Department of Computer Science and Information Systems

Permanent URI for this collectionhttp://localhost:4000/handle/123456789/1928

Browse

Author: search.filters.author.Bhatia, AshutoshSubject: search.filters.subject.Deep learning

Search Results

Now showing 1 - 5 of 5
  • No Thumbnail Available
    Item
    Bayesian deep learning meets self-attention: a risk-aware approach to advertisement optimization
    (IEEE, 2025-05) Bhatia, Ashutosh; Tiwari, Kamlesh
    In the highly competitive landscape of e-commerce advertising, maximizing Return on Advertising Spend (ROAS) is critical, yet remains inherently uncertain due to auction-based bidding dynamics and fluctuating market conditions. Traditional deterministic models fail to capture this uncertainty, necessitating a probabilistic approach that balances predictive accuracy with interpretability. To address this challenge, the paper proposes a novel Hierarchical Bayesian Deep Learning framework that integrates a Bayesian Belief Network (BBN) for structured probabilistic reasoning and a Mixture Density Network (MDN) for full distributional modeling of ROAS. The BBN models dependencies among campaign variables, offering interpretable insights, while the hierarchical deep learning architecture overcomes scalability limitations in high-dimensional settings through self-attention mechanisms. Experiments demonstrate up to 22.8% lower RMSE and 27.4% better Negative Log Likelihood (NLL) and up to 31.2% lower Kullback-Leibler divergence (KLD) than state-of-the-art methods (DeepAR, Prophet, NGBoost), achieving an R2 of 98% with an inference speed of 5.2 ms per campaign, making real-time bidding feasible. Ablation studies confirm that attention-driven feature selection and calibrated uncertainty quantification significantly enhance both predictive performance and explainability, identifying key drivers of campaign success. By providing precise, uncertainty-aware, and explainable predictions, this approach enables adaptive bidding strategies, optimized budget allocation, and risk management, setting a new benchmark for intelligent decision-making in digital advertising.
  • No Thumbnail Available
    Item
    Deep learning approaches for driver distraction detection using driver facing cameras: literature review and empirical study using cnn classifiers on a 100-driver image dataset
    (2025-05) Bhatia, Ashutosh; Sharma, Yashvardhan; Tiwari, Kamlesh
    Distracted driving contributes to thousands of fatalities and injuries globally. According to India’s Ministry of Road Transport and Highways (MoRTH), distraction-related behaviors such as rear-end and off-road collisions accounted for nearly one-fourth of all traffic incidents in 2022. The U.S. National Highway Traffic Safety Administration (NHTSA) reported 3,275 deaths and over 324,000 injuries from distraction-related crashes in 2023. In Europe, the European Road Safety Observatory (ERSO) observed handheld phone use by drivers in up to 9.4% of vehicles across member states, with self-reported texting rates reaching 53%. Despite numerous studies and surveys on driver distraction detection, existing literature remains fragmented, often combining multiple sensor modalities or distraction with related driver states such as fatigue. Prior empirical efforts also lack a unified benchmarking strategy to assess model generalization under shifts in viewpoint or spectral input. This paper presents a focused survey and empirical study of visiononly distraction detection using deep learning models applied to driver-facing camera inputs. It introduces a conceptual model linking behavioral cues to cognitive distraction, defines the visionbased Driver Distraction Detection (vDDD) system with alert logic, and develops structured taxonomies of datasets, architectures, and learning strategies. Using the 100-Driver dataset, the empirical study evaluates 26 CNN classifiers under 64 crossdomain configurations, systematically analyzing generalization across modality and camera view changes. Results show that frontal RGB-trained models generalize better than their NIRtrained counterparts and that lightweight models trade off accuracy under rare class scenarios for faster inference. The study establishes the vDDD paradigm as a vision-based behavioral modeling approach for distraction detection using driver-facing camera data. It outlines future research directions in spectrumaligned augmentation, attention modeling, and lightweight visuallanguage fusion, emphasizing deployment-focused strategies such as quantization, contrastive learning, and progressive fine-tuning.
  • No Thumbnail Available
    Item
    User Profiling Using Smartphone Network Traffic Analysis
    (IEEE, 2021) Bhatia, Ashutosh; Tiwari, Kamlesh
    The recent decade has witnessed phenomenal growth in communication technology. Development of user friendly software platforms, such as Facebook, WhatsApp etc. have facilitated ease of communication and thereby people have started freely sharing messages and multimedia over the Internet. Further, there is a shift in trends with services being accessed from smartphones over personal computers. To protect the security and privacy of the smartphone users, most of the applications use encryption that encapsulates communications over the Internet. However, research has shown that the statistical information present in a traffic can be used to identify the application, and further, the activity performed by the user inside that application. In this paper, we extend the scope of analysis by proposing a learning framework to leverage application and activity data to profile smartphone users in terms of their gender, profession age group etc. This will greatly help the authoritative agencies to conduct their investigations related to national security and other purposes
  • No Thumbnail Available
    Item
    SmartDriveAuth: Enhancing Vehicle Security with Continuous Driver Authentication via Wearable PPG Sensors and Deep Learning
    (Springer, 2024-04) Bhatia, Ashutosh; Tiwari, Kamlesh
    The paper introduces a novel approach for continuous driver authentication in vehicle security, utilizing wearable photoplethysmography (PPG) sensors and Long Short-Term Memory (LSTM)–based deep learning. This study aims to overcome the limitations of traditional one-time authentication (OTA) methods, which typically involve passwords, PINs, or physical keys. While effective for initial identity verification, these conventional methods do not continuously validate the driver’s identity during vehicle operation. The proposed system leverages an LSTM-based prediction model to efficiently predict the subsequent PPG values using the raw PPG signals from wrist-worn devices. The predicted values are continuously compared with actual real-time data (received from the sensors) for authentication. The proposed system eliminates the need to permanently store user biometrics in a database. Motion artifacts and momentary disruptions have minimal impact on system performance. Experimental validation was conducted with 15 participants driving in varied conditions to simulate real-life driving conditions. The study evaluated the system’s accuracy, achieving an Equal Error Rate (EER) of 4.8%, demonstrating its potential as a viable solution for continuous driver authentication in dynamic environments.
  • No Thumbnail Available
    Item
    Enhancing security through continuous biometric authentication using wearable sensors
    (Elsevier, 2024) Bhatia, Ashutosh; Tiwari, Kamlesh
    The paper presents a novel approach for biometric continuous driver authentication (CDA) for secure and safe transportation using wearable photoplethysmography (PPG) sensors and deep learning. Conventional one-time authentication (OTA) methods, while effective for initial identity verification, fail to continuously verify the driver’s identity during vehicle operation, potentially leading to safety, security, and accountability issues. To address this, we propose a system that employs Long Short-Term Memory (LSTM) models to predict subsequent PPG values from wrist-worn devices and continuously compare them with real-time sensor data for authentication. Our system calculates a confidence level representing the probability that the current user is the authorized driver, ensuring robust availability to genuine users while detecting impersonation attacks. The raw PPG data is directly fed into the LSTM model without pre-processing, ensuring lightweight processing. We validated our system with PPG data from 15 volunteers driving for 15 min in varied conditions. The system achieves an Equal Error Rate (EER) of 4.8%. Our results demonstrate that the system is a viable solution for CDA in dynamic environments, ensuring transparency, efficiency, accuracy, robust availability, and lightweight processing. Thus, our approach addresses the main challenges of classical driver authentication systems and effectively safeguards passengers and goods with robust driver authentication.