ABSTRACT Cervical cancer remains one of the leading causes of cancer‐related mortality among women worldwide, emphasising the need for early and accurate diagnosis. Traditional deep learning approaches for cervical cancer detection often suffer from low accuracy, limited generalisation, and high computational complexity. To overcome these challenges, this study introduces CerviScan‐Net, a novel hybrid deep learning framework that integrates the Shifted Window Transformer (Swin Transformer), EfficientNet, and bidirectional long short‐term memory (BiLSTM)‐Attention modules for precise and efficient cervical cell classification. The Swin Transformer captures global contextual features, EfficientNet extracts fine‐grained local representations with optimised computational efficiency, and the BiLSTM‐Attention mechanism enhances spatial–semantic learning by focusing on diagnostically significant regions. Interpretability is achieved through a hybrid visualisation strategy combining enhanced gradient‐weighted class activation mapping (Grad‐CAM++; for convolutional layers) and attention rollout (for transformer layers), enabling transparent model predictions. The framework was trained and evaluated on the publicly available SipakMed cervical cell dataset, encompassing five cytological categories – dyskeratotic, koilocytotic, metaplastic, parabasal, and superficial–intermediate cells. Comparative analysis against benchmark models, including ResNet‐50 + LSTM, visual geometry group network (VGG)‐16 + LSTM, Xception (extreme inception network), and DenseNet, demonstrated that CerviScan‐Net achieved superior performance with 98.5% accuracy, 96.3% sensitivity, and 99.2% specificity. These results affirm CerviScan‐Net's robustness, clinical interpretability, and potential as a reliable AI‐assisted diagnostic tool for early detection and screening of cervical cancer.

The growth of resource-constrained embedded and mobile IoT devices has increased the need to find a security-related solution that meets the demand for good security as well as computational efficiency. While conventional schemes, AES and RSA, have good security properties, they are inefficient in devices that operate under these constraints due to their computational and memory limitations. This paper proposes a new framework for lightweight cryptography that combines chaotic key generation for high entropy with hybrid symmetric operations. The chaotic sensitivity and unpredictability of the logistic map allow our system to produce high-entropy keys in-memory using small seeds, and secure keys are provided through index permutation for confusion and diffusion. Furthermore, a hybrid system, with two models of XOR and modular arithmetic, is used to introduce nonlinear transformations with minimal additional overhead. A security analysis suggests the proposed system is resilient against brute-force or statistical attacks by the combined space complexity of a very large dynamic keyspace and the chaotic sensitivity of the mapping function. In practice, the experiment results support the analysis and align with NIST SP 800 − 22 randomness tests while achieving encryption within 412 ms for 10 MB of file data and outperforming AES-128 by 34% and reducing memory usage as well. Given the straightforward nature, adaptability, efficiency, and fit to IoT, sensor networks, and real-time mobile software applications, the proposed framework represents a means to reconcile IoT between theoretical security robustness, in order to deploy lightweight cryptography for use in pervasive computing environments.

ABSTRACT Heart disease continues to be a primary cause of mortality globally, highlighting the critical necessity for efficient early prediction and classification techniques. This study presents a new hybrid model attention‐based CNN‐Bi‐LSTM that integrates the SMOTE with an attention‐driven improved convolutional neural network‐recurrent neural network architecture to improve the classification of heart sounds, especially from imbalanced datasets. Heart sounds are difficult to classify because of their complex acoustic properties and the variability of their characteristics across frequency and temporal domains. The proposed model utilises an advanced CNN to effectively extract global and local features, in conjunction with a bidirectional long short‐term memory network to improve the architecture by capturing contextual information from both preceding and subsequent time sequences. The incorporation of spatial attention within the CNN and temporal attention in the RNN enables the model to concentrate on the most pertinent audio segments. To address the challenges presented by imbalanced and noisy datasets that may impede the efficacy of deep learning algorithms, our model employs SMOTE to improve data representation. The hybrid model outperformed popular models such as CNN, LSTM and CNN‐LSTM, achieving a classification accuracy of more than 97% on the PCG and PASCAL heart sound datasets. The findings demonstrate the model's reliability as an initial evaluation tool in clinical settings, thereby improving support for cardiovascular disease diagnosis.

ABSTRACT The Internet of Things (IoT) has revolutionised the maintenance of parking systems. By incorporating IoT technology, parking systems can assist drivers in quickly locating appropriate parking spaces, alleviating traffic congestion, and minimising emissions from vehicles meandering in search of parking. This paper proposes a machine learning multilayer model that efficiently handles data generated by the sensors; it works at two layers to accurately predict appropriate parking spaces according to the vehicle type. Based on the survey conducted by the researchers, we use different models at different layers. At the first layer, an ensemble technique predicts the available parking spot; in the second layer, a random forest technique detects the type of vehicle. The integration of these techniques makes the proposed multilayer model novel, time‐efficient, and cost‐effective. These techniques are selected because they can handle complex data patterns, use the model's different strengths, and achieve high accuracy in prediction and classification tasks. The proposed multilayer model is implemented on the sensor dataset extracted from the Harvard Dataverse. The model is implemented using Python in a Jupyter notebook, and the evaluation metrics include accuracy, recall, precision, F1 score, and ROC curve. The proposed multilayer model obtained an accuracy of 97.88%, a precision of 96.11%, a recall of 95.69%, an F1 score of 96.04%, and an AUC score of 0.89. The results prove that the proposed multilayer model achieves the highest accuracy among all existing models in predicting the parking space with appropriate vehicle detection for the most effective parking space allocation in real‐time scenarios.

Abstract Protein sequence classification needs to be performed quickly and accurately to progress bioinformatics advancements and the production of pharmaceutical products. Extensive comparisons between large databases of known proteins and unknown sequences are necessary in traditional protein classification methods, which can be time‐consuming. This labour‐intensive and slow manual matching and classification method depends on functional and biological commonalities. Protein classification is one of the many fields in which deep learning has recently revolutionized. The data on proteins are organized hierarchically and sequentially, and the most advanced algorithms, such as Deep Family‐based Method (DeepFam) and Protein Convolutional Neural Network (ProtCNN), have shown promising results in classifying proteins into relative groups. On the other hand, these methods frequently refuse to acknowledge this fact. We propose a novel hybrid model called ProteinCNN‐BLSTM to overcome these particular challenges. To produce more accurate protein sequence classification, it combines the techniques of amino acid embedding with bidirectional long short‐term memory (BLSTM) and convolutional neural networks (CNNs). The CNN component is the most effective at capturing local features, while the BLSTM component is the most capable of modeling long‐term dependencies across protein sequences. Through the process of amino acid embedding, sequences of proteins are transformed into numeric vectors, which significantly improves the precision of prediction and the representation of features. Using the standard protein samples PDB‐14189 and PDB‐2272, we analyzed the proposed ProteinCNN‐BLSTM model and the existing deep‐learning models. Compared to the existing models, such as CNN, LSTM, GCNs, CNN‐LSTM, RNNs, GCN‐RNN, DeepFam, and ProtCNN, the proposed model performed more accurately and better than the existing models.