Rohan Chandrashekar

Peer-Reviewed Conferences & Journals

• Acoustic Horizons: Low-Latency Assistive Navigation Systems
  • Venue: IEEE International Conference on Emerging Technologies in Computer Science for Interdisciplinary Applications (ICETCS) - To be published 2025
  • Abstract: This paper introduces an innovative wearable lightweight, wireless headset designed to assist visually impaired individuals in navigating their environment more effectively and safely. The device aims to address common challenges faced by this demographic, including obstacle detection and spatial orientation, thereby enhancing their ability to move independently and confidently in various settings using 2 low powered esp32 camera modules which provide live, low-latency video feeds to the respective YOLOv8 model based on the scene which act as an extremely low cost alternative to AI cameras which are orders of magnitude more expensive.
• Detection of Bank Transaction Anomalies using Gradient Boosted Federated Learning
  • Venue: Submitted to IEEE Access (Under review to address reviewer comments)
  • Abstract: The advent of the internet has brought forth a new era in digital technology. This has revolutionized many businesses and sectors particularly banking and retail. Although this transformation has solved many problems and created numerous opportunities, it has also given rise to many challenges, the foremost being security. The increase in fraudulent transactions has been a major concern for both banks and customers. While customers are at risk of losing their savings, fraudulent transactions negatively affect the reputation of retail banks. The accurate and timely detection of bank transactions as fraudulent ranks among the most critical challenges confronting the banking sector. This study aims to research and introduce two novel machine learning techniques to address this pressing challenge. We leveraged the decentralized capabilities of the Federated Learning framework to bolster data security and privacy while analyzing sensitive and confidential data from banking transactions. What sets our methods apart is the strategic application of federated learning with gradient-boosting algorithms, celebrated for their adeptness in managing significantly imbalanced datasets. This innovative approach not only enhances our ability to detect anomalies in banking transactions but also minimizes false positives and achieves greater accuracy. The outcome of our research shows that by adopting these techniques, we can mitigate data privacy concerns and achieve significant improvements in both the detection of anomalies and the efficiency of the process. Thus, it paves the way for a transformative impact of this method in the realm of contemporary online banking security.
• Mental Health in the Digital Era-NLP Models for Depression and Suicidal Tendency Detection
  • Venue: Fifth International Conference on Computing and Network Communications (CoCoNet'23) - To be published in Springer Lecture Notes in Electrical Engineering Proceedings of CoCoNet 2023, Volume 2 in Jan 2025
  • Abstract: The advent of Natural Language Processing (NLP) as a novel technology, especially with the integration of the BERT (Bidirectional Encoder Representations from Transformers) model, presents a transformative approach in addressing mental health issues in today’s digital landscape. BERT’s advanced capabilities in understanding the context and nuances of language enable more accurate detection of mental health concerns, including depression and suicidal tendencies, in user-generated text. This exploration presents an in-depth exploration of a cutting-edge NLP-based model designed to detect signs of mental health issues, depression, and suicidal tendencies in user-generated text. The authors emphasise the importance of leveraging the advancements in AI and machine learning for early detection of mental health issues. We introduce a novel approach using NLP models, distinguishing it from the traditional machine learning models that are commonly employed. The end goal is to facilitate better intervention methods, offering a more comprehensive view of an individual’s mental well-being, especially when a vast majority of communication has shifted to digital platforms.
• PyEdit Pro: The Ultimate Advanced Text Editor in Python - Empowering Text Editing Experience
  • Venue: IEEE International Conference on Multidisciplinary Research in Technology and Management (MRTM) 2023 - To be published in IEEE XPlore in Jan 2025. (Recipient of the Best Paper Award)
  • Abstract: In today's fast-paced digital world, text editors have become more than just tools; they're essential for a wide range of professional and creative tasks. From coding to content writing, the versatility of these editors is remarkable. In our exploration, we dive into the creation of a cutting-edge text editor using Python, a language known for its flexibility and power. This research unpacks the core principles of design, highlighting the features and techniques that are crucial in building a robust and user-friendly editor. Key areas like interface design, text manipulation, syntax highlighting, and plugin support are thoroughly examined. We also shine a light on the complexities and challenges encountered during development, offering insights that are particularly useful for Python enthusiasts and developers. Lastly, the paper ventures into potential future research directions in this ever-evolving field, inviting further innovation and exploration.

Industry White Papers & Technical Reports

• Detecting Malicious Activities and Behavorial Analysis of Runtime Containers
  • Venue: HPE Parasparam 2024
  • Limited Technical Description: I contributed to designing and implementing an innovative container security solution that integrates Falco Sidekick with a custom web UI and Slack plugin for real-time alerting. My work focused on developing the web UI, configuring Slack integration, and streamlining deployment, enabling comprehensive monitoring and immediate incident response. This solution, validated in controlled environments, enhances the security posture of runtime containers.
• Securing Large Language Models
  • Venue: HPE Parasparam 2024
  • Limited Technical Description: In the white paper 'Securing Large Language Models,' I contributed to developing a comprehensive security framework that identifies and mitigates vulnerabilities in LLMs, such as adversarial attacks and data breaches. My work involved simulating cybersecurity scenarios in a controlled environment and implementing robust defense mechanisms, including encryption and anomaly detection, aligned with international standards like NIST and ISO.