Search Results (89)

In real-world applications, user demands for new functionalities and activities constantly evolve, requiring action recognition systems to incrementally incorporate new action classes without retraining from scratch. This class-incremental learning (CIL) paradigm is essential for enabling adaptive and scalable systems that can grow over time. However, Wi-Fi-based indoor action recognition under incremental learning faces two major challenges: catastrophic forgetting of previously learned knowledge and uncontrolled model expansion as new classes are added. To address these issues, we propose CAREC, a class-incremental framework that balances dynamic model expansion with efficient compression. CAREC adopts a multi-branch architecture to incorporate new classes without compromising previously learned features and leverages balanced knowledge distillation to compress the model by 80% while preserving performance. A data replay strategy retains representative samples of old classes, and a super-feature extractor enhances inter-class discrimination. Evaluated on the large-scale XRF55 dataset, CAREC reduces performance degradation by 51.82% over four incremental stages and achieves 67.84% accuracy with only 21.08 M parameters, 20% parameters compared to conventional approaches. Full article

This paper explores the application of deep learning techniques for evaluating and classifying deadlift weightlifting techniques from video input. The increasing popularity of weightlifting, coupled with the injury risks associated with improper form, has heightened interest in this area of research. To address these concerns, we developed an application designed to classify three distinct styles of deadlifts: conventional, Romanian, and sumo. In addition to style classification, our application identifies common mistakes such as a rounded back, overextension at the top of the lift, and premature lifting of the hips in relation to the back. To build our model, we created a comprehensive custom dataset comprising lateral-view videos of lifters performing deadlifts, which we meticulously annotated to ensure accuracy. We adapted the MoveNet model to track keypoints on the lifter’s joints, which effectively represented their motion patterns. These keypoints not only served as visualization aids in the training of Convolutional Neural Networks (CNNs) but also acted as the primary features for Long Short-Term Memory (LSTM) models, both of which we employed to classify the various deadlift techniques. Our experimental results showed that both models achieved impressive F1-scores, reaching up to 0.99 for style and 1.00 for execution form classifications on the test dataset. Furthermore, we designed an application that integrates keypoint visualizations with motion pattern classifications. This tool provides users with valuable feedback on their performance and includes a replay feature for self-assessment, helping lifters refine their technique and reduce the risk of injury. Full article

With the increasing demand for drones in diverse tasks, the Internet of Drones (IoD) has recently emerged as a significant technology in academia and industry. The IoD environment enables various services, such as traffic and environmental monitoring, disaster situation management, and military operations. However, IoD communication is vulnerable to security threats due to the exchange of sensitive information over insecure public channels. Moreover, public key-based cryptographic schemes are impractical for communication with resource-constrained drones due to their limited computational capability and resource capacity. Therefore, a secure and lightweight key agreement scheme must be developed while considering the characteristics of the IoD environment. In 2024, Alzahrani proposed a secure key agreement protocol for securing the IoD environment. However, Alzahrani’s protocol suffers from high computational overhead due to its reliance on elliptic curve cryptography and is vulnerable to drone and mobile user impersonation attacks and session key disclosure attacks by eavesdropping on public-channel messages. Therefore, this work proposes a lightweight and security-enhanced key agreement scheme for the IoD environment to address the limitations of Alzahrani’s protocol. The proposed protocol employs a physical unclonable function and simple cryptographic operations (XOR and hash functions) to achieve high security and efficiency. This work demonstrates the security of the proposed protocol using informal security analysis. This work also conducted formal security analysis using the Real-or-Random (RoR) model, Burrows–Abadi–Needham (BAN) logic, and Automated Verification of Internet Security Protocols and Applications (AVISPA) simulation to verify the proposed protocol’s session key security, mutual authentication ability, and resistance to replay and MITM attacks, respectively. Furthermore, this work demonstrates that the proposed protocol offers better performance and security by comparing the computational and communication costs and security features with those of relevant protocols. Full article

This research introduces Fortified-Edge 2.0, a novel authentication framework that addresses critical security and privacy challenges in Physically Unclonable Function (PUF)-based systems for collaborative edge computing (CEC). Unlike conventional methods that transmit full binary Challenge–Response Pairs (CRPs) and risk exposing sensitive data, Fortified-Edge 2.0 employs a machine-learning-driven feature-abstraction technique to extract and utilize only essential characteristics of CRPs, obfuscating the raw binary sequences. These feature vectors are then processed using lightweight cryptographic primitives, including ECDSA, to enable secure authentication without exposing the original CRP. This eliminates the need to transmit sensitive binary data, reducing the attack surface and bandwidth usage. The proposed method demonstrates strong resilience against modeling attacks, replay attacks, and side-channel threats while maintaining the inherent efficiency and low power requirements of PUFs. By integrating PUF unpredictability with ML adaptability, this research delivers a scalable, secure, and resource-efficient solution for next-generation authentication in edge environments. Full article

The Controller Area Network (CAN) bus has been implemented in most modern Vehicles. Various attacks can be launched against the CAN bus protocol because it is designed without security mechanisms. It is essential to develop a highly accurate intrusion detection system (IDS) for CAN bus attacks. We design an effective ensemble learning-based IDS scheme for detecting and classifying DoS, fuzzing, replay, and spoofing attacks. These are common CAN bus attacks that can threaten the safety of a vehicle’s driver, passengers, and pedestrians. For this purpose, we utilize supervised machine learning in combination with ensemble methods. We first perform data balancing and feature selection. We build and fine-tune random forest, Xtreme gradient boosting, and decision tree supervised learning models. We then combine these models with voting, stacking, and bagging ensemble methods. The results obtained demonstrate the effectiveness of the proposed scheme when trained on real-life CAN traffic datasets to detect and classify these four attacks. The stacking method achieved the highest performance in terms of accuracy, precision, recall, F1-score, and area-under-the-curve receiver operator characteristic (ROC-AUC). The stacking method outperformed other recently proposed methods with an F1-score, precision, recall, and accuracy of 0.993, 0.993, 0.993, and 0.986, respectively. Full article

Reinforcement learning has achieved significant success in sequential decision-making problems but exhibits poor adaptability in non-stationary environments with unknown dynamics, a challenge particularly pronounced in multi-agent scenarios. This study aims to enhance the adaptive capability of multi-agent systems in such volatile environments. We propose a novel cooperative Multi-Agent Reinforcement Learning (MARL) algorithm based on MADDPG, termed MACPH, which innovatively incorporates three mechanisms: a Composite Experience Replay Buffer (CERB) mechanism that balances recent and important historical experiences through a dual-buffer structure and mixed sampling; an Adaptive Parameter Space Noise (APSN) mechanism that perturbs actor network parameters and dynamically adjusts the perturbation intensity to achieve coherent and state-dependent exploration; and a Huber loss function mechanism to mitigate the impact of outliers in Temporal Difference errors and enhance training stability. The study was conducted in standard and non-stationary navigation and communication task scenarios. Ablation studies confirmed the positive contributions of each component and their synergistic effects. In non-stationary scenarios featuring abrupt environmental changes, experiments demonstrate that MACPH outperforms baseline algorithms such as DDPG, MADDPG, and MATD3 in terms of reward performance, adaptation speed, learning stability, and robustness. The proposed MACPH algorithm offers an effective solution for multi-agent reinforcement learning applications in complex non-stationary environments. Full article

With the explosion of vehicle-to-infrastructure (V2I) communications in the internet of vehicles (IoV), it is still very important to ensure secure authentication and efficient key agreement because of the vulnerabilities in the existing protocols such as physical capture attacks, privacy leakage, and low computational efficiency. This paper proposes a physical unclonable function (PUF)-based multi-factor authentication and key agreement protocol tailored for V2I environments, named as PMAKA-IoV. The protocol integrates hardware-based PUFs with biometric features, utilizing fuzzy extractors to mitigate biometric template risks, while employing dynamic pseudonyms and lightweight cryptographic operations to enhance anonymity and reduce overhead. Security analysis demonstrates its resilience against physical capture attacks, replay attacks, man-in-the-middle attacks, and desynchronization attacks, and it is verified by formal verification using the strand space model and the automated Scyther tool. Performance analysis demonstrates that, compared to other related schemes, the PMAKA-IoV protocol maintains lower communication and storage overhead. Full article

Tor is the most widely used anonymous communication network. Tor has developed a series of pluggable transports (PTs) to obfuscate traffic and avoid censorship. These PTs use different traffic obfuscation techniques, and many of them have been maintained and updated. In order to achieve continual learning against PTs and their updates, this paper proposes an incremental learning model for Tor traffic detection. First, we analyzed several common traffic obfuscation techniques, including randomization, mimicry, and tunneling. A feature set was designed for Tor obfuscation traffic detection. Second, this paper constructs the Tor incremental learning framework and proposes edge exemplar enhancement to enhance the memory of trained models for previous classes. It can enhance the previous class memory of the model through edge feature enhancement and selective replay to alleviate the catastrophic forgetting problem of incremental learning. Finally, we combined public and self-collected datasets to simulate the development of Tor PTs and verify the effectiveness of our model. The experimental results show that the improved model in this paper has the highest accuracy rate of 87.6% in the simulated environment. This means that the incremental learning model can effectively cope with the updating of PTs. Full article

In the field of unmanned aerial vehicle (UAV) path planning, the conventional deep Q-network (DQN) algorithm encounters the issue of action space discretization, which results in the generation of unsmooth and inefficient planned paths. To address this issue, we introduce the particle swarm optimization (PSO) algorithm into DQN to convert the discrete action space into a continuous one. This method divides the agent’s surrounding space into discrete and continuous action spaces. The PSO algorithm performs a global search in the continuous space to obtain a continuous candidate solution, while DQN learns a policy in the discrete space to obtain a discrete candidate solution. Then, the two candidate solutions are combined using a weighted vector method to determine a direction that balances global search and policy learning. Additionally, we introduce a novel feature matrix as the state space for DQN, providing more accurate environmental and positional representations. Furthermore, we incorporate a mechanism into the base prioritized experience replay (PER) and N-step updates, which combines the current temporal difference error (TD-error) with historical priorities and includes a policy entropy penalty term, thereby enhancing DQN’s ability to learn long-term dependencies. The performance of the PSO-DQN model is further improved through an enhanced $\epsilon$-greedy policy and learning rate decay strategy. Simulation results and experiments using the Flightmare simulator demonstrate that the proposed method generates smoother and more efficient paths for drones, exhibiting strong robustness in complex environments. Full article

Face anti-spoofing (FAS) has always been a hidden danger in network security, especially with the widespread application of facial recognition systems. However, some current FAS methods are not effective at detecting different forgery types and are prone to overfitting, which means they cannot effectively process unseen spoof types. Different loss functions significantly impact the classification effect based on the same feature extraction without considering the quality of the feature extraction. Therefore, it is necessary to find a loss function or a combination of different loss functions for spoofing detection tasks. This paper mainly aims to compare the effects of different loss functions or loss function combinations. We selected the Swin Transformer as the backbone of our training model to extract facial features to ensure the accuracy of the ablation experiment. For the application of loss functions, we adopted four classical loss functions: cross-entropy loss (CE loss), semi-hard triplet loss, L1 loss and focal loss. Finally, this work proposed combinations of Swin Transformers and different loss functions (pairs) to test through in-dataset experiments with some common FAS datasets (CelebA-Spoofing, CASIA-MFSD, Replay attack and OULU-NPU). We conclude that using a single loss function cannot produce the best results for the FAS task, and the best accuracy is obtained when applying triplet loss, cross-entropy loss and Smooth L1 loss as a loss combination. Full article

Background/Objectives: Motor neurorehabilitation can be realized by gradually learning diverse motor imagery (MI) tasks. EEG-based brain-computer interfaces (BCIs) provide an effective solution. Nevertheless, existing MI decoding methods cannot balance plasticity for unseen tasks and stability for old tasks. This paper proposes a generative diffusion-based task Incremental Learning (IL) method called GD-TIL. Methods: First, data augmentation is employed to increase data diversity by segmenting and recombining EEG signals. Second, to capture temporal-spatial features (TSFs) from different temporal resolutions, a multi-scale temporal-spatial feature extractor (MTSFE) is developed via integrating multiscale temporal-spatial convolutions, a dual-branch pooling operation, multiple multi-head self-attention mechanisms, and a dynamic convolutional encoder. The proposed self-supervised task generalization (SSTG) mechanism introduces a regularization constraint to guide MTSFE and unified classifier updating, which combines labels and semantic similarity between the augmentation with original views to enhance model generalizability for unseen tasks. In the IL phase, a prototype-guided generative replay module (PGGR) is used to generate old tasks’ TSFs by training a lightweight diffusion model based on the prototype and label of each task. Furthermore, the generated TSF is merged with a new TSF to fine-tune the convolutional encoder and update the classifier and PGGR. Finally, GD-TIL is evaluated on a self-collected ADL-MI dataset with two MI pairs and a public dataset with four MI tasks. Results: The continuous decoding accuracy reaches 80.20% and 81.32%, respectively. The experimental results exhibit the excellent plasticity and stability of GD-TIL, even beating the state-of-the-art IL methods. Conclusions: Our work illustrates the potential of MI-based BCI and generative AI for continuous neurorehabilitation. Full article

The increasing incorporation of Industrial Internet of Things (IIoT) devices into critical industrial operations and critical infrastructures necessitates robust security measures to safeguard confidential information and ensure dependable connectivity. Particularly in Cyber Physical Systems (CPSs), IIoT system security becomes critical as systems become more interconnected and digital. This paper introduces a novel Lightweight Industrial IoT Authentication (LI2A) method as a solution to address security concerns in the industrial sector and smart city infrastructure. Mutual authentication, authenticated message integrity, key agreement, soundness, forward secrecy, resistance to a variety of assaults, and minimal resource consumption are all features offered by LI2A. Critical to CPS operations, the approach prevents impersonation, man-in-the-middle, replay, eavesdropping, and modification assaults, according to a security study. The method proposed herein ensures the integrity of CPS networks by verifying communication reliability, identifying unauthorized message modifications, establishing a shared session key between users and IIoT devices, and periodically updating keys to ensure sustained security. A comprehensive assessment of performance takes into account each aspect of storage, communication, and computation. The communication and computing capabilities of LI2A, which are critical for the operation of CPS infrastructure, are demonstrated through comparisons with state-of-the-art systems from the literature. LI2A can be implemented in resource-constrained IIoT devices found in CPS and industrial environments, according to the results. By integrating IIoT devices into critical processes in CPS, it is possible to enhance security while also promoting urban digitalization and sustainability. Full article

The Internet of Vehicles (IoV), a key component of smart transportation systems, leverages 5G communication for low-latency data transmission, facilitating real-time interactions between vehicles, roadside units (RSUs), and sensor networks. However, the open nature of 5G communication channels exposes IoV systems to significant security threats, such as eavesdropping, replay attacks, and message tampering. To address these challenges, this paper proposes the Efficient Cluster-based Mutual Authentication and Key Update Protocol (ECAUP) designed to secure IoV systems within 5G-enabled sensor networks. The ECAUP meets the unique mobility and security demands of IoV by enabling fine-grained access control and dynamic key updates for RSUs through a factorial tree structure, ensuring both forward and backward secrecy. Additionally, physical unclonable functions (PUFs) are utilized to provide end-to-end authentication and physical layer security, further enhancing the system’s resilience against sophisticated cyber-attacks. The security of the ECAUP is formally verified using BAN Logic and ProVerif, and a comparative analysis demonstrates its superiority in terms of overhead efficiency (more than 50%) and security features over existing protocols. This work contributes to the development of secure, resilient, and efficient intelligent transportation systems, ensuring robust communication and protection in sensor-based IoV environments. Full article

Reinforcement Learning (RL) is increasingly being applied to complex decision-making tasks such as financial trading. However, designing effective reward functions remains a significant challenge. Traditional static reward functions often fail to adapt to dynamic environments, leading to inefficiencies in learning. This paper presents a novel approach, called Self-Rewarding Deep Reinforcement Learning (SRDRL), which integrates a self-rewarding network within the RL framework. The SRDRL mechanism operates in two primary phases: First, supervised learning techniques are used to learn from expert knowledge by employing advanced time-series feature extraction models, including TimesNet and WFTNet. This step refines the self-rewarding network parameters by comparing predicted rewards with expert-labeled rewards, which are based on metrics such as Min-Max, Sharpe Ratio, and Return. In the second phase, the model selects the higher value between the expert-labeled and predicted rewards as the RL reward, storing it in the replay buffer. This combination of expert knowledge and predicted rewards enhances the performance of trading strategies. The proposed implementation, called Self-Rewarding Double DQN (SRDDQN), demonstrates that the self-rewarding mechanism improves learning and optimizes trading decisions. Experiments conducted on datasets including DJI, IXIC, and SP500 show that SRDDQN achieves a cumulative return of 1124.23% on the IXIC dataset, significantly outperforming the next best method, Fire (DQN-HER), which achieved 51.87%. SRDDQN also enhances the stability and efficiency of trading strategies, providing notable improvements over traditional RL methods. The integration of a self-rewarding mechanism within RL addresses a critical limitation in reward function design and offers a scalable, adaptable solution for complex, dynamic trading environments. Full article

As one of the most crucial parts of face detection, the accuracy and the generalization of face anti-spoofing are particularly important. Therefore, it is necessary to propose a multi-branch network to improve the accuracy and generalization of the detection of unknown spoofing attacks. These branches consist of several frequency map encoders and one depth map encoder. These encoders are trained together. It leverages multiple frequency features and generates depth map features. High-frequency edge texture is beneficial for capturing moiré patterns, while low-frequency features are sensitive to color distortion. Depth maps are more discriminative than RGB images at the visual level and serve as useful auxiliary information. Supervised Multi-view Contrastive Learning enhances multi-view feature learning. Moreover, a two-stage feature fusion method effectively integrates multi-branch features. Experiments on four public datasets, namely CASIA-FASD, Replay–Attack, MSU-MFSD, and OULU-NPU, demonstrate model effectiveness. The average Half Total Error Rate (HTER) of our model is 4% (25% to 21%) lower than the Adversarial Domain Adaptation method in inter-set evaluations. Full article