PBPramesh Baral

Pramesh Baral

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Pramesh Baral has a total of 4 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Privacy Sentinel: A Multilingual Context-Aware Framework for Real-Time PII Detection and Risk-Based Redaction

    by Pramesh Baral, Raghab Singh, Akshey Sigdel, Rista Baral

    The fast rate of development of web content has augmented the risk of privacy violations, specifically, the incidental disclosure of Personally Identifiable Information (PII). To cope with this problem, this paper introduces Privacy Sentinel, a multi-layered system, which is capable of identifying and redacting sensitive information in unstructured and semi-structured web content in real time. The system uses the latest context-sensitive transformer models, such as DeBERTa-v3 and hybrid RoBERTa-LLM models, to detect possible PII leakage, and calculate the risk dynamically depending on the context information. Privacy Sentinel implements a modular design, which is a microservice-based backend based on FastAPI and Redis with a lightweight browser extension on the client. The extension communicates directly with the Document Object Model (DOM), which allows the redaction to be reversible in order to allow sensitive information to be concealed temporarily and still make it available to authorized users when needed. We test the system on big web text corpus and multilingual datasets, where it has a high performance of about 96% accuracy, 94% recall and a F1-score of about 95, which is higher than the traditional regex based systems, single RoBERTa models and the Privacy Shield baseline. The system also has a low latency of approximately 50 milliseconds per DOM element, which ensures that the user can easily switch between redaction on or off without interfering with the page readability. Altogether, our findings indicate that the integration of sophisticated NLP frameworks with the workable system design can offer an efficient and convenient solution to the security of confidential data in contemporary internet-based settings.

  • Open Access

    A Machine Learning Framework for Detecting and Mitigating Adversarial Prompt Attacks in Large Language Models

    by Raghab Singh, Pramesh Baral, Binamra Neupane

    Large Language Models LLMs have changed numerous industries including education cybersecurity healthcare robotics and Internet of Things IoT by empowering them to converse and comprehend the context and generate intelligent arguments to power intelligent systems. Since these models are deployed as part of software platforms robotics and linked infrastructures their reliability and trustworthiness becomes a key issue. Nonetheless, the bulk of the available literature is concerned with the optimization of model performances and its usage and security risks and adversarial threats are not given much consideration. New attacks like jailbreaking can bypass safety nets and cause great harm to users and systems based on LLMs like multi turn manipulation and injection by emergency attacks. This paper fills this gap by building and analyzing a varied dataset of prompt and safe and adversarial prompts generated by human professionals and advanced LLMs including GPT4o and Claude 3.5 Sonnet. The controversial intent is detected by use of TF IDF feature extraction and machine learning classifiers such as K Nearest Neighbors Decision Tree Logistic Regression random forest Naive Bayes Support Vector machine and ensemble model. The SVM model is the most successful model with 98.4 percent accuracy and 98.7 percent F1 score as well as 99.7 percent recall.


  • Open Access

    Anticipatory Autonomic Management of Poly-Cloud Environments: A Machine Intelligence Paradigm

    by Pramesh Baral

    This work introduces a new type of machine intelligence aiming to improve resilience in complex multi-cloud systems. With the help of advanced algorithms like anomaly detection and trend forecasting, this technique enables it to process telemetry from various clouds, to detect when an infrastructure failure will start occurring and when a resource will run out, before services are affected. A graduated automation remediation engine integrates with cloud orchestration tools. Dynamic and appropriate response actions are enabled to dramatically reduce the time to resolve incidents and service outages compared to reactive approaches. This framework can enhance smart budgeting via scaling predictions and balancing workloads. It has been tested on different degradation scenarios. This presents a work that offers unified resilience governance as a remedy for the challenges of cross-provider visibility and agile data functioning. Moreover, it reduces service reliability and operational cost in distributed cloud environments.


  • Open Access

    Uncovering Neural Dynamics: A Statistical Approach to Analyzing Brain Activity Across Timescales

    by Pramesh Baral

    A novel approach is proposed to assess the temporal evolution of neural activity. A brain-reading technique that uses its own pattern recognition and re-corrects for major errors lets researchers investigate some regions of the brain. It provides an alternative method for inferring outdoor air quality that takes into account factors not taken into account by other methods. Simulation results combined with real-life large-scale brain-data show our method to be correct and computationally efficient. The study plays a crucial role in improving our understanding of the brain, its processes, and its functions.

  • Open Access

    Integrating Spatially Continuous Environmental Covariates into HMMs for Animal Behavior via Kriging

    by Pramesh Baral

    The advent of high-resolution animal movement tracking technologies has significantly advanced our ability to study animal behavior in natural environments. Hidden Markov Models (HMMs) have emerged as a powerful tool for decoding behavioral states, such as exploratory versus encamped movements, based on telemetry data. However, a common challenge lies in incorporating spatially continuous environmental variables that influence animal behavior, as these covariates are often collected at different times and locations than the movement data. In this study, we present an innovative approach that integrates kriging, a spatial interpolation technique, with HMMs to account for continuous environmental covariates in animal behavior modeling. By first fitting a continuous-time correlated random walk (CTCRW) model to animal tracks, followed by spatial interpolation of environmental data, we predict the necessary covariate values at the regular time intervals required by HMMs. This methodology allows researchers to incorporate complex environmental dynamics, such as changes in temperature or habitat structure, into behavioral state analysis. The proposed framework enhances the precision of behavioral predictions, providing a more nuanced understanding of how animals interact with their environment across both space and time.

  • Open Access

    Deep Learning for Text-Based Sentiment Prediction

    by Pramesh Baral

    Understanding sentiment in textual data is essential for analyzing user opinions and trends. This study explores a deep learningbased sentiment classification approach using recurrent neural networks (LSTM). The model is trained on the IMDb dataset and achieves an accuracy of 86.0%, outperforming traditional machine learning baselines. Results demonstrate that LSTM effectively captures contextual dependencies, enhancing sentiment analysis capabilities in real-world applications.


  • Open Access

    Healthcare Copilot: A Modular Framework for Safe and Dynamic Medical Consultations Using Large Language Models

    by Pramesh Baral

    This paper introduces a modular Healthcare Copilot system powered by large language models (LLMs), designed to enhance the quality, safety, and accuracy of medical consultations. The system is composed of three principal components: a Dialogue Module for interactive engagement, a Memory Module for contextual continuity, and a Processing Module for automated summarization and reporting. Evaluated across simulated telehealth scenarios, the Copilot demonstrates superior performance in coherence, inquiry depth, response accuracy, and ethical compliance compared to baseline LLMs. This framework offers a scalable solution for LLM deployment in medical assistance, eliminating the need for extensive fine-tuning while enabling structured, expert-supervised interactions.

  • Open Access

    Unobserved Health: The Impact of Reporting Error on Health Dynamics

    by Pramesh Baral

    This paper examines the limitations of using self-reported health status (SRHS) as a direct measure of true health in dynamic economic models. Motivated by empirical evidence from major panel datasets (MEPS, HRS, and PSID) showing duration dependence and violations of the Markov property in SRHS transitions, we introduce a latent health model that accounts for transitory reporting error and individual heterogeneity. The model treats health as a continuous latent variable following an autoregressive process with mixture-distributed shocks, mapped to discrete SRHS outcomes through individual-specific reporting thresholds. Estimation results reveal strong persistence in latent health (ρ ≈ 0.9) and systematic reporting heterogeneity: older and less-educated individuals report worse SRHS for identical latent health levels. The model outperforms standard Markov specifications in capturing observed transition dynamics and demonstrates strong predictive validity through external validation using mortality data and labor force outcomes. Our framework enables more accurate policy simulations for social insurance programs and provides a template for handling noisy ordinal outcomes in other domains of applied microeconomics.