Accurate estimation of battery state including state of charge (SoC) and state of health (SoH) are crucial for ensuring safety in energy storage applications. The SOC and SOH estimators were independently trained using the same input vector but with different objective functions, no integration between SOC and SOH estimations were explored. In this paper, a unified algorithm, for identifying both SoC and SoH states, is introduced by considering the Bayesian optimization for hyperparameter tuning. This approach allows seamless transition between SoC and SoH estimation without needing separate models for each task. In addition, equipping the dual estimation framework with a unified algorithm for identifying both states would impact the algorithm’s complexity. The suggested BiLSTM model reduces complexity in real-time Battery Management System (BMS) applications by eliminating the need for a separate model to estimate SoH. When compared to other machine learning and deep learning models such as Support Vector Machines (SVM), Decision Trees (DT), Random Forest (RF), Radial Basis Function Neural Networks (RBF-NN), Recurrent Neural Networks (RNN), and LSTM, the suggested BiLSTM method demonstrates the highest efficiency. Finally, to verify the proposed method’s effectiveness, a comparison among the different evaluation metrics was conducted. The proposed BiLSTM model achieved an average MAE (Mean Absolute Error) of 0.08 and NRMSE (Normalized Root Mean Squared Error) of 0.15 for SoC estimation across various temperatures (5?C,15?C, 35?C, and 45?C), and an MAE of 3.12 and NRMSE of 0.23 for SoH estimation with a degradation rate of 47% of the cell estimated from the predicted capacity values

Facial recognition systems are widely used in various applications such as security, healthcare, and authentication, but face significant challenges in uncontrolled environments. Poor lighting conditions can obscure facial features, introduce shadows, and distort spatial relationships, while changes in pose are critical for accurate identification. Existing methods often struggle to strike a balance between accuracy, computational efficiency, and robustness. Deep learning has become popular for automatically learning features through convolution layers. This study proposes a robust framework that integrates contrast-limited adaptive histogram equalization (CLAHE) and adaptive gamma correction for illumination normalization and multi-task cascaded convolutional networks (MTCNN) for precise face detection under varying poses and lighting conditions. This study proposes a deep learning-based approach for face recognition utilising multiple models, including VGG16, VGG19, ResNet-50, ResNet-101, and MobileNetV2. For classification, an ensemble of SVM, XGBoost, and random forest classifiers is combined using weighted averaging. The approach is tested on datasets such as CASIA3D and 105PinsFace, which include variations in illumination conditions. Using deep learning for automated hierarchical feature extraction and ensemble strategies, experimental results demonstrate significant improvements in recognition accuracy and enhanced robustness against lighting and pose variations while ensuring scalability for real-world applications. The approach achieved 99.91% accuracy on the CASIA3D dataset and 98.77% on the 105PinsFace dataset, showcasing its effectiveness across challenging conditions.

The integration of PSO with CNN provides a promising approach for classifying ASD using sMRI data. ASD is a behavioral disorder that impacts a person’s lifetime tendency to reciprocate with society. The variability and intensity of ASD symptoms, in addition to the fact that they share symptoms with other mental disorders, make an early diagnosis difficult. The key limitation of CNN is selecting the best parameters. To overcome this, we use PSO as an optimization approach within the CNN to choose the most relevant parameters to train the network. In the proposed approach, we initialize a swarm of particles, where each particle represents a unique configuration of CNN hyperparameters, including the number of convolutional layers, learning rates, filter sizes, and batch sizes. To evaluate the swarm in PSO, we use a fitness function, such as accuracy, to measure each particle’s performance. The performance of the proposed approach for ASD prediction outperformed that of the other optimizers with a high convergence rate.

In the field of vehicular communication, the Internet of Vehicles (IoV) serves as a new era that guarantees increased connectivity, efficiency, and safety. The modern area and new technology have their challenges and constraints, though. This paper thoroughly examines these constraints significantly; we show how blockchain technology is being used to overcome them. This paper primarily explores the complexities of Blockchain-enabled Internet of Vehicles (BIoV) architectures, the applications they serve, and the robust security features they provide through a systematic literature review (SLR). In addition, we look at the several ways that blockchain and IoV might be integrated and investigate the subtle factors that should be considered when choosing consensus algorithms to maximize performance on different blockchains. This paper also addresses the methods and tools used to identify and avoid fraudulent activities in BIoV networks at a maximum level of security. It also reveals the wide range of BIoV applications and analyzes the different security levels they provide. In closing, we give an idea of the possibilities that will continue to develop the blockchain and IoV environment, reducing the roadblocks and advancing this combination toward a more secure, effective, and connected future for vehicle communication systems

The requirement of a person to be available during the authentication made the biometric authentication to be used in abundant applications over password or token-based authentication systems. Privacy and security are two major concerns still to be addressed in biometric authentication system. In the last couple of years, researchers used the homomorphic encryption (HE) to propose the privacy-preserving biometric authentication systems which overcomes the limitations of cancelable biometrics and biometric cryptosystems. But these methods fail to achieve overall performance and security measures. To handle this, we introduce a privacy-preserving Bimodal authentication system (PPBA) utilizing Fan-Vercauteren scheme. An optimized method is proposed to compute the hamming distance between the encrypted templates that helps to carry out the computation without disclosing the user sensitive data. PPBA is tested on publicly available databases to analyze its efficiency. PPBA satisfies the diversity, irreversibility, revocability properties and also achieves decent performance.

The use of a biometric authentication system (BAS) for reliable automatic human recognition has increased exponentially over traditional authentication systems in recent years. Multimodal BAS was introduced to solve unimodal BAS's difficulties and improve security. Privacy and security are two significant concerns to be addressed in BAS, as biometric traits are irrevocable. Researchers employed cancelable biometrics in the past few years to propose several privacy-preserving BAS. We propose a privacy-preserving bimodal cancelable BAS (PPBCBAS) to overcome these problems. The traits used in our method are iris and fingerprint. Features are extracted from both the traits, and feature level fusion is done by concatenating the feature vectors of iris and fingerprint. PPBCBAS uses a quotient filter to generate the cancelable template, and the comparison is made on these transformed templates using the modified Hamming distance. PPBCBAS has been tested on three publicly available databases to analyze its efficiency. PPBCBAS satisfies the diversity, irreversibility, and revocability properties and achieves decent performance.

Human beings can describe scenarios and objects in a picture through vision easily whereas performing the same task with a computer is a complicated one. Generating captions for the objects of an image helps everyone to understand the scenario of the image in a better way. Instinctively describing the content of an image requires the apprehension of computer vision as well as natural language processing. This task has gained huge popularity in the field of technology and there is a lot of research work being carried out. Recent works have been successful in identifying objects in the image but are facing many challenges in generating captions to the given image accurately by understanding the scenario. To address this challenge, we propose a model to generate the caption for an image. Residual Neural Network (ResNet) is used to extract the features from an image. These features are converted into a vector of size 2048. The caption generation for the image is obtained with Long Short-Term Memory (LSTM). The proposed model is experimented on the Flickr8K dataset and obtained an accuracy of 88.4%. The experimental results indicate that our model produces appropriate captions compared to the state of art models.

Everyday dealing with enormous amounts of sensitive data requires its protection and communication over the insecure network. The field of ‘Steganography’ always attracted researchers for significant amount of scientific attention to protect and communicate sensitive data. This paper presents a secure steganography scheme for hiding Gray-scale secret image into a Color cover image by replacing cover image bits in frequency domain using modified quantum substitution box (S-Box). The inclusion of modified quantum S-Box for concealing secret bits in randomly selected any of the two channels of cover image ensures enhanced security. In the proposed scheme, we first performed discrete cosine transform (DCT) on the cover image. Then, quantum S-box is applied to locate the position of DCT coefficients where least significant bits are substituted intelligently based on the relative ordering of DCT frequencies. This relative ordering is achieved by traversing DCT coefficients in a zigzag manner where less important pixels have been altered more effectively without any major loss in image quality. The security of proposed method is examined by key space, key sensitivity parameters and robustness analysis. Additionally, the conducted simulation results demonstrate that our proposed steganography scheme has better visual image quality in terms of MSE, PSNR, UQI, SSIM, RMSE parameters as compared to other state-of-the-art works.

Homomorphic encryption (HE) is the most widely explored research area in the construction of privacy-preserving biometric authentication systems because of its advantages over cancellable biometrics and biometric cryptosystems. However, most of the existing privacy-preserving biometric authentication systems using HE assume that the server performs computations honestly. In a malicious server setting, the server may return an arbitrary result to save computational resources, resulting in a false accept/reject. To address this, secure and verifiable multi-instance iris authentication using blockchain (SviaB) is proposed. Paillier HE provides confidentiality for the iris templates in SviaB. The blockchain offers the integrity of the encrypted reference iris templates as well as the trust of the comparator result. The challenges of using blockchain in biometrics are also addressed in SviaB. Extensive experimental results on benchmark iris databases demonstrate that SviaB provides privacy to the iris templates with no loss of accuracy and trust in the comparator result.

Diabetic retinopathy is a common diabetic disease that affects the retina and can result to blindness if not treated initially. Deep learning (DL)-based models are proposed to detect the blood abnormalities in the retinal tissue due to diabetes mellitus obtained from fundus camera. The drawback with these models is the lack of performance. To address this, we propose to automate the process of detection of severity of diabetic retinopathy (DR) using ensembles of pretrained models, thus exploring the power of transfer learning in the field of automated diagnosis. Deep learning models perform well when the model is trained on a large amount of data. In this regard, we also put forth data augmentation and preprocessing techniques to generate the synthetic images and to improve image quality. Extensive experimental results on publicly available database illustrate that the proposed ensemble model achieves fair accuracy when compared to existing models. Thus, the proposed model shows good scope for deployment in real-time diagnosis.

Many government and commercial organizations are using biometric authentication systems instead of a password or token-based authentication systems. They are computationally expensive if more users are involved. To overcome this problem, a biometric system can be created and deployed in the cloud which then can be used as a biometric authentication service. Privacy is the major concern with cloud-based authentication services as biometric is irrevocable. Many biometric authentication systems based on cancelable biometrics are developed to solve the privacy concern in the past few years. But the existing methods fail to maintain the trade-off between speed, security, and accuracy. To overcome this, we present a multi-instance cancelable iris system (MICBTDL). MICBTDL uses a convolutional neural network trained using triplet loss for feature extraction and stores the feature vector as a cancelable template. Our system uses an artificial neural network as the comparator module instead of the similarity measures. Experiments are carried on IITD and MMU iris databases to check the effectiveness of MICBTDL. Experimental results demonstrate that MICBTDL accomplishes fair performance when compared to other existing works.

In this digital age, many learning technologies and tools are suitable for synchronous and asynchronous learning. There is an interaction between participants, instructor, and training in synchronous learning at fixed timing. In synchronous learning, there is real interaction between participants. In asynchronous learning, there is no real-time interaction between the participants. Students can learn at their own time and pace. So, in asynchronous learning, there is a need to understand whether the learner has understood the concepts. The evaluation can be achieved using H5P, an interactive course content creation tool. This study aims to measure the learning outcomes by making the students understand the concepts through the active learning experience. The learning enhancement is achieved by creating interactive content through H5P. The learners can study through the interactive content and revise the concept using the engagement, which leads to improved performance in their end exams. The participants included 60 engineering students of IV B. Tech Information Technology at a women-only engineering educational institution. The participants are allowed to watch prerecorded self-made videos, participate in activities like quiz at a particular duration of the video, and get feedback immediately. Summaries were also added at the end of the videos. The course instructor gets the report of all students’ participation status and scores of the entire class in the LMS platform Moodle. H5P helps the instructor understand the students’ learning difficulties, and it will be addressed enabling the attainment of improved learning outcomes.

Programming Skills play a crucial role in any computer engineering student's life to apply the concepts in solving any real world problem as well to crack a secure job in the dream company. To achieve this they should assess their performance in programming, analyze and improve their skills regularly. Many students are even undergoing mental stress and depression and even attempting suicides out of the stress if the considered scores and performance are not met. With the help of analyzing the programming skills one can enhance their scores and performance on a regular basis, introspect and can deliberately practice for better improvement. This reduces the stress, anxiety and depression on students' minds in securing good scores in their academics and in building their career to achieve the goal. This analysis helps even professors to improvise the teaching and learning outcomes of students and increase their performance in whichever field they are working in. We made a comparison of different machine learning algorithms based on 200 classification instances. This analysis helped us in analyzing the statistics of students' performance.

Lattice-based crypto systems are regarded as secure and believed to be secure even against quantum computers. lattice-based cryptography relies upon problems like the Shortest Vector Problem. Shortest Vector Problem is an instance of lattice problems that are used as a basis for secure cryptographic schemes. For more than 30 years now, the Shortest Vector Problem has been at the heart of a thriving research field and finding a new efficient algorithm turned out to be out of reach. This problem has a great many applications such as optimization, communication theory, cryptography, etc. This paper introduces the Shortest Vector Problem and other related problems such as the Closest Vector Problem. We present the average case and worst case hardness results for the Shortest Vector Problem. Further this work explore efficient algorithms solving the Shortest Vector Problem and present their efficiency. More precisely, this paper presents four algorithms: the Lenstra-Lenstra-Lovasz (LLL) algorithm, the Block Korkine-Zolotarev (BKZ) algorithm, a Metropolis algorithm, and a convex relaxation of SVP. The experimental results on various lattices show that the Metropolis algorithm works better than other algorithms with varying sizes of lattices.