Copy-move (CM) forgery is a common type of image manipulation that involves copying and pasting a region within an image to conceal or duplicate content. Detection of such forgeries acts as an important part of digital image forensics. Deep learning techniques, such as convolutional neural networks (CNNs), are employed to extract informative features from images. CNNs are known for their ability to capture complex patterns and structures, making them well-suited for image-related tasks like forgery detection. This paper introduces a reptile search algorithm with a deep transfer learning-based CM forgery detection (RSADTL-CMFD) approach. The presented model uses Neural Architectural Search Network (NASNet) for feature extraction in forgery detection which allows the network to effectively capture relevant and discriminative features from the input images. To enhance the performance of the NASNet model, we employ the reptile search algorithm (RSA) for hyperparameter tuning. This algorithm optimizes the network’s hyperparameters, enabling the model to quickly adapt to different forgery detection tasks and achieve superior performance. Finally, extreme gradient boosting (XGBoost) effectively utilizes the extracted features from the deep learning network to classify regions within the image as genuine or manipulated/forged. The experimental result analysis of the RSADTL-CMFD model is tested using benchmark datasets. An extensive comparative study highlighted the enhanced outcomes of the RSADTL-CMFD method over recent techniques.

Blockchain (BC) and Artificial intelligence (AI) based technologies have earned a better reputation amongst the research community, especially in the medical field. BC technology has emerged as a promising solution to revolutionize the medical field by addressing challenges related to efficiency, data security, and interoperability. A BC-aided smart healthcare system leverages the immutable and decentralized nature of BC to construct a secured and transparent ecosystem to manage processes and healthcare data. It leverages the secure and decentralized nature of BC to optimize the processes, security, interoperability, and efficiency of medical data. The existing system is exposed to security attacks on healthcare data. It can be necessary to construct a real-time detection device utilizing a cyber-physical system (CPS) with BC technology in a significant way. This article designs a novel Blockchain-Enabled Smart Healthcare System using Jellyfish Search Optimization with Dual-Pathway Deep Convolutional Neural Network (JSO-DPCNN) technique. The presented JSO-DPDCNN technique exploits the concept of BC-enabled secure data transmission and DL-based diagnosis model for moneypox disease on smart healthcare monitoring. To accomplish this, the JSO-DPCNN technique uses Ethereum-based public BC to secure the privacy of healthcare images. In addition, the JSO-DPCNN technique applies a feature extraction module using DPCNN, which extracts the suitable set of features in the input images. Moreover, the multiplicative long short-term memory (MLSTM) approach was used for the disease detection process. Lastly, the JSO system can be employed for the parameter tuning of the MLSTM model. The simulation result of the JSO-DPCNN system was executed on a benchmark medical dataset. The comprehensive outcomes highlighted the significant outcome of the JSO-DPCNN approach in terms of different measures.

Accurate and timely traffic flow prediction not just allows traffic controllers to evade traffic congestion and guarantee standard traffic functioning, it even assists travelers to take advantage of planning ahead of schedule and modifying travel routes promptly. Therefore, short-term traffic flow prediction utilizing artificial intelligence (AI) techniques has received significant attention in smart cities. This manuscript introduces an autonomous short-term traffic flow prediction using optimal hybrid deep belief network (AST2FP-OHDBN) model. The presented AST2FP-OHDBN model majorly focuses on high-precision traffic prediction in the process of making near future prediction of smart city environments. The presented AST2FP-OHDBN model initially normalizes the traffic data using min–max normalization. In addition, the HDBN model is employed for forecasting the traffic flow in the near future, and makes use of DBN with an adaptive learning step approach to enhance the convergence rate. To enhance the predictive accuracy of the DBN model, the pelican optimization algorithm (POA) is exploited as a hyperparameter optimizer, which in turn enhances the overall efficiency of the traffic flow prediction process. For assuring the enhanced predictive outcomes of the AST2FP-OHDBN algorithm, a wide-ranging experimental analysis can be executed. The experimental values reported the promising performance of the AST2FP-OHDBN method over recent state-of-the-art DL models with minimal average mean-square error of 17.19132 and root-mean-square error of 22.6634.

Intelligent transportation systems (ITS) are pivotal in modern urban development by enhancing mobility and transit efficiency. However, the challenges of accurate short-term traffic prediction persist due to real-time traffic data’s dynamic, nonlinear, and non-stationary nature. In this study, the author addresses these challenges and proposes a novel approach to improve traffic flow prediction. Our research introduces a hybrid model, combining Long-Short Term Memory (LSTM) and the Kalman filter-based Rauch-Tung-Striebel (RTS) noise reduction technique, tailored to mitigate the limitations of low market penetration of connected vehicles and data availability. This paper aims to provide a robust traffic prediction solution with direct applications in smart cities and real-time traffic management. To evaluate our model’s efficacy, The author conducted an empirical case study using the Enhanced Next Generation Simulation (NGSIM) dataset, which offers highly granular vehicle trajectory data. The author rigorously analyses our methodology and algorithms to assess the quality of traffic predictions. Our results show that the proposed model outperforms traditional LSTM models. In particular, the Bi-LSTM/RTS model achieved a remarkable accuracy of 99% in predicting traffic patterns during sunny weather conditions, signifying a significant advancement in short-term traffic prediction accuracy. The evaluation metrics demonstrate that Bi-LSTM outperforms LSTM by a wide margin, with a coefficient of determination value of 0.99 against 0.97. The trend anticipated by Bi-LSTM (8.6%) is more in line with the actual trend than that projected by LSTM (8.9%), as seen by the smaller MAPE. In conclusion, Bi-LSTM outperforms LSTM at predicting traffic patterns.

Automated facial emotion recognition (FER) is one of the important fields of human-computer interaction (HCI). FER remains challenging due to facial accessories, non-uniform illumination, pose variation, etc. Emotion detection exploiting conventional algorithms has the demerit of mutual optimization of classification and feature extraction. Artificial intelligence (AI) techniques can be employed to identify FER automatically. Deep learning (DL) driven FER models have recently allowed for designing an end-to-end learning process. Therefore, this study designs a Henry Gas Solubility Optimization with Deep Learning Based FER (HGSO-DLFER) technique for HCI. The HGSO-DLFER technique aims to recognize and identify various kinds of facial emotions. To accomplish this, the HGSO-DLFER technique employs adaptive fuzzy filtering (AFF) for noise removal. In addition, the MobileNet model is used for feature vector generation, and the HGSO algorithm optimally chooses its hyperparameter scan. For the recognition of facial emotions, the HGSO-DLFER technique uses an autoencoder (AE) classifier with a Nadam optimizer. A widespread experimental analysis is made to facilitate a better understanding of the FER results by the HGSO-DLFER technique. The comparative analysis showed the effective performance of the HGSO-DLFER technique over other FER techniques with maximum accuracy of 98.65%.

The Internet of Things (IoT) refers to a technology enabler to enhance the urban physical architecture and render public services. But, public access to accumulated heterogeneous IoT urban information is prone to hackers attacking connected devices to the internet intellectual property as well. IoT security serves a dynamic part in the smart city. Some IoT devices are connected in smart homes, and these connections were centred on gateways. In smart homes, the gateways gain a lot of significance; but their centralized structure causes many security vulnerabilities like availability, integrity, and certification. Unified “cloud-like” computing networks and Blockchain (BC) type systems should be used to sort out these problems. Therefore, this article develops a Blockchain-Assisted Secure Smart Home Network using Gradient Based Optimizer with Hybrid Deep Learning (BSSHN-GBOHDL) model. The presented BSSHN-GBOHDL technique employs BC technology to improve the confidentiality of the data in the smart home environment. In addition, the BSSHN-GBOHDL technique identifies malicious activities in the smart home environment via three sub-processes namely data preprocessing, hybrid deep learning (HDL)-based malicious activity classification, and GBO-based hyperparameter tuning. The GBO algorithm assists in the proficient hyperparameter selection of the HDL model, which aids in accomplishing increased detection efficiency. The experimental validation of the BSSHN-GBOHDL approach is tested on a benchmark NSL-KDD dataset with 65495 normal and 60743 attack samples. The results highlight the betterment of the BSSHN-GBOHDL approach over other recent methods with maximum accuracy of 98.29%.