Intrusion Detection Systems (IDS) are essential for securing modern communication networks but face challenges in handling concept drift, the evolving nature of traffic and threats, while resisting adversarial attacks. Conventional IDSs often require computationally expensive full retraining, which limits their real-time deployment in resource-constrained communication environments. This paper presents a meta-learning–driven adaptive IDS framework for attack detection and prevention in dynamic networked systems. The proposed design leverages First-Order Model-Agnostic Meta-Learning (FOMAML) to enable rapid online adaptation using a small memory buffer, eliminating the need for complete retraining. To strengthen resilience, adversarial robustness is enhanced by integrating Fast Gradient Sign Method (FGSM)-based adversarial training directly into the meta-adaptation loop. Furthermore, an adaptive controller utilizes rolling-window accuracy monitoring to trigger updates only when significant performance degradation is observed, minimizing computational overhead. Experiments on the IDSIoT2024 dataset demonstrate that the proposed framework maintains robust detection performance, sustaining an F1-score above 0.93 across varying drift severities and exhibiting rapid recovery under simultaneous multi-class drift and adversarial attack scenarios. Its compact model size of 102.40KB and fast adaptation time of 41.41ms confirm its suitability for real-time, resource-constrained communication scenarios, highlighting the potential of meta-learning for attack detection and prevention in next-generation communication systems.

This is an erratum for the article "Adversarial Machine Learning in IoT Security: A Comprehensive Survey" published in ACM Comput. Surv. 58, 8, Article 202 (February 2026), 35 pages.

This survey presents a comprehensive analysis of Adversarial Machine Learning (AML) as a critical approach for enhancing the security of Internet of Things (IoT) ecosystems, with a focus on its integration into Intrusion Detection Systems (IDSs). It synthesizes existing research, identifies current limitations, and outlines future directions for effective AML deployment. The survey explores emerging technologies to improve IDS robustness, adaptability, and threat anticipation capabilities. It further addresses the challenges of implementing AML in resource-constrained IoT environments. A structured, multi-phase attack framework inspired by the MITRE ATT&CK model is proposed to illustrate the evolving lifecycle of adversarial threats. Ethical and regulatory considerations are examined, emphasizing responsible deployment and accountability. The cross-domain applicability of AML is discussed across various environments, including smart grids, autonomous systems, cloud, and edge computing. Real-world use cases are analyzed, with attention to adversarial datasets and evaluation practices. Additionally, the role of Human-in-the-Loop (HITL) systems is highlighted for enhancing interpretability and trust in AML-powered IDSs. This survey serves as a valuable resource for researchers and practitioners aiming at advancing AML in securing IoT ecosystems with innovative, ethical, and practical defense strategies.

The increasing connectivity of Internet of Things (IoT) devices and networks has significantly raised security concerns. Intrusion detection systems (IDSs) serve as a firstline defense mechanism to detect and identify various cyber threats. However, traditional IDSs frameworks come with their own challenges, such as high computational costs, limited generalization to evolving variants of cyberattacks, increasing complexity, and vulnerability to adversarial attacks. This paper proposes a meta-learning-based IDS framework using Model-Agnostic Meta-Learning (MAML) to particularly combat adversarial attacks in IoT networks. The designed architecture optimizes model initialization by performing inner-loop updates using adversarially perturbed data. It aggregates gradients from these adversarially adapted models in the outer loop to achieve a resilient initialization that generalizes well against adversarial attacks. The proposed approach is rigorously evaluated by training and testing the model on three major adversarial attacks: Fast Gradient Sign Method (FGSM), Projected Gradient Descent (PGD), and DeepFool. The experimental outcomes indicate the promising performance of the proposed architecture with an average attack detection accuracy of 95.97%. The compact model size of 0.06MB makes it suitable for deployment on resource-constrained IoT devices and networks. Furthermore, the lower inferencing time ensures the timely detection of intrusion, which is significant for real-time IDSs.

The Internet of Things (IoT) has become an integral part of modern societies, with devices, networks, and applications offering industrial, economic, and social benefits. However, these devices and networks generate vast amounts of data, making them a favourite target for cybercriminals. This article introduces an attention-based convolutional neural network (ACNN) for cyberattack detection in IoT. The proposed ACNN incorporates an Inception architecture with a spatial attention mechanism. This integration optimizes feature extraction and localization, bolstering the IDS's accuracy and efficiency. The efficacy of the designed IDS is investigated through several parameters using the CIC-IDS 2018 dataset. The experimental outcomes indicate that the proposed ACNN achieved a higher attack detection accuracy of 98.11% and successfully classified 14 classes. The model size was determined as 21.172 MB which can be easily deployed on resource-constrained IoT devices. The model exhibited a computational performance of 5.804 GFLOPS when executed on an NVIDIA T4 GPU, and its inference time was recorded as 0.1867 ms.

Internet of Vehicles (IoV) is an application of the Internet of Things (IoT) network that connects smart vehicles to the internet, and vehicles with each other. With the emergence of IoV technology, customers have placed great attention on smart vehicles. However, the rapid growth of IoV has also caused many security and privacy challenges that can lead to fatal accidents. To reduce smart vehicle accidents and detect malicious attacks in vehicular networks, several researchers have presented machine learning (ML)-based models for intrusion detection in IoT networks. However, a proficient and real-time faster algorithm is needed to detect malicious attacks in IoV. This article proposes a hybrid deep learning (DL) model for cyber attack detection in IoV. The proposed model is based on long short-term memory (LSTM) and gated recurrent unit (GRU). The performance of the proposed model is analyzed by using two datasets-a combined DDoS dataset that contains CIC DoS, CI-CIDS 2017, and CSE-CIC-IDS 2018, and a car-hacking dataset. The experimental results demonstrate that the proposed algorithm achieves higher attack detection accuracy of 99.5% and 99.9% for DDoS and car hacks, respectively. The other performance scores, precision, recall, and F1-score, also verify the superior performance of the proposed framework.

The Internet of Things (IoT) devices, networks, and applications have become an integral part of modern societies. Despite their social, economic, and industrial benefits, these devices and networks are frequently targeted by cybercriminals. Hence, IoT applications and networks demand lightweight, fast, and flexible security solutions to overcome these challenges. In this regard, artificial-intelligence-based solutions with Big Data analytics can produce promising results in the field of cybersecurity. This article proposes a lightweight dense random neural network (DnRaNN) for intrusion detection in the IoT. The proposed scheme is well suited for implementation in resource-constrained IoT networks due to its inherent improved generalization capabilities and distributed nature. The suggested model was evaluated by conducting extensive experiments on a new generation IoT security dataset ToN_IoT. All the experiments were conducted under different hyperparameters and the efficiency of the proposed DnRaNN was evaluated through multiple performance metrics. The findings of the proposed study provide recommendations and insights in binary class and multiclass scenarios. The proposed DnRaNN model attained attack detection accuracy of 99.14% and 99.05% for binary class and multiclass classifications, respectively.

Vehicular Sensor Networks (VSN) introduced a new paradigm for modern transportation systems by improving traffic management and comfort. However, the increasing adoption of smart sensing technologies with the Internet of Things (IoT) made VSN a high-value target for cybercriminals. In recent years, Machine Learning (ML) and Deep Learning (DL) techniques attracted the research community to develop security solutions for IoT networks. Traditional ML and DL approaches that operate with data stored on a centralized server raise major privacy problems for user data. On the other hand, the resource-constrained nature of a smart sensing network demands lightweight security solutions. To address these issues, this article proposes a Federated Learning (FL)-based attack detection framework for VSN. The proposed scheme utilizes a group of Gated Recurrent Units (GRU) with a Random Forest (RF)-based ensembler unit. The effectiveness of the suggested framework is investigated through multiple performance metrics. Experimental findings indicate that the proposed FL approach successfully detected the cyberattacks in VSN with the highest accuracy of 99.52%. The other performance scores, precision, recall, and F1 are attained as 99.77%, 99.54%, and 99.65%, respectively.

Plugin hybrid electric vehicles (PHEVs) are a suitable choice to achieve enhanced performance and reduced toxic gases. The considered PHEV consists of an integrated charging unit and a hybrid energy storage system (HESS). The proposed HESS comprises of a battery with high energy density and a supercapacitor with high power density coupled together to fulfill the load demands of the vehicle. A DC–DC buck converter with an uncontrolled rectifier has been used for an on-board and balanced charging and two bi-directional DC–DC buck-boost converters have been incorporated to ensure a smooth transition of energy. A rule-based algorithm has been employed as a supervisory controller by incorporating total power inflow and state of charge of the power sources to fulfill the load demands. Moreover, a robust integral backstepping-based nonlinear controller has been designed for the smooth execution and energy management of PHEV in terms of output voltage regulation, reference generation, and smooth tracking of current. Lyapunov stability theory has been used to ensure asymptotic stability of the PHEV. The performance of the proposed controller has been validated using simulations on MATLAB/Simulink by comparing it with the Lyapunov redesign and backstepping controllers already proposed in the literature. The robustness of the proposed controller has been verified by introducing uncertainty in the state model. Finally, the real-time applicability and effectiveness of the proposed work have been ensured using controller hardware in loop (C-HIL) test bench.

The Industrial Internet of Things (IIoT) refers to the use of traditional Internet of Things (IoT) concepts in industrial sectors and applications. IIoT has several applications in smart homes, smart cities, smart grids, connected cars, and supply chain management. However, these systems are being more frequently targeted by cybercriminals. Deep learning and big data analytics have great potential in designing and developing robust security mechanisms for IIoT networks. In this paper, a novel hybrid deep random neural network (HDRaNN) for cyberattack detection in the IIoT is presented. The HDRaNN combines a deep random neural network and a multilayer perceptron with dropout regularization. The proposed technique is evaluated using two IIoT security-related datasets: (i) DS2OS and (ii) UNSW-NB15. The performance of the proposed scheme is analyzed through a number of performance metrics such as accuracy, precision, recall, F1 score, log loss, Region of Convergence (ROC), and Area Under the Curve (AUC). The HDRaNN classified 16 different types of cyberattacks using with higher accuracy of 98% and 99% for DS2OS and UNSW-NB15, respectively. To measure the effectiveness of the proposed scheme, the performance metrics are also compared with several state-of-the-art attack detection algorithms. The findings of HDRaNN proved its superior performance over other DL-based schemes. The deployment perspective of the proposed work is also highlighted in this work.

The Industrial Internet of Things (IIoT) refers to the use of smart sensors, actuators, fast communication protocols, and efficient cybersecurity mechanisms to improve industrial processes and applications. In large industrial networks, smart devices generate large amounts of data, and thus IIoT frameworks require intelligent, robust techniques for big data analysis. Artificial intelligence (AI) and deep learning (DL) techniques produce promising results in IIoT networks due to their intelligent learning and processing capabilities. This survey article assesses the potential of DL in IIoT applications and presents a brief architecture of IIoT with key enabling technologies. Several well-known DL algorithms are then discussed along with their theoretical backgrounds and several software and hardware frameworks for DL implementations. Potential deployments of DL techniques in IIoT applications are briefly discussed. Finally, this survey highlights significant challenges and future directions for future research endeavors.

The blockchain has emerged as an innovative and powerful technology that shows the tremendous potential to enhance the smart industrial frameworks by providing encryption, immutable storage, and decentralization. In the past few years, several applications in the industrial Internet of Things (IIoT) have emerged and blockchain technologies gained great attention from academia and industry. To discover the great potential of blockchain technology for the IIoT, we present a comprehensive survey on security issues, blockchain architectures, and applications from the industrial perspective. This article starts with a comparison of exiting state‐of‐the‐art surveys of blockchain technologies for IoT/IIoT applications. A four‐layer reference architecture of IIoT is presented along with the functionalities and security issues of each layer. To address these challenges, we assess the potential of blockchain technology by considering the key characteristics, architectures, consensus algorithms, and implementation platforms. Furthermore, we also discussed some use cases of blockchain for the IIoT frameworks. Finally, this survey is concluded by highlighting some open issues and future research directions.

Plugin hybrid electric vehicles (PHEVs) are getting the attention of electric transportation market and the endusers. They comprise of smart charging mechanism and a hybrid energy storage system (HESS). In this study, a topology for HESS based on battery/ultracapacitor (UC), coupled with two bi-directional DC-DC buck-boost converters has been considered. For controlled integrated charging and smooth execution of energy management algorithm, a unidirectional DC-DC converter has been used. The mathematical model of the complete HESS with integrated charging unit for PHEV has been designed. A nonlinear controller termed as adaptive terminal sliding mode control (ATSMC) along with adaptive law has been proposed. The controller parameters have been tuned using genetic algorithm. Also an algorithm of high level control has been presented to switch between static and dynamic behaviors of the PHEV. The objective of the proposed control strategy is to adapt the unknown parameters of the system, deliver power for load well in time, output DC bus voltage regulation and smooth tracking of reference currents for the battery and UC with varying demands of the vehicle. The asymptotic stability of the system has been ensured by using Lyapunov stability theory. Finally, the energy management algorithm using the state of charge (SoC) as decisive factor is incorporated to maintain the stability of the system under varying load conditions. The performance of proposed controller has been compared with conventional sliding mode controller (SMC) and finite time synergetic controller (FTSC) using MATLAB/Simulink. The performance of the system is further verified by testing it on real-time hardware-in-loop setup.

The problem of extracting maximum power from a photovoltaic (PV) system with negligible power loss is concerned with the power generating capability of the PV array and nature of the output load. Changing weather conditions and nonlinear behavior of PV systems pose a challenge in tracking of varying maximum power point. A robust nonlinear controller is required to ensure maximum power point tracking (MPPT) by handling nonlinearities of a system and making it robust against changing environmental conditions. Sliding mode controller is robust against disturbances, model uncertainties and parametric variations. It depicts undesirable phenomenon like chattering, inherent in it causing power and heat losses. In this paper, a supertwisting sliding mode algorithm based nonlinear robust controller has been designed for MPPT of a PV system which not only removes the chattering but also enhances the overall system’s dynamic response. Moreover, supertwisting sliding mode controller is robust against changing environmental conditions like change in temperature and irradiance. Noninverting DC-DC Buck-Boost converter has been used as an interface between source and the load. The efficiency of MPPT of a PV system depends upon the accuracy of reference for peak power voltage, therefore an efficient mechanism for reference generation has also been proposed in this work. The reference for peak power voltage has been generated by using a trained artificial neural network, which is to be tracked by proposed nonlinear controllers. Sliding mode controller (SMC) and synergetic controllers have also been designed for MPPT of a PV system in order to compare them with supertwisting sliding mode controller (ST-SMC). Global asymptotic stability of the system has been ensured by using Lyapunov stability criterion. The performance of the proposed nonlinear controllers has been validated in MATLAB/Simulink ODE 45 environment. ST-SMC has also been compared with recently proposed integral backstepping controller and other conventional MPPT controllers given in the literature. The simulation results show the better performance of ST-SMC in terms of best dynamic response and robustness.

This paper presents a prototype design of a robot for household purpose which aims to act as an aide for humans especially bed-ridden or handicapped people. It can perform various tasks without the need of human assistance. These tasks include lifting and moving objects that are performed with the help of robotic arm and chassis. It can also be configured as surveillance robot within home premises. The robotic chassis has been fitted with 2 gear motors that can displace weight up to 15 kg while the robotic arm is controlled using six servo motors of different torque capacities. These servo motors can be moved at any angle as per requirement. The robotic arm is a six Degrees of Freedom (DOF) arm which has 6 different motions, one associated with every servo motor. DOF can be defined as the freedom of movement of a rigid body in a 3-dimensional space. These movements include surging, swaying, heaving, rolling, pitching and yawing. Its Main Control Unit (MCU) is based upon Raspberry Pi and Arduino. The robot has been wirelessly controlled with the help of RF Transceivers NRF24L01, which can provide user the control of robot up to range of 100m. Joysticks have ben interfaced with an Arduino which is used to transmit instructions to the MCU, via RF transceivers, which correspondingly operate the robot as desired. Video transmission capability has been added to the robot enabling it to operate beyond visual range and provide surveillance of premises for the user. This transmission can be viewed within the local network on the web browser of any device.