to align data from different devices to the accuracy of one sample time. The synchronization method also considers situations of data loss, congested Bluetooth channels, and multiple receiving hubs. With further interpolation on the received data, a synchronization accuracy of 1 ms is achieved. Unlike existing methods that rely on embedded timestamps, synchronized start events, or firmware modifications, our technique operates entirely on the receiver side, achieving single sample accuracy even under packet loss and clock drift.

Cuffless Blood Pressure (BP) monitoring using photoplethysmography (PPG) is attractive for continuous wearable assessment, however, it is highly susceptible to motion artifacts and poor signal quality. We propose a simple and interpretable multi-stage signal-quality (SQ) framework that embeds three sequential quality checks directly into the PPG → BP estimation pipeline: (1) raw-signal metrics (signal jitters, zero crossings), (2) beat-level plausibility (peak-to-peak amplitude and inter-beatinterval changes), and (3) multi-sensor consistency (pulse arrival time variability). Each detected pulse is labelled good / bad, aggregated over a one-minute window, and gated by a tunable Good Signal Threshold (GST). GST was calibrated on a small test set, and GST$=0.7$was chosen because it retained 99 % of stationary windows while rejecting 90 % of motion-contaminated windows. In a feasibility study with subjects undergoing oneday measurements of 24 hours, PPG-derived heart rate (HR) was compared against ECG ground truth. Ablation analysis showed incremental benefits from each SQ stages: during sleep, error reduced from 2.08 bpm to 1.22 bpm with 80 % coverage; while during daytime activity, error decreased from 3.94 bpm to 1.69 bpm, albeit with reduced coverage of 50 %. These results confirm that the SQ framework substantially improves pulse fidelity, achieving$40-55 \%$error reduction, while maintaining interpretable design and low computational cost. The proposed approach is thus well-suited for low-power wearable systems and provides a practical foundation for cuffless BP monitoring.

The advent of Cellular Vehicle-to-Everything (CV2X) technology has revolutionised intelligent transportation systems (ITS), but poses challenges for secure and efficient data sharing due to its dynamic nature. Traditional centralised systems are inadequate, prompting the need for decentralised solutions like blockchain. However, applying blockchain technologies in C-V2X always faces scalability issues. This paper proposes a scalable C-V2X blockchain network with a hierarchical consensus by integrating a dynamic load-balancing sharding mechanism and zero-knowledge proofs (ZKPs). Our scheme ensures scalability in the C-V2X environment through sharding while utilising ZKPs to enhance cross-shard validation efficiency, reducing its complexity to$O(1)$. Additionally, our approach reduces bandwidth consumption by 90.8% compared to Merkle tree-based solutions and its consensus time is lower than 360 ms.

Blockchain technology has gained traction in Food Supply Chain Management (FSCM), enhancing traceability and transparency. The existing deployments of public or private blockchains face issues in achieving an optimal balance between transparency and decentralization. This work proposes a hybrid blockchain model complemented by an Access Control (AC) mechanism to bolster security, reliability, and usability within FSCM systems. Furthermore, the integration of a recommender system is proposed to utilize data analytics and machine learning for personalizing product offerings and optimizing inventory management, aiming to boost efficiency and consumer satisfaction. The synergy between the hybrid blockchain framework and the recommender system is anticipated to cultivate a more engaged, efficient, and gratified supply chain ecosystem. The model significantly enhances monitoring in 30% of the use cases and supports transparency in a quarter. It also reduces vulnerability cases by 20%. Inventory management is markedly improved, reducing overstock by 25%, confirming the effectiveness of the proposed hybrid blockchain approach.

Document Image Inpainting (DII) has been applied to degraded documents, including financial and historical documents, to enhance the legibility of images for: (1) human readers by providing high visual quality images; and (2) machine recognizers such as Optical Character Recognition (OCR), thereby reducing recognition errors. With the advent of Deep Learning (DL), DL-based DII methods have achieved remarkable enhancements in terms of either human or machine legibility. However, focusing on improving machine legibility causes visual image degradation, affecting human readability. To address this contradiction, we propose an adaptive learning-based DII method, namely ALDII, that applies domain adaptation strategy, our approach acts like a plug-in module that is capable of constraining a total feature space before optimizing legibility of human and machine, respectively. We evaluate our ALDII on a Chinese handwritten character dataset, which includes single-character and text-line images. Compared to other state-of-the-art approaches, experimental results demonstrated superior performance of our ALDII with metrics of both human and machine legibility.

The ultra-dense network (UDN) is considered a vital technology for 5G mobile communications due to its ability to transmit high data rates in high-traffic environments. However, it also creates new challenges, such as increased interference and difficulty managing mobility. To ensure seamless base station connectivity and maintain a high quality of service, a reliable handover algorithm is necessary, especially in a UDN where the cell size is small. This paper proposes an optimisation method for handover parameters based on the Deep Deterministic Policy Gradient (DDPG) algorithm. It adjusts the handover margin (HOM) to determine the handover trigger points accurately and dynamically. Simulation results indicate that the system’s mobility performance has been greatly improved while maintaining high throughput and low latency at different speeds.

File transfer is always handled by a separate service, sometimes it is a third-party service in videoconferencing. When sending files during a video session, file data flow and video stream are independent of each other. Encryption is a mature method to ensure file security. However, it still has the chance to leave footprints on the intermediate forwarding machines. These footprints can indicate that a file once passed through, some protocol-related logs give clues to the hackers' later investigation. This work proposes a file-sending scheme through the video stream using blockchain and steganography. Blockchain is used as a file slicing and linkage mechanism. Steganography is applied to embed file pieces into video frames that are continuously generated during the session. The scheme merges files into the video stream with no file transfer protocol use and no extra bandwidth consumed by the file to provide trackless file transmission during the video communication.

Android compilers play a crucial role in Android app development. The correctness of the apps relies on the compilers because the source code of the app is translated into the target language by the compilers. The use of obfuscators is becoming the standard in app development to prevent reverse engineering or code tampering. Despite their importance, both compilers and obfuscators lack an oracle, which is the mechanism to determine the correctness of the execution, and hence they can be called untestable software. Metamorphic Testing (MT) is a state-of-the-art testing method that can test untestable software. MT tests software based on Metamorphic Relations (MRs). Recent studies have shown that program transformation, an MT-based compiler-testing strategy, is highly effective in revealing bugs in compilers. However, this strategy requires sophisticated tools that could take significant time to develop. Therefore, program transformation using obfuscators is proposed. Based on research into testing obfuscators using MT, it is suggested that an MT-based compiler-testing strategy could be achieved by using obfuscators. In addition, this method has the potential to detect bugs in both compilers and obfuscators. This paper reports on our experience using MT techniques to test compilers and obfuscators. We present three related MRs, two of which uncover evidence of faults.

Diabetes, a chronic condition with a growing global prevalence, exerts lasting effects on individuals' health and well- being, necessitating continuous control and monitor of blood glucose for stable levels. Meeting this fact, recent years have seen an increasing adoption of machine learning algorithms to accurately predict blood glucose values. In this work we present a novel multi-model approach, BUMS (Balanced Multi-model Scheme), designed to accurately predict blood glucose levels in real time. The primary goal of this system is to mitigate the risks associated with critical blood glucose events, such as hypoglycemia and hyperglycemia, which significantly impact individuals living with diabetes. BUMS combines three distinct algorithms: Long Short-Term Memory (LSTM), Random Forest (RF), and Extreme Gradient Boosting (XGBoost), all leveraging continuous glucose monitoring data collected at 5-minute intervals. To ensure robustness and balance in the predictive models, we introduce a pre-trained Balancer into the multi-model architecture. Our approach is validated using data from the publicly available DirectNet dataset, featuring continuous blood glucose measurements from 30 patients. The Balancer module is pre-trained on data from 5 patients before being tested on data from the remaining 25 patients, employing Linear Regression as its foundation. We evaluate the performance of our system across various prediction horizons, ranging from 25 to 855 minutes, using 169 test cases. The results demonstrate an overall Root Mean Square Error (RMSE) of 4.8125, indicating the model's high predictive accuracy. Notably, among the 169 test cases, only one case was incorrectly identified, resulting in an accuracy rate of 96.29% in detecting hypoglycemic events.

Named Data Network (NDN) advocates the philosophy of accessing IoT data owing to its location independence feature. This enables routers to pre-cache content and serves the future requests for the same content on a local basis. Such architecture demonstrates huge application potential in the E-health field. In order to achieve efficient healthcare treatment and administration for both patients and medical professions, the optimization of storing patients’ real-time, large-scale physical data is of necessity. We propose aDynamicMulti-AttributeCaching mechanism forNDN-Based remote health monitoring system (DMACN). In our model, we adopted a predictable consumer-driven freshness mechanism with low computation cost to satisfy the freshness-sensitive nature of health data. A novel content popularity model based on Analytic Hierarchy Process (AHP) and medical-grade parameters are proposed to handle the doctor-decision-making-coupled Interest sending mechanism of a remote health monitoring system. The final simulation results show DMACN has strong robustness against intensive requesting and complex contents. It is also shown that its performance surpasses existing mechanisms. TheCache Hit Ratioexceeds 37.5% to FIFO and LRU, 220% compared with CPFC, and 55% for CTDICR for both consumer and producer tests. On the other hand, theLatencyis about 23.8% lower than FIFO and LRU, 46.6% lower compared with CPFC, and 35.4% for CTDICR.

Nowadays, social media platforms play a significant role in real-world events, which can cause both positive and negative effects. The popularity of image-based content on social media has been dramatically increased, which brings the problem that the quality of content is rather spotty. Therefore, automated techniques of identifying fake images have drawn significant attention. The traditional detection methods focus on the elements’ consistency of the image, which requires massive computing resources and huge datasets for pairs of real and fake images. Many studies on detecting rumors on social media showed that there are propagation patterns for the spreading of fake content that can be used as clues of detection. Thus, the proposed approach attempts to characterize the propagation patterns of fake images on social media using several user features and tweet features. The detection model applies recurrent neural networks to capture the variation of suggested features along the propagation path over time. The results of the experiment on a Weibo dataset of image tweets show that the model can achieve 89% accuracy in classifying fake images from real ones. Moreover, the model already reaches high performance as the detection deadline is smaller than 24 h, which demonstrates the strong capability of early detection. The positive outcomes indicate that the proposed detection model has great potential to be further developed to an automated technique that can be used in classifying real images from fake images posted on social media.

Temperature detection aiming for Covid-19 prevention is widely in demand from 2020. A system combining Covid-19 detection and authentication plays a significant role in ensuring safety and security. Nevertheless, after reviewing the literature, it is found that most of such systems existing contain two devices including a visible light camera for recognition and a thermal camera for temperature detection. In this paper, we propose an integrated single device framework based on a thermal camera, which combines authentication and Covid-19 detection using thermal infrared imagery. It features customizing machine learning models for face recognition, temperature detection with thermal images captured by the thermal camera and sending warnings to the house owner remotely if necessary. The whole work is carried out from two aspects: framework design and simulation. The framework designing part concerns modules involved and allocating the modules into edge devices and servers. For the software implementation part, work is generally divided into the initialization phase for training the machine learning model and the authentication phase for face recognition and temperature detection.

Since the end of 2019, the world has been caught in the crisis of the COVID-19 which is a serious epidemic disease. This paper seeks to come up with a fast and efficient COVID-19 detection and monitoring easy to use system which can be used in the facilities of densely populated areas, such as community centers and school clinics, to quickly identify suspected COVID-19 patients. This system could detect the probability of a person getting infected by COVID-19 using an android smartphone and thermal camera. Three types of data are collected from users: breathe sound, thermal video, and health status. Generally, the breathe audio and thermal video are preprocessed into two-time series, which indicate the breath status of the user. Then, the two series are inputted into the Bidirectional Gated Recurrent Unit (BI-GRU) neural network model separately to get the infection rates. Since the real data is difficult to get due to privacy reasons, a synthetic dataset is generated based on mathematical equations to train the model. For health status, the application requires the user to fill a questionnaire and calculates an infection rate through a medical prediction model. Finally, the two values from the machine learning model and the infection rate from the user report are added together with weight to calculate the final predictive infection rate.

The main reason which makes epidemics so dangerous and difficult to contain is their highly infectious nature. In the case of Covid-19 also, data shows that its contagious nature is increasing along with its various mutant strains. One of the primary methods adopted to fight the pandemic has been to break the infection chain and thus reduce the rate of persons getting infected every day, through lockdowns, self-isolation, social distancing, and other measures. But although there are already many existing epidemic models, to predict and track the spread of the disease, it is evident from the difference in the rates of infection and fatalities in different countries, that a uniform set of parameters is not sufficient to accurately predict the curves. In this paper, we have suggested some additional benchmarks that could be considered and at a higher granularity for more accurate predictions at more local levels. We also propose an IoT-based framework for the collection of such types of data through smartphones for more consolidated information to be made available to the authorities, for the effective management of epidemics. The framework also issues warnings to other users through smartphones if the app detects the presence of a potentially infected person within close range.

IoT based remote health monitoring systems play an important role to provide a technological solution for health monitoring and reach essential medical services to the patients. However, a review of the existing literature reveals that the existing remote health monitoring systems rarely consider difficulty of interoperability among devices in the system. In this paper, we propose a Named Data Networking (NDN) based remote health monitoring system, which involves the functions of measuring and combining patient’s health data, real-time named based data transmission between gateway and health center, data analysis at hospital side, and overall the realization of two-way communication between doctors and patients. The whole work is carried out from three aspects: simulation, hardware implementation, client platform building (App development). Simulation part focuses on the construction of the whole system in the virtual NDN network environment on the ndnSIM platform. The hardware part implements system construction and basic functions in reality. For the client platform, we build a mobile application, which achieves the functionality of online communication between doctors and patients.

In recent times, COVID-19 is the most severe epidemic disease and it needs to be controlled as soon as possible. Promising ideas and mathematical models have been proposed to predict the number of infected people in a particular timeline and project its development tendency. In this paper, to further increase the accuracy of prediction, we propose a new model named AMSD model at an agent scale by combining three models that are widely used in this field: the social network model, the mobility model, and the Susceptible-Exposed-Infected-Recovered (SEIR) model. Initially, the mobility model could identify people's mobility patterns on weekdays and weekends, and during the day and at night. Then, by combining this model with the social network model, we could classify people by their social connections in the network, with a more accurate prediction of infected people. The basic SEIR model is enhanced to find the spread/growth of viruses between and within people and has four stages from susceptible to recovered. AMSD model, as the combination of these three models is a more comprehensive approach to better present and predict the propagation of COVID-19, which involves many more important social factors.

Internet of Things (IoT) is a network of heterogeneous physical devices connected over the Internet. Each of the devices is capable of collecting and processing data. Due to the connection with the Internet, the IoT devices become more susceptible to attacks by malicious nodes, which may result in privacy loss and security breaches. Thus, network security is necessary for the privacy of transmitted messages. In this context, we propose a scheme, Service-Oriented game-theoretic Security (SOS), which provides a simple yet robust security solution for IoT networks. Here, we have amalgamated our scheme with Named Data Networking (NDN), which is more of a data content-specific approach, unlike the traditional IP address search. In this scheme, at first, the hop count between the sender and the receiver is used to generate the public key to encrypt the messages by the sender. When the receiver receives this message, it decrypts the message with the help of the decryption function generated by the sender using the hop count between them as the private key. A non-cooperative Stackelberg game-theoretic model is used to model defenders and attackers, which helps to decide strategies to maximize the payoff (profit) of the defenders to protect the network from malicious attacks. The results are further extended for a modified public key encryption technique, which results in the robustness of the security scheme to be used for all real-life network scenarios. Simulation results show that the proposed scheme, SOS, has a better performance compared to the existing state-of-the-art security schemes, UAKMP and CLS, in terms of time complexity, message overhead, throughput, and attack probability.

Mobile edge caching is an emerging approach to manage high mobile data traffic in fifth‐generation wireless networks that reduces content access latency and offloading data traffic of backhaul links. This paper proposes a novel cooperative caching policy based on long short‐term memory (LSTM) neural networks considering the characteristics between the features of the heterogeneous layers and the user moving speed. Specifically, LSTM is applied to predict content popularity. Size‐weighted content popularity is utilised to balance the impact of the predicted content popularity and content size. We also consider the moving speeds of mobile users and introduce a two‐level caching architecture consisting of several small base stations (SBSs) and macro base stations (MBSs). To avoid content requests of fast‐moving users affecting the content popularity distribution of the SBS since fast‐moving users frequently handover among SBSs, fast‐moving users are served by MBSs no matter which SBS they are in. SBSs serve low‐speed users, and SBSs in the same cluster can communicate with one another. The simulation results show that compared to common cache methods, for example, the least frequently used and least recently used methods, our proposed policy is at least 8.9% lower and 6.8% higher in terms of the average content access latency and offloading ratio, respectively.

Named Data Networking (NDN) is a content-centric networking, where the publisher of the packet signs and encapsulates the data packet with a name-content-signature encryption to verify the authenticity and integrity of itself. This scheme can solve many of the security issues inherently compared to IP networking. NDN also support mobility since it hides the point-to-point connection details. However, an extreme attack takes place when an NDN consumer newly connects to a network. A Man-in-the-middle (MITM) malicious node can block the consumer and keep intercepting the interest packets sent out so as to fake the corresponding data packets signed with its own private key. Without knowledge and trust to the network, the NDN consumer can by no means perceive the attack and thus exposed to severe security and privacy hazard. In this paper, the N ame-Signature Lookup System (NSLS) and corresponding Name-Signature Lookup Protocol (NSLP) is introduced to verify packets with their registered genuine publisher even in an untrusted network with the help of embedded keys inside Network Interface Controller (NIC), by which attacks like MITM is eliminated. A theoretical analysis of comparing NSLS with existing security model is provided. Digest algorithm SHA-256 and signature algorithm RSA are used in the NSLP model without specific preference.

In this work, the problem of topology control for self-adaptation in stationary Wireless Sensor Networks (WSNs) is revisited, specifically for the case of networks with a subset of nodes having temporary connection impairment between them. This study focuses on misbehaviors arising due to the presence of\enskip “dumb” nodes [Misra et al. 2014; Roy et al. 2014a, 2014b, 2014c; Kar and Misra 2015], which can sense its surroundings but cannot communicate with its neighbors due to shrinkage in its communication range by the environmental effects attributed to change in temperature, rainfall, and fog. However, a dumb node is expected to behave normally on the onset of favorable environmental conditions. Therefore, the presence of such dumb nodes in the network gives rise to impaired connectivity between a subset of nodes and, consequently, results in change in topology. Such phenomena are dynamic in nature and are thus distinct from the phenomena attributed to traditional isolation problems considered in stationary WSNs. Activation of all the sensor nodes simultaneously is not necessarily energy efficient and cost-effective. In order to maintain self-adaptivity of the network, two algorithms, named Connectivity Re-establishment in the presence of Dumb nodes ( CoRD ) and Connectivity Re-establishment in the presence of Dumb nodes Without Applying Constraints ( CoRDWAC ), are designed. The performance of these algorithms is evaluated through simulation-based experiments. Further, it is also observed that the performance of CoRD is better than the existing topology control protocols—LETC and A1—with respect to the number of nodes activated, overhead, and energy consumption.

Faults in a gas pipeline network is one of the major impairments towards the safe gas distribution among consumers. So, it is significant to detect and locate the faults in a pipeline network. In this work, a novel sensor based fault localization scheme, FLoGPN is proposed for detection and localization of faults in a gas distribution network. Here leak in gas pipeline network is considered as fault. This scheme uses Josang's Beta Reputation model to combine information from the sensors to select the nearest sensor to the fault. The aim is to identify the nearest sensor to the fault that indicates the faulty pipe segment in the whole network. The method is tested in the presence of noise to check its reliability and the simulation results show that the scheme can localize fault in a gas pipeline network accurately with the SNR value 20 dB or more.

In this work, we propose a scheme, named CoRAD , for the reestablishment of lost connectivity using sensor nodes with adjustable communication range in stationary wireless sensor networks (WSNs), when “dumb” behavior occurs some of the nodes. Due to the occurrence of such behavior, there may be temporary loss of connectivity between among the nodes. Such a phenomenon is different from the commonly known node isolation problem in stationary WSNs. The mere activation of intermediate sleep nodes cannot guarantee reestablishment of connectivity, because there may not exist neighbor nodes of the isolated nodes. On the contrary, the increase in communication range of a single sensor node may make it die quickly. Including this, a sensor node has maximum limit of increase in communication range that may not be sufficient to reestablish connectivity. Therefore, considering all these factors for self-organization of the network and isolated node re-connection, we propose a price-based scheme, which addresses the issue by activating intermediate sleep nodes or by adjusting the communication range of some of the other nodes in the network. The scheme also deactivates the additional activated nodes and reduces the increased communication range when the dumb nodes resume their normal behavior, upon the return of favorable environmental conditions. To implement the proposed scheme, CoRAD it is required to construct the network using GPS-enabled adjustable communication range sensor nodes. Through simulation we compare our proposed scheme with the existing topology management schemes -- LETC and A1 -- in the same scenario by considering the number of activated nodes, message overhead, and energy consumption. We find that the proposed scheme shows improved performance compared to the existing topology management schemes.

A collection of spatially distributed sensor nodes in a wireless sensor network (WSN) work collaboratively to sense the physical phenomena around them and then send the sensed information to the sink node through single-hop or multihop paths. In this work, we propose a scheme, named ReDAST, for reliable and efficient data acquisition in a stationary WSN in the presence of transfaulty nodes. Due to the transfaulty behavior, a sensor node gets temporarily isolated from the network. Temporary node isolation leads to the formation of dynamic communication holes in the network, which form and disappear dynamically. Furthermore, they may increase or decrease in size dynamically as well. These effects result in loss of information in the radiation-affected area. To prevent information loss in WSN due to transfaulty behavior of sensor nodes, in the proposed scheme, we construct the network using sensor nodes having dual mode of communication-RF and acoustic. To get redundant coverage within a radiation affected area, all the sensor nodes in the area become activated and switch to the acoustic communication mode after detecting themselves to be affected by radiations. In-network data fusion is performed to get actual information from the redundant information received from the radiation-affected area. Simulation results exhibit that the proposed scheme, ReDAST, achieves better energy efficiency and reduced average end-to-end delay than sensor nodes having only acoustic mode of communication.