Wearable Sensor-Based Behavioral User Authentication Using a Hybrid Deep Learning Approach with Squeeze-and-Excitation Mechanism

The tremendous development and rapid evolution in computing advancements has urged a lot of organizations to expand their data as well as computational needs. Such type of services offers security concepts like confidentiality, integrity, and availability. Thus, a highly secured domain is the fundamental need of cloud environments. In addition, security breaches are also growing equally in the cloud because of the sophisticated services of the cloud, which cannot be mitigated efficiently through firewall rules and packet filtering methods. In order to mitigate the malicious attacks and to detect the malicious behavior with high detection accuracy, an effective strategy named Multiverse Fractional Calculus (MFC) based hybrid deep learning approach is proposed. Here, two network classifiers namely Hierarchical Attention Network (HAN) and Random Multimodel Deep Learning (RMDL) are employed to detect the presence of malicious behavior. The network classifier is trained by exploiting proposed MFC, which is an integration of multi-verse optimizer and fractional calculus. The proposed MFC-based hybrid deep learning approach has attained superior results with utmost testing sensitivity, accuracy, and specificity of 0.949, 0.939, and 0.947.

COVID-19 has affected daily life in unprecedented ways, with dramatic changes in mental health, sleep time and level of physical activity. These changes have been especially relevant in the elderly population, with important health-related consequences. In this work, two different sensor technologies were used to quantify the energy expenditure of ageing adults. To this end, a technological platform based on Raspberry Pi 4, as an elaboration unit, was designed and implemented. It integrates an ambient sensor node, a wearable sensor node and a coordinator node that uses the information provided by the two sensor technologies in a combined manner. Ambient and wearable sensors are used for the real-time recognition of four human postures (standing, sitting, bending and lying down), walking activity and for energy expenditure quantification. An important first aim of this work was to realize a platform with a high level of user acceptability. In fact, through the use of two unobtrusive sensors and a low-cost processing unit, the solution is easily accessible and usable in the domestic environment; moreover, it is versatile since it can be used by end-users who accept being monitored by a specific sensor. Another added value of the platform is the ability to abstract from sensing technologies, as the use of human posture and walking activity for energy expenditure quantification enables the integration of a wide set of devices, provided that they can reproduce the same set of features. The obtained results showed the ability of the proposed platform to automatically quantify energy expenditure, both with each sensing technology and with the combined version. Specifically, for posture and walking activity classification, an average accuracy of 93.8% and 93.3% was obtained, respectively, with the wearable and ambient sensor, whereas an improvement of approximately 4% was reached using data fusion. Consequently, the estimated energy expenditure quantification always had a relative error of less than 3.2% for each end-user involved in the experimentation stage, classifying the high level information (postures and walking activities) with the combined version of the platform, justifying the proposed overall architecture from a hardware and software point of view.

Text classification is an essential component in a variety of applications of natural language processing. While the deep learning-based approach is becoming more popular, using vectors of word as an input for the models has proved to be a good way for the machine to learn the relation between words in a document. This paper proposes a solution for the text classification using hybrid deep learning approaches. Every existing deep learning approach has its own advantages and the hybrid deep learning model we are introducing is the combination of the superior features of CNN and LSTM models. The proposed models CNN-LSTM, LSTM-CNN show enhanced accuracy over another approach.

Cyberbullying on social media platforms has become widespread in society. Cyberbullying can take many forms, including hate speech, trolling, adult content, racism, harassment, or rants. One social media platform that has many cyberbullies is Twitter, which has been renamed 'X'. The anonymous nature of this 'X' platform allows users from all over the world to commit cyberbullying as they can freely share their thoughts and expressions without having to account for their identity. This research aims to explore the influence of IndoBERT’s semantic features on hybrid deep learning models for cyberbullying detection while integrating TF-IDF feature extraction and FastText feature expansion to enhance text classification performance. Specifically, this study examines how IndoBERT’s semantic capabilities affect the hybrid deep learning model in detecting cyberbullying on platform 'X'. This study has 30,084 tweets with a hybrid deep learning approach that combines CNN and LSTM. In the IndoBERT scenario, IndoBERT features were first combined with TF-IDF, then expanded using FastText before being applied to the hybrid deep learning model. The test results produced the highest accuracy rate by: CNN (80.69%), LSTM (80.67%), CNN- LSTM (81.18%), CNN-LSTM-IndoBERT (82.05%). This research contributes to informatics by integrating hybrid deep learning (CNN-LSTM) with IndoBERT and TF-IDF, demonstrating its effectiveness in improving cyberbullying detection in Indonesian text. Future research can explore the use of other transformer-based models such as RoBERTa or ALBERT to enhance contextual understanding in cyberbullying classification.

RIFATA: Remora improved invasive feedback artificial tree algorithm-enabled hybrid deep learning approach for root disease classification

In nearly all current systems, user authentication mechanism is one time and static. Although such type of user authentication is sufficient for many types of applications, in some scenarios, continuous or periodic re-verification of the identity is desirable, especially in high-security application. In this paper, we study user authentication based on 3D foot motion, which can be suitable for periodic identity re-verification purposes. Three-directional (3D) motion of the foot (in terms of acceleration signals) is collected using a wearable accelerometer sensor attached to the ankle of the person. Ankle accelerations from three directions (up-down, forward-backward and sideways) are analyzed for person authentication. Applied recognition method is based on detecting individual cycles in the signal and then finding best matching cycle pair between two acceleration signals. Using experimental data from 30 subjects, obtained EERs (Equal Error Rates) were in the range of 1.6–23.7% depending on motion directions and shoe types. Furthermore, by combining acceleration signals from 2D and 3D and then applying fusing techniques, recognition accuracies could be improved even further. The achieved performance improvements (in terms of EER) were up to 68.8%.

User authentication is a critical process in both corporate and home environments due to the ever-growing security and privacy concerns. With the advancement of smart cities and home environments, the concept of user authentication is evolved with a broader implication by not only preventing unauthorized users from accessing confidential information but also providing the opportunities for customized services corresponding to a specific user. Traditional approaches of user authentication either require specialized device installation or inconvenient wearable sensor attachment. This paper supports the extended concept of user authentication with a device-free approach by leveraging the prevalent WiFi signals made available by IoT devices, such as smart refrigerator, smart TV and thermostat, etc. The proposed system utilizes the WiFi signals to capture unique human physiological and behavioral characteristics inherited from their daily activities, including both walking and stationary ones. Particularly, we extract representative features from channel state information (CSI) measurements of WiFi signals, and develop a deep learning based user authentication scheme to accurately identify each individual user. Extensive experiments in two typical indoor environments, a university office and an apartment, are conducted to demonstrate the effectiveness of the proposed authentication system. In particular, our system can achieve over 94% and 91% authentication accuracy with 11 subjects through walking and stationary activities, respectively.

Recognition of locomotion mode is a crucial process for control of wearable soft robotic devices to assist humans in walking activities. We present a probabilistic Bayesian approach with a sequential analysis method for recognition of locomotion and phases of the gait cycle. Our approach uses recursive accumulation of evidence, as biological systems do, to reduce uncertainty present in the sensor measurements, and thus improving recognition accuracy. Data were collected from a wearable sensor, attached to the shank of healthy human participants, from three locomotion modes; level-ground walking, ramp ascent and ramp descent. We validated our probabilistic approach with recognition of locomotion in steady-state and gait phases in transitional states. Furthermore, we evaluated the effect, in recognition accuracy, of the accumulation of evidence controlled by increasing belief thresholds. High accuracy results achieved by our approach, demonstrate its potential for robust control of lower limb wearable soft robotic devices to provide natural and safe walking assistance to humans.

On wearable and Virtual Reality (VR) platforms, user authentication is a basic function, but usually a keyboard or touchscreen cannot be provided to type a password. Hand gesture and especially in-air-handwriting can be potentially used for user authentication because a gesture input interface is readily available on these platforms. However, determining whether a login request is from the legitimate user based on a piece of hand movement is challenging in both signal processing and matching, which leads to limited performance in existing systems. In this paper, we propose a multifactor user authentication framework using both the motion signal of a piece of in-air-handwriting and the geometry of hand skeleton captured by a depth camera. To demonstrate this framework, we invented a signal matching algorithm, implemented a prototype, and conducted experiments on a dataset of 100 users collected by us. Our system achieves 0.6% Equal Error Rate (EER) without spoofing attack and 3.4% EER with spoofing only data, which is a significant improvement compared to existing systems using the Dynamic Time Warping (DTW) algorithm. In addition, we presented an in-depth analysis of the utilized features to explain the reason for the performance boost.

Adaptive Bayesian inference system for recognition of walking activities and prediction of gait events using wearable sensors

Abstract- This research proposes a hybrid deep learning model for automated rice leaf disease detection, addressing the limitations of traditional methods. A pre-trained ResNet50V2 CNN is combined with a custom fully connected network, enhanced with batch normalization, dropout, and L2 regularization. Trained on 15,023 images from nine disease classes and healthy leaves, the model utilizes extensive data augmentation techniques such as rotation, shifts, zoom, and flipping to improve robustness. The model achieved 99.53% accuracy, surpassing existing benchmarks, and provides a reliable system for timely disease detection, enhancing rice crop management and yield. Keywords- Hybrid deep learning model, Rice leaf disease detection, Fully connected layer, Data augmentation, Dropout, Batch normalization, L2 regularization.

A hybrid deep learning approach for Assamese toxic comment detection in social media

Mammography and ultrasound based dual modality classification of breast cancer using a hybrid deep learning approach

Advancing breast ultrasound diagnostics through hybrid deep learning models

Alzheimer's disease (AD) is a progressive neurodegenerative disorder that primarily affects older adults, leading to memory loss, cognitive decline, and behavioural changes.It is marked by the accumulation of amyloid-beta plaques and tau tangles in the brain, which impair neuronal function and eventually cause cell death.Diagnosis typically involves clinical evaluations, cognitive assessments, and imaging techniques such as MRI (Magnetic Resonance Imaging) and PET (Positron Emission Tomography) scans.Early detection is crucial, as it allows for interventions that can potentially slow the disease's progression, enabling patients and their families to manage symptoms more effectively.The transition from Normal Control (NC) to Early Mild Cognitive Impairment (EMCI) may be reversible, whereas other stages are not.This makes it a particularly challenging classification problem, as the changes in biomarker values are typically minimal during the progression from NC to very mild stages.To address this issue, we created a hybrid deep learning model, GDP (combining GoogLeNet, DenseNet-121, and Principal Component Analysis (PCA)), to extract hybrid features using 26639 sMRI images from well-known MRI datasets, such as the Open Access Series of Imaging Studies (OASIS) and the Alzheimer's Disease Neuroimaging Initiative (ADNI).These features fed into the SVM binary classification algorithm to distinguish very mild cases from healthy images.We created synthetic images using CycleGAN augmentation to balance the dataset across labels and expand its size.The proposed GDP model stands out from current state-of-the-art deep learning techniques, such as ResNet-50, DenseNet-169, VGG-16, EfficientNet, and Inceptionv3, by achieving an impressive accuracy of 98.98% and an AUC of 99.91%.These results significantly surpass the performance of existing methods, making it a promising tool for early Alzheimer's disease prediction.