Naveed Islam

The Internet of things (IoT) is an emerging technology for many smart applications due to its efficient resource utilization, scalability, and fast interaction with the physical world. Software‐defined network (SDN), on the other hand, provides dynamic services for controlling and managing real‐time systems. However, collected data are sent to a central location, which requires balancing energy resources with redundant channels to maximize the availability of smart functions. Furthermore, the IoT network faces numerous security vulnerabilities as a result of its open communication space, including malicious messages and privacy concerns. Thus, this paper presents a distributed and artificial intelligence‐based energy‐efficient model for IoT‐SDN architecture, which aims to improve data aggregation and power distribution. It also provides security and authentication for smart communication systems. First, the proposed model introduces the heuristic evaluation using artificial intelligence and decreases the power consumption for sensor nodes in a real‐time system. Moreover, it optimizes the paradigm of distributed processing and efficiently increases the green energy technology with nominal management costs using the mobile edges. Second, the aggregated data of the environment is secured using a centralized controller to attain the most trustworthy data availability. The experimental results show a comparative analysis of the proposed model in terms of energy efficiency, packet drop ratio, and waiting time by 22%, 23%, 40%, and 49% as compared to existing studies.

Accurate diagnosis of brain tumors is critical in understanding the prognosis in terms of the type, growth rate, location, removal strategy, and overall well-being of the patients. Among different modalities used for the detection and classification of brain tumors, a computed tomography (CT) scan is often performed as an early-stage procedure for minor symptoms like headaches. Automated procedures based on artificial intelligence (AI) and machine learning (ML) methods are used to detect and classify brain tumors in Computed Tomography (CT) scan images. However, the key challenges in achieving the desired outcome are associated with the model's complexity and generalization. To address these issues, we propose a hybrid model that extracts features from CT images using classical machine learning. Additionally, although MRI is a common modality for brain tumor diagnosis, its high cost and longer acquisition time make CT scans a more practical choice for early-stage screening and widespread clinical use. The proposed framework has different stages, including image acquisition, pre-processing, feature extraction, feature selection, and classification. The hybrid architecture combines features from ResNet50, AlexNet, LBP, HOG, and median intensity, classified using a multilayer perceptron. The selection of the relevant features in our proposed hybrid model was extracted using the SelectKBest algorithm. Thus, it optimizes the proposed model performance. In addition, the proposed model incorporates data augmentation to handle the imbalanced datasets. We employed a scoring function to extract the features. The Classification is ensured using a multilayer perceptron neural network (MLP). Unlike most existing hybrid approaches, which primarily target MRI-based brain tumor classification, our method is specifically designed for CT scan images, addressing their unique noise patterns and lower soft-tissue contrast. To the best of our knowledge, this is the first work to integrate LBP, HOG, median intensity, and deep features from both ResNet50 and AlexNet in a structured fusion pipeline for CT brain tumor classification. The proposed hybrid model is tested on data from numerous sources and achieved an accuracy of 94.82%, precision of 94.52%, specificity of 98.35%, and sensitivity of 94.76% compared to state-of-the-art models. While MRI-based models often report higher accuracies, the proposed model achieves 94.82% on CT scans, within 3-4% of leading MRI-based approaches, demonstrating strong generalization despite the modality difference. The proposed hybrid model, combining hand-crafted and deep learning features, effectively improves brain tumor detection and classification accuracy in CT scans. It has the potential for clinical application, aiding in early and accurate diagnosis. Unlike MRI, which is often time-intensive and costly, CT scans are more accessible and faster to acquire, making them suitable for early-stage screening and emergency diagnostics. This reinforces the practical and clinical value of the proposed model in real-world healthcare settings.

Software-defined networks (SDN) have gained a lot of attention in recent years as a technique to develop smart systems with a help of the Internet of Things (IoT). Its powerful and centralized architecture makes a balanced contribution to the management of sustainable applications through efficient processes. These networks also systematically keep track of mobile devices and decrease the extra overheads in the communication cost. Many solutions are proposed to cope with data transferring for the critical system, however, mobile devices, on the other hand, require long-distance communication links with minimal retransmissions. Furthermore, the mobile network is highly infected by security attacks and compromised the IoT architecture for both the intermediate layers and end-users. Therefore, this paper presents an adaptive routes migration model for sustainable applications with the collaboration of SDN architecture and limits the disconnectivity time in data transporting along with efficient management of network services. Moreover, its centralized controller fetches the updated information from low-level smart devices and supervised their monitoring efficiently. The proposed model also secures the cloud of things (CoTs) from network threats and protects private data. It provides three levels of security algorithms and supports adaptive computing systems. The proposed model was tested using simulations, and the findings showed that it outperformed other existing studies in terms of packet delivery ratio by 13%, packet loss rate by 15%, transmission error by 22%, computing cost by 17%, and latency by 18%.

Pashto is the native language of Afghanistan and one of Pakistan's most essential and regional languages. The Pashto language has a vast number of native speakers who live in various parts of the world. The handwritten Pashto textual trajectories are hard to recognize and detect due to the cursive style and handwriting variation. The transformation behaviour, i.e., scaling, rotation, and shifting of handwritten text, are the prominent but challenging factors. A lightweight deep learning-based model construction for low and medium-resource devices in a less-constrained environment is challenging. This paper provides a practical, light deep learning-based model for predicting handwritten Pashto words. A massive Pashto-transformed invariant inverted handwritten text dataset is prepared with the help of the Pashtun community. A lightweight MobileNetV2 has been highly tuned for Pashto handwritten text classification, extracting images' features (MoI). We inverted the dataset to make the model more accurate and restrict it to fifteen epochs. Extensive experiments have been conducted to validate the suggested model's performance. The proposed transformed invariant lightweight Pashto deep learning (TILPDeep) technique achieves a training accuracy of 0.9839 and a validation accuracy of 0.9405 for transformed invariant Pashto handwritten inverted text using recognition matrices.

This paper provides an analysis of the combining effect of novel activation function and loss function based on M-estimation in application to extreme learning machine (ELM), a feed-forward neural network. Due to the computational efficiency and classification/prediction accuracy of ELM and its variants, they have been widely exploited in the development of new technologies and applications. However, in real applications, the performance of classical ELMs deteriorates in the presence of outliers, thus, negatively impacting the precision and accuracy of the system. To further enhance the performance of ELM and its variants, we proposed novel activation functions based on the psi function of M and redescend the M-estimation method along with the smooth l2-norm weight-loss functions to reduce the negative impact of the outliers. The proposed psi functions of several M and redescending M-estimation methods are more flexible to make more distinct features space. For the first time, the idea of the psi function as an activation function in the neural network is introduced in the literature to ensure accurate prediction. In addition, new robust l2 norm-loss functions based on M and redescending M-estimation are proposed to deal with outliers efficiently in ELM. To evaluate the performance of the proposed methodology against other state-of-the-art techniques, experiments have been performed in diverse environments, which show promising improvements in application to regression and classification problems.

The use of handwritten recognition tools has increased yearly in various commercialized fields. Due to this, handwritten classification, recognition, and detection have become an exciting research subject for many scholars. Different techniques have been provided to improve character recognition accuracy while reducing time for languages like English, Arabic, Chinese and European languages. The local or regional languages need to consider for research to increase the scope of handwritten recognition tools to the global level. This paper presents a machine learning-based technique that provides an accurate, robust, and fast solution for handwritten Pashto text classification and recognition. Pashto belongs to cursive script division, which has numerous challenges to classify and recognize. The first challenge during this research is de-veloping efficient and full-fledged datasets. The efficient recognition or prediction of Pashto handwritten text is impossible by using ordinary feature extraction due to natural transformations and handwriting variations. We propose some useful invariant features extracting techniques for handwritten Pashto text, i.e., radial, orthographic grid, perspective projection grid, retina, the slope of word trajectories, and cosine angles of tangent lines. During the dataset creation, salt and pepper noise was generated, which was removed using the statistical filter. Another challenge to face was the invalid disconnected handwritten stroke trajectory of words. We also proposed a technique to minimize the problem of disconnection of word trajectory. The proposed approach uses a linear support vector machine (SVM) and RBF-based SVM for classification and recognition.

In recent years, 5G and the Internet of Things (IoT) have been integrated into a variety of applications to support sustainable communication systems. In the presence of intermediate hardware, IoT devices collect the network data and transfer them to cloud technologies. The interconnect machines provide essential information to the connected devices over the Internet. Many solutions have been proposed to address the dynamic and unexpected characteristics of IoT-based networks and to support smart developments. However, more work needs to explore efficient quality-aware data routing for distributed processing. Additionally, to handle the massive amount of data created by smart cities and achieve the transportation objectives for resource restrictions, artificial intelligence (AI)-oriented approaches are necessary. This research proposes a secured protocol with collaborative learning for IoT-enabled sustainable communication using AI techniques. This approach increases systems' reaction times in critical conditions and also controls the smart functionalities for inter-device communication. Furthermore, fitness computing can help in balancing the contribution of quality-aware metrics to achieve load balancing and efficient energy consumption. To deal with security, IoT communication is broken down into stages, resulting in a more dependable network for unpredictable environments. The simulation results of the proposed protocol have been compared to existing approaches and improved the performance of response time by 17%, energy consumption by 14%, number of re-transmissions by 16%, and computing overhead by 16%, under a varying number of nodes and data packets.

Explainable Artificial Intelligence (XAI) is one of the most advanced research areas of Artificial Intelligence (AI). To explain the deep learning (DL) model is the main objective of XAI. It deals with artificial models which are understandable to humans, including the users, developers, policymakers, etc. XAI is very important in some critical domains like security, healthcare, etc. The purpose of XAI is only to provide a clear answer to the question of how the model made its decision. The explanation is very important before any system decision-making. As an example, if a system responds to a decision, it is necessary to have inside knowledge of the model about that decision. The decision can be positive or negative, but it is more important to know the decision based on characteristics. The decision of the model should be trusted when we know the internal structure of the DL model. Generally, DL models come under the black box models. So for security purposes, it is very necessary to explain a system internally for any decision-making. Security is very crucial in healthcare as well as in any other domain. The objective of this research is to provide a decision about security based on XAI which is a big challenge. We can improve security systems based on XAI for the next level. For medical/healthcare security, when we recognize human action using transfer learning techniques, one pre-trained model is considered good for action and the same action is not good in terms of accuracy using another pre-trained model. This is called the black-box model problem, and it needs to know what is the internal mechanism of both models for the same action. Why one model considers good for action and why the same action is not very well using another model? Here need a model-specific approach of post-hoc interpretability to know the internal structure and characteristics of both models for the same action.

In modern years, the Internet of Things (IoT) has gained tremendous growth and development in various sectors because of its scalability, self-configuring, and heterogeneous factors. It performs a vital role in improving multimedia communication and reducing production costs. The multimedia data consist of various types and formats (text, audio, videos, etc.), which are forwarded in the form of blocks of bits in the network layer of TCP/IP. Due to limited resources available to IoT-built devices, most of the Multimedia Internet of Things (MIoT)-based applications are delay constraints, especially for big data content. Similarly, multimedia-based applications are more vulnerable to security burdens and lower the trust of data processing. In this paper, we present a secure and sustainable prediction framework for MIoT data transmission using machine learning, which aims to offer intelligent behavior of the system with information protection. Firstly, the network edges exploit a regression analysis for a real-time multimedia routing scheme and achieve precise delivery towards the media servers. Secondly, an efficient and low-processing asymmetric process is proposed to provide secure data transmission between the IoT devices, edges, and data servers. Extensive experiments are performed over the OMNET++ network simulator, and its significance is achieved by an average for energy consumption by 71%, throughput by 30.5%, latency by 22%, bandwidth by 34.5%, packets overheads by 38.5%, computation time by 12.5%, and packet drop ratio by 35% in the comparison of existing schemes.

Breast cancer is one of the common disease in female gender population all over the world. The classical methods of segmentation and classification for malignant cells are not only repetitive but also very time-consuming. Therefore, a computer-aided diagnosis is needed for automatic segmentation and classification of malignant cells in breast cytology images. In this article, a machine learning-based approach is proposed for malignant cell segmentation and classification in breast cytology images. In the proposed approach, the segmentation of cells is performed by a level set algorithm which is used to extract statistical information related to the malignant and benign cells. Similarly, the gray level co-occurrence matrix is computed to exploit the texture information, and support vector machine-based classification is used for the classification of malignant and benign cells. It has been observed through experiments that the proposed approach achieved high accuracy (96.3%) in the classification of malignant and benign cells.