Multi-modal action recognition via advanced image fusion techniques for cyber-physical systems

Recently, with the growth of cyber physical systems (CPS), several applications have begun to deploy in the CPS for connecting the cyber space with the physical scale effectively. Besides, the cloud computing (CC) enabled CPS offers huge processing and storage resources for CPS that finds helpful for a range of application areas. At the same time, with the massive development of applications that exist in the CPS environment, the energy utilization of the cloud enabled CPS has gained significant interest. For improving the energy effectiveness of the CC platform, virtualization technologies have been employed for resource management and the applications are executed via virtual machines (VMs). Since effective scheduling of resources acts as an important role in the design of cloud enabled CPS, this paper focuses on the design of chaotic sandpiper optimization based VM scheduling (CSPO-VMS) technique for energy efficient CPS. The CSPO-VMS technique is utilized for searching for the optimum VM migration solution and it helps to choose an effective scheduling strategy. The CSPO algorithm integrates the concepts of traditional SPO algorithm with the chaos theory, which substitutes the main parameter and combines it with the chaos. In order to improve the process of determining the global optimum solutions and convergence rate of the SPO algorithm, the chaotic concept is included in the SPO algorithm. The CSPO-VMS technique also derives a fitness function to choose optimal scheduling strategy in the CPS environment. In order to demonstrate the enhanced performance of the CSPO-VMS technique, a wide range of simulations were carried out and the results are examined under varying aspects. The simulation results ensured the improved performance of the CSPO-VMS technique over the recent methods interms of different measures.

The load model and its parameters have great effect on the accuracy of power system simulation. This paper proposed a method of real-timely collecting load data based on intelligent measuring equipment, charring out the bottom-up accurate modeling of power load under cyber physical system (CPS) environment. Using intelligent measuring equipment as online measurement and identification device, identified load model on different voltage levels based on online load data, and established the aggregated load model using the weighting-mean method based on the capacity. Finally, considering the distribution system parameter, established the integrated load model of 220/110kV bus by gradually converting. Proposed method can realize real-time dynamical update of power load model, and improve the accuracy of power system simulation. Simulation indicated that the load model obtained by proposed method can better reflect the load dynamic characteristics compared to the traditional overall identification method.

Effectively utilizing multimodal information (e.g., heart rate and acceleration) is a promising way to achieve wearable sensor based human activity recognition (HAR). In this paper, an activity recognition approach MARCEL (Multimodal Activity Recognition with Classifier Ensemble) is proposed, which exploits the diversity of base classifiers to construct a good ensemble for multimodal HAR, and the diversity measure is obtained from both labeled and unlabeled data. MARCEL uses neural network (NN) as base classifiers to construct the HAR model, and the diversity of classifier ensemble is embedded in the error function of the model. In each iteration, the error of the model is decomposed and back-propagated to base classifiers. To ensure the overall accuracy of the model, the weights of base classifiers are learnt in the classifier fusion process with sparse group lasso. Extensive experiments show that MARCEL is able to yield a competitive HAR performance, and has its superiority on exploiting multimodal signals.

Cyber Physical Systems (CPS) are redefining modern engineering practices by bringing the computational systems together with physical processes, boosting efficiency, security, and sustainability in different industries. This research highlights the influence of CPS in different sectors: its power to facilitate realtime decision making, predictive maintenance, and cybersecurity issues. The context in which this study occurs is to the application of digital twin technology in CPS — which is physical systems’ simulation for performance optimization and downtime reduction. The research includes the analysis of the efficiency of cybersecurity measures applied in CPS environment, the identification of vulnerabilities, and the suggestions of solutions for the mitigation of a cyber threat. Digital twins and predictive maintenance appears to improve the system performance by 30%, reduces the downtime by 25% due to application. Moreover, the research addresses upcoming concerns of sustainable practices in CPS and emerging technologies such as Industry 4.0 that promote resource optimization and waste reduction. The results show that CPS can improve both the operational efficiency and the environmental sustainability by enabling a smarter resource management. This work advances our understanding of trends, challenges and future directions in CPS as a means to inform engineering management practices, which are enabled by CPS.

Cyber physical system (CPS) is a network of cyber and physical elements, which interact with one another in a feedback form. CPS approves critical infrastructure and is treated as essential in day to day since it forms the basis of futuristic smart devices. An increased usage of CPSs poses security as a challenging issue and intrusion detection systems (IDS) can be applied for the identification of network intrusions. The latest advancements in the field of artificial intelligence (AI) and deep learning (DL) enables to design effective IDS models for the CPS environment. At the same time, metaheuristic algorithms can be employed as a feature selection approach in order to reduce the curse of dimensionality. With this motivation, this study develops a novel Poor and Rich Optimization with Deep Learning Model for Blockchain Enabled Intrusion Detection in CPS Environment, called PRO-DLBIDCPS technique. The proposed PRO-DLBIDCPS technique initially introduces an Adaptive Harmony Search Algorithm (AHSA) based feature selection technique for proper selection of feature subsets. For intrusion detection and classification, and attention based bi-directional gated recurrent neural network (ABi-GRNN) model is applied. In addition, the detection efficiency of the ABi-GRNN technique has been enhanced by the use of Poor and rich optimization (PRO) algorithm based hyperparameter optimizer, which resulted in enhanced intrusion detection results. Furthermore, blockchain technology is applied for enhancing security in the CPS environment. In order to demonstrate the enhanced outcomes of the PRO-DLBIDCPS technique, a wide range of simulations was carried out on benchmark dataset and the results reported the better outcomes of the PRO-DLBIDCPS technique in terms of several measures.

Action recognition, a field within human-centered computing, is vital for identifying and understanding human actions, benefiting applications like surveillance, autonomous vehicles, and human-computer interaction. Artificial intelligence’s enduring goal is to develop robust models for perceiving and understanding the visual world around us. Deep neural networks have shown exceptional performance in various tasks, profoundly impacting real-world applications, including action prediction and recognition, which have advanced significantly in recent years. Most work in visual human action recognition focuses on single viewpoints or modalities from complete observation. Yet, the real significance lies in predicting future actions from incomplete observations to prevent real-world tragedies. With the availability of multiple cameras and data from multiple modalities (RGB, Depth, and Skeleton) available today, it becomes possible to model human action in multi-view and multi-modality context, minimizing the data loss due to occlusions and signal quality issues to improve recognition accuracy on the strength of state-of-the-art deep learning models. However, deep neural network models are susceptible to adversarial attacks, where imperceptible perturbations can compromise action recognition model performance. This thesis focuses on identifying latent vulnerabilities and proposing a defense mechanism against such threats in a multi-modality and multi-view setting. This work introduces an efficient and effective attack mechanism that perturbs skeleton data by targeting key joints and segments while employing a graph attention mechanism that learns the semantics to perturb other modalities. Additionally, an approach has been developed that not only adds noise but also alters the visual spatial structure of skeleton data through generative modeling. Furthermore, this dissertation introduces a defense mechanism known as the Collaborative Knowledge Distillation Network, which leverages graph attention and knowledge distillation techniques. This network leverages the knowledge from compromised multi-view data and integrates information from clean data to address incomplete observations and noisy action videos, enhancing the robustness of action recognition models for real-world applications.

Cyber-attacks on cyber-physical systems (CPSs) resulted to sensing and actuation misbehavior, severe damage to physical object, and safety risk. Machine learning (ML) models have been presented to hinder cyberattacks on the CPS environment; however, the non-existence of labelled data from new attacks makes their detection quite interesting. Intrusion Detection System (IDS) is a commonly utilized to detect and classify the existence of intrusions in the CPS environment, which acts as an important part in secure CPS environment. Latest developments in deep learning (DL) and explainable artificial intelligence (XAI) stimulate new IDSs to manage cyberattacks with minimum complexity and high sophistication. In this aspect, this paper presents an XAI based IDS using feature selection with Dirichlet Variational Autoencoder (XAIIDS-FSDVAE) model for CPS. The proposed model encompasses the design of coyote optimization algorithm (COA) based feature selection (FS) model is derived to select an optimal subset of features. Next, an intelligent Dirichlet Variational Autoencoder (DVAE) technique is employed for the anomaly detection process in the CPS environment. Finally, the parameter optimization of the DVAE takes place using a manta ray foraging optimization (MRFO) model to tune the parameter of the DVAE. In order to determine the enhanced intrusion detection efficiency of the XAIIDS-FSDVAE technique, a wide range of simulations take place using the benchmark datasets. The experimental results reported the better performance of the XAIIDS-FSDVAE technique over the recent methods in terms of several evaluation parameters.

Healing Process is a major role in developing resiliency in cyber-physical system where the environment is diverse in nature. Cyber-physical system is modelled with Multi Agent Paradigm and biological inspired Danger Theory based-Artificial Immune Recognization2 Algorithm Methodology towards developing healing process. The Proposed methodology is implemented in a simulation environment and percentage of Convergence rates shown in achieving accuracy in the healing process to resiliency in cyber-physical system environment is shown.

Cyber-Physical Systems (CPS) are increasingly deployed in critical infrastructure yet remain vulnerable to adversarial attacks that manipulate sensor data to mislead AI-based decision-making. These threats demand not only high-accuracy detection but also transparency in model reasoning. This study proposes a Hybrid Explainable AI (XAI) Framework that integrates Convolutional Neural Networks (CNN), SHAP-based feature interpretation, and rule-based reasoning to detect adversarial inputs in CPS environments. The framework is tested on two simulation scenarios: industrial sensor networks and autonomous traffic sign recognition. Using datasets of 10,000 samples (50% adversarial via FGSM and PGD), the model achieved an accuracy of 97.25%, precision of 96.80%, recall of 95.90%, and F1-score of 96.35%. SHAP visualizations effectively distinguished between normal and adversarial inputs, and the added explainability module increased inference time by only 8.5% over the baseline CNN (from 18.5 ms to 20.1 ms), making it suitable for real-time CPS deployment. Compared to prior methods (e.g., CNN + Grad-CAM, Random Forest + LIME), the proposed hybrid framework demonstrates superior performance and interpretability. The novelty of this work lies in its tri-level integration of predictive accuracy, explainability, and rule-based logic within a single real-time detection system—an approach not previously applied in CPS adversarial defense. This research contributes toward trustworthy AI systems that are robust, explainable, and secure by design.

Cyber-physical Systems (CPS) have raised serious security concerns and thus have been subjected to intensive security research lately. Recent publications have shown that there is a potential to transfer hidden information through CPS environments. In comparison to these existing studies, we demonstrate that CPS cannot only be used to covertly transfer secret data but also to store secret data. Using an analogy to the biological concept of animal scatter hoarding behavior we exemplify CPS secret data storage using automated buildings.1

The security of Cyber-Physical Systems (CPS) has been recently receiving significant attention from the research community. To this end, this paper sheds the light on a number of security approaches for CPS from two perspectives, namely, control-theoretic and cyber security. Further, threat detectors in various CPS environments are highlighted and discussed. The aim is to demonstrate the lack of coherent approaches that systematically tackle both security aspects of such systems, in addition to pinpointing several insightful research gaps that endeavor to shape future CPS security solutions.

Recent research on modeling human daily activities in smart showed substantive challenges due to the heterogeneity of incoming data and the natural interleaving among human activities. This paper introduces an approach for handling multimodal activity recognition systems in smart homes by modeling atomic actions using multi-layer perceptron neural network; enriched with background knowledge from sensors' profiles. The proposed methodology defines the activity recognition problem as multiclass classification problem, in which an atomic action might belong to more than one activity at a given time. For each sensor, a profile of selected features has been created to identify the spatiotemporal aspect of its triggered actions. Then, a multimodal segmentation technique is applied to assemble actions that formulate a given activity into independent blocks of tuples. Such segmentation considers the interleaving among actions by applying multiclass classification. Experiments on multimodal dataset have been conducted to measure the performance of the proposed methodology. Measures such as F-measure and accuracy have been used to compare among different classifiers. The preliminary results showed a significant enhancement especially on activities that share high number of similar actions (performed concurrently).

The aim of this paper is to provide a conceptual formulation of a cyber-physical system (CPS) as a sociotechnical system (STS). CPS has become a major area for research and development; however, for successful integration with society, the sociotechnical dimension of CPS should be addressed. Examining basic issues will provide a strong foundation for CPS technologies along with providing stakeholder acceptance and engagement for wellbeing of society. While CPS can be considered as subsets of STS, they are not merely an application of STS principles. Rather, as this article demonstrates, considerations of fundamental principles involve a careful scrutiny of basic assumptions. This paper examines the basic assumptions involved in STS based on insights derived from a multitude of fields ranging from human–computer interaction, risk management, history of technology, philosophy of technology, science and technology studies and systems engineering. Further, based on sociotechnical models from three different fields, and in light of the basic assumptions of STS, a conceptual model of CPS is presented. This model is devised with a view towards human–technology interaction in CPS. It consists of technology and societal elements as hybrid entities available at several levels of functional abstraction. Using already existing example of CPS projects, the applicability of the proposed conceptual model is demonstrated. The paper concludes with a discussion of three main future challenges for human–technology interaction in CPS.

Chaotic marine predators optimization based task scheduling scheme for resource limited cyber-physical systems

Understanding semantic meaning of human actions captured in unconstrained environments has broad applications in fields ranging from patient monitoring, human-computer interaction, to surveillance systems. However, while great progresses have been achieved on automatic human action detection and recognition in videos that are captured in controlled/constrained environments, most existing approaches perform unsatisfactorily on videos with uncontrolled/unconstrained conditions (e.g., significant camera motion, background clutter, scaling, and light conditions). To address this issue, the authors propose a robust human action detection and recognition framework that works effectively on videos taken in controlled or uncontrolled environments. Specifically, the authors integrate the optical flow field and Harris3D corner detector to generate a new spatial-temporal information representation for each video sequence, from which the general Gaussian mixture model (GMM) is learned. All the mean vectors of the Gaussian components in the generated GMM model are concatenated to create the GMM supervector for video action recognition. They build a boosting classifier based on a set of sparse representation classifiers and hamming distance classifiers to improve the accuracy of action recognition. The experimental results on two broadly used public data sets, KTH and UCF YouTube Action, show that the proposed framework outperforms the other state-of-the-art approaches on both action detection and recognition.