International Journal of Information Technology最新文献

One of the key components of a smart city is thought to be the traffic control system. Road traffic congestion is prevalent in big cities due to increasing population density and rising transportation in cities. A smart traffic control system using cutting-edge computational intelligence algorithms has been developed to address numerous challenges related to traffic management on road networks and to assist regulators in making sound decisions. The current endeavor seeks to develop a new type of Smart Traffic Management System (SmartTMS) using state-of-the-art deep learning and optimization methods. The hybrid Gated Recurrent Deep Convoluted Network (GRCNet) approach is applied to accurately forecast the traffic congestion from the smart vehicular networks. In order to improve the classifier's decision-making ability and prediction accuracy, the parameters of the deep learning algorithm are tuned using a revolutionary Starling Murmuration Optimizer (StMO) methodology. Moreover, traffic congestion in vehicle networks can be precisely diagnosed and decreased with a low error rate and high accuracy by using the GRCNet-StMO model combination. The proposed SmartTMS's main benefits are its ease of deployment, quick congestion forecast time, and minimal computing complexity. To evaluate the effectiveness of the suggested model, a comprehensive performance and comparison study is carried out in this work, taking into account a number of factors like error rate, accuracy, miss rate, and journey duration.

The Visual Cryptography Scheme (VCS) allows secret images to be hidden in two or more shares. A solution to the problem of pixel expansion in conventional VCS was created: the Probability Scale Invariant VCS (P-SIVCS). However, a codebook needs to be written for P-SIVCS and VCS. Kafri’s Random Grid VCS (RG-VCS) did not require a codebook, however the revealed secret image was noisy. The random gird algorithms presented in this paper are both aesthetically pleasing and security relevant. This is also the first time that the first share of RG-VCS was chosen as a block instead of a uniform distribution, and it is also the first time that the concept of a multi-pixel was implemented in RG-VCS . In addition, compared to previous thresholding algorithms, the proposed algorithms have better performance in terms of contrast accuracy and visual quality according to theoretical analysis and experimental results.

Computer-aided diagnosis has emerged as one of the main areas of study for radiology diagnosis and medical imaging in recent years. Also, developing a single prediction methodology for handling multiple types of medical images is remains one of the most significant issues in recent times. For handling various kinds of medical images, this research presents Smart Multimodal Disease Detection (SMD₂), an innovative and powerful automated method. The proposed framework’s contribution is the ability to use various kinds of medical images to carry out an accurate and efficient disease diagnosis. The Woodpecker Mating Optimization Algorithm (WpMO) approach is used to optimally choose the most important features from the provided inputs, simplifying the classification process. In addition, the innovative Squeeze Excitation Capsule Network (SECNet) model is used to accurately identify and classify the disease class with a reduced computational time and complexity. A range of various medical imaging datasets, including X-ray, CT, and MRI, are considered for study in order to validate the performance outcomes of the proposed model. The results of the investigation indicate that the loss value of the proposed approach has dropped to 1.3, but its average accuracy has grown by 99%.

Electroencephalogram (EEG) data are susceptible to artifacts, such as lapses in concentration or poor imagination, which can significantly impact the accuracy of disease diagnosis in e-health applications. To mitigate this issue, the use of machine learning (ML) and potentially artificial intelligence (AI) solutions to accurately identify outliers becomes crucial. Unlike many AI methods that incorporate unnecessary or redundant input variables, our study focuses on detecting anomalous values in EEG data through the K nearest neighbor (KNN) process and Euclidean distance metric. Our proposed unsupervised non-parametric algorithm, known as the smart KNN outlier detector (SKOD), eliminates the need for initial parameter configurations such as the number of neighbors (K), while achieving high performance. Evaluation of SKOD using real EEG data from 140 trials demonstrated sensitivity and specificity exceeding 60%, with nearly perfect accuracy in detecting outliers reaching close to 100%.

Renewable energy sources (RES) reveal potential for the near future since they are sustainable and generate clean energy. Currently grid-connected solar photovoltaic (SPV) systems are becoming increasingly significant in meeting energy demand and contributing to clean energy production. The power electronic converter (PEC) plays a significant role in conversion, regulating and controlling the flow of power from RES. This research primarily focuses on high-gain cubic boost converter (HG-CBC) and simulated with the help of MATLAB/Simulink. The results are evaluated and contrasted with both traditional and other high-gain boost converters, focusing on the boost factor (B) and the total part count. It is essential to integrate SPV modules with MPPT system to capture the peak power available under varying temperature (T) and solar irradiation (G) levels. This research introduces a new hybrid based flying squirrel search optimization (FSSO) approach combined with Perturb & Observe (P&O) MPPT approach which allows for faster MPP, lesser oscillations at the output and high accuracy with less convergence time in contrast with P&O and FSSO MPPT. The proposed hybrid based HFSSO with P&O MPPT with HG-CBC exhibits low output voltage ripples of 0.039 % and a contrast lower rising time of 0.264 s and settled at 0.6 s.

Network security in the midst of evolving and complex cyber-attacks is a growing concern. As the complexity of network architectures grows, so does the need for advanced methods in network traffic analysis and classification. This study explores the application of a novel Graph Convolutional Networks (GCNs) to address the challenges associated with intelligent network traffic analysis. The network interactions are modeled as a graph, where nodes represent devices or IP addresses, and edges capture the communication channels between them. In this work, dataset which contains packet information of some network devices was obtained from an online repository. The data was preprocessed, normalized and label-encoded. Seven baseline models, including Feed Forward Network (FFN) were developed as reference to the proposed GCN. The parameters were tuned to optimize the performance and the dataset was split into average train-test to avoid overfitting. Two convolutional fully-connected layers were used also as more could cause oversmoothing. Performance of the novel GCN was compared with the reference models. The improved GCN model gave classification accuracy of 94.3% compared to classical GCN with 92.5% and FFN with 88%. Results also showed that the enhanced GCN proposed in this study outperformed the classical GCN and FFN in precision, recall, F1 score and area under curve metrics. Through the utilization of a GCN architecture and proposed enhancements, the proposed model demonstrates notable effectiveness in accurately classifying diverse types of network traffic. This research showed the efficacy of GCNs in intelligent network traffic analysis, offering a promising approach to augmenting cybersecurity efforts in an evolving digital landscape.

Due to their uniqueness and simplicity, handwritten signatures are used as a behavioral biometric feature to identify and authenticate individuals. Due to the increase in the criminal activity of forgers in forging signatures, organizations are forced to use computer systems to verify the authenticity of signatures. For this reason, offline signature verification system is widely used in most organizations. Despite the abundance of research conducted on signature verification, it is difficult to distinguish real and forged signature samples due to the lack of information in the signing process. On the other hand, the small number of training samples is a challenge for offline signature recognition systems. In recent years, to improve these problems, systems based on machine learning and deep learning methods have been presented. In this paper, we have proposed a graph neural network-based architecture for offline signature verification. In this work, the features in the signature images, which are the pixels that make up the signature, are extracted by the SIFT algorithm and sent to the graph-based neural network as a graph structure. After training the network, the data of the test samples are classified into one of two classes, genuine or forged. The proposed model was evaluated on two datasets, MCYT-75 and UTSig, and Accuracy (Acc), Average Error Rate (AER), False Acceptance Rate (FAR) and False Positive Rate (FPR) were considered as performance measures. In this model, the values of Acc, AER, FAR and FPR for the MCYT-75 data set are equal to 1,0, 0, and 0, respectively, and for the UTSig database, these values are equal to 0.092, 0.007, 0.014 and 0.

Cybercriminals, terrorists, political activists, whistleblowers, and others are drawn to the darknet market and its use for illicit purposes. Various methods are employed to identify the people who are behind these identities and websites. Since DNMs are more recent than other platforms, there are more unexplored research possibilities in this field. Research has been done to identify the buying and selling of products connected to hacking from Darknet Marketplaces, the promotion of cyber threats in hacker’s forums and DNMs, and the supply chain elements of content related to cyber threats. The proposed research covers one of the most promising research areas: darknet markets and social media platforms exploitation tools and strategies. The research uses 6 DNMs publicly available data and then identified the most popular social media platform and intent of discussion based on the interaction available in form of the user remarks and comments. The research caters the social media platform and cybercrimes or threats associated to them, by help of the machine learning algorithms Logistic Regression, RandomForestClassifier, GradientBoostingClassifier, KNeighborsClassifier, XGBClassifier, Voting Classifier and Deep Learning based model LSTM and Transformer based Model used. In existing research, natural language processing techniques were employed to identify the kinds of commodities exchanged in these markets, while machine learning approaches were utilized to classify product descriptions.In proposed research work advanced and lighter version of BERT and LSTM model used yielding accuracy of 90.12% and 91.35% respectively. LSTM performed best to extract multiclass classification of actual intension of social media usage by intelligent analysis on hackers’ discussions. Strategies on social media platforms such as Facebook, twitter, Instagram, Snapchat to exploit them using darknet platforms also explored. This paper contributes on cyber threat intelligence that leverages social media applications to work proactively to save their assets based on the threats identified in the Darknet.

In today’s technology landscape, systems must adapt to diverse conditions to be practically useful. Thermal imaging’s intersection with adverse weather presents a challenge for existing heavy networks designed for RGB images. This research addresses this gap by using knowledge distillation to optimise networks for thermal imaging in challenging weather. Current networks struggle with interpreting thermal images effectively in adverse conditions like fog or rain. Through knowledge distillation, our work aims to enhance these networks, ensuring compatibility and efficiency with thermal imaging. This effort holds promise for enhancing object detection in thermal images during adverse weather, benefiting surveillance systems, improving safety in self-driving vehicles under harsh conditions, and aiding search and rescue operations with limited visibility. This research doesn’t just refine networks; it empowers technology to excel in adverse conditions, promising practical applications that enhance safety, efficiency, and reliability across various technological domains.

Internet of Things (IoT) devices with limited energy and storage resources can access sensing services from wireless sensor networks (WSNs), which are collections of specialized transducers. Power consumption becomes one of the most important design considerations in WSN because battery replacement or recharge in sensor nodes is almost impossible. For the energy-constrained network, clustering algorithms are crucial for power conservation. By carefully balancing the network’s demand, a cluster head (CH) can lower energy usage and extend lifespan. The primary topic of this study is an effective CH election mechanism that alternates the CH position among nodes with higher energy levels than the others. In order to accomplish balanced load clustering in WSN, the method in this study takes into account a total of seven such factors and coordinates to select the optimal CHs from among them. To choose the set of CHs that can most effectively meet the coordination criterion from the available possibilities, modified MADM techniques are used. The improved version performs better, according to simulation analysis.