Embedded Files
Below, is the outline of current main areas of focus in Cyber Innovations Lab, where we aim to contribute to cutting-edge advancements in technology and security. Our research revolves around five primary categories, each representing a critical aspect of the evolving landscape in computer sciences and cybersecurity.
1. Cyber Systems Security & Privacy
The security and privacy of sensitive data remain paramount in today’s digital landscape. Through our research in applied cryptography and security, we aim to develop sophisticated cryptographic algorithms and secure communication protocols. By exploring cryptographic primitives, encryption techniques, and secure key management, we strive to fortify the foundations of secure data transmission and storage
2. Trustworthy AI in Cybersecurity
In this research area, we investigate both how machine learning can enhance cybersecurity defenses and how to protect AI systems from adversarial attacks. Our recent work develops intelligent security mechanisms for threat detection while simultaneously creating novel methodologies to protect AI models from manipulation through adversarial inputs. This dual approach ensures we build both effective and resilient AI-powered security solutions.
3. Decentralized Systems and Applications: Architecture, Security, and Real-Word Implementation
Our research in decentralized systems encompasses both foundational architectures and their practical applications. continue to advance the field through innovative P2P frameworks, as demonstrated in our recent IEEE Internet of Things Journal publication and other significant works. Our current research focuses on location-aware cryptography and secure fog computing architectures, particularly in healthcare applications. By integrating blockchain technology and advanced cryptographic protocols, we address critical challenges in scalability, security, and efficiency
4. Brain-Computer Interfaces and their Security
In this emerging area, we advance the frontiers of neural computing and its security implications. Our recent work, published in IEEE Transactions on Cognitive and Developmental Systems, presents a comprehensive review of EEG-to-text conversion technologies, analyzing current methodologies, challenges, and future directions in brain signal decoding while considering privacy and security implications . This research aims to establish novel methodologies for secure neural data processing and transmission, addressing critical challenges in brain-computer interface security.
5. Digital Warfare & Cybersecurity Policy
Our research in cyber warfare examines the evolving landscape of digital conflicts, defensive strategies, and their implications for national security. This work encompasses the analysis of sophisticated cyber threats, development of counter-measures, and investigation of emerging battlefield dynamics in the digital domain. Through policy-focused research, we develop frameworks for enhancing organizational and national cyber resilience while addressing modern threats. This includes creating comprehensive approaches for securing e-commerce systems through risk management and compliance frameworks, ensuring robust protection for digital transactions and critical infrastructure.
6. Computing Sciences Education
As an educator and researcher, We are am committed to advancing cybersecurity education through innovative approaches . Our current work, supported by the NSF NAIRR pilot program, focuses on developing advanced machine learning curricula for cybersecurity education. As Curriculum Developer for the IN-CITE program ($1.5M DOL grant), I lead the development of cyber warfare training curricula bridging academic and military needs. I also successfully led USM’s effort to achieve the NSA CAE-CD designation, strengthening our institution’s position in cybersecurity education.
Saydul Akbar Murad- PhD Candidate
Derek Taylor
PhD Student
Oluseyi Olukola(Ola)
PhD Student
Tomas Nader- MS Student
Ian Harris - MS Student
Zachariah McCullough - MS
Alkendria McNair - MS
Ashim Dahal Undergrad- Lab Admin
Ankit Ghimire Undergrad
Luana Ferreira
Undergrad
Pappu Jha
Undergrad
Mukesh Pudel Undergrad
Hanzela Hamid - Undergrad
EEG-to-Text Translation: A Model for Deciphering Human Brain Activity
Saydul Akbar Murad, Ashim Dahal, Nick Rahimi, arXiv, 2025
We propose a new model, R1 Translator, which aims to improve the performance of EEG-to-text decoding.
Multi-Lingual Cyber Threat Detection in Tweets/X Using ML, DL, and LLM: A Comparative Analysis
Saydul Akbar Murad, Ashim Dahal, Nick Rahimi, arXiv, 2025
Threats over X (Twitter) are given on multiple languages. This paper proposes a new dataset and methodology to detect cyber threats spread over tweets on X.
Accepted: IEEE Transactions on Computational Social Systems
Redemption Score: A Multi-Modal Evaluation Framework for Image Captioning via Distributional, Perceptual, and Linguistic Signal Triangulation
Ashim Dahal, Ankit Ghimire, Saydul Akbar Murad, Nick Rahimi, arXiv2025
A robust framework to evaluate image-text pairs under perceptual, semantic, pragmatic and distributional alignment.
Embedding Shift Dissection on CLIP: Effects of Augmentations on VLM's Representation Learning
Ashim Dahal, Saydul Akbar Murad, Nick Rahimi
MIV at CVPR (Proceedings Track), 2025
How exactly does the representation on ViT model like CLIP change when we apply different levels and kinds of augmentations to them? This short paper explains this very question.
Heuristical Comparison of Vision Transformers Against Convolutional Neural Networks for Semantic Segmentation on Remote Sensing Imagery
Ashim Dahal, Saydul Akbar Murad, Nick Rahimi
IEEE Sensors Journal Impact Factor: 4.3, 2025
Analysis of Vision Transformers (ViT) against Convolutional Neural Networks (UNet CNN) for image segmentation on Remote Sensing iSAID dataset. We propose a novel loss function that helps a smaller CNN model to perform equally to a 5x larger ViT model.
Efficiency Bottlenecks of Convolutional Kolmogorov-Arnold Networks: A Comprehensive Scrutiny with ImageNet, AlexNet, LeNet and Tabular Classification
Ashim Dahal, Saydul Akbar Murad, Nick Rahimi, arXiv, 2025
This paper analyzes Convolutional Kolmogorov Arnold Networks on ImageNet with Alexnet, MNIST with LeNet and Tabular CNN modification with MoA datasets.
Cyber Warfare: Strategies, Impacts, and Future Directions in the Digital Battlefield [Link]
Nick Rahimi, Henry Jones
Journal of Information Security 16 (2), 252-269
This paper explores the evolution of cyber warfare, its significant impacts on global security, and the urgent need to establish the legal and ethical guidelines currently lacking in this domain.
Human Activity Recognition Using an Ensemble Learning Approach [Link]
Tomas Nader, SA Murad, Nick Rahimi
International Conference on Computers and Their Applications, 98-112 [BEST PAPER AWARD WINNER]
This research proposes a new ensemble learning method, using a meta-model to combine predictions, to overcome common Human Activity Recognition (HAR) challenges like low accuracy and poor generalization in complex datasets.
Heuristical Comparison of Vision Transformers Against Convolutional Neural Networks for Semantic Segmentation on Remote Sensing Imagery
Ashim Dahal, Saydul Akbar Murad, Nick Rahimi
IEEE Sensors Journal Impact Factor: 4.3, 2025
Analysis of Vision Transformers (ViT) against Convolutional Neural Networks (UNet CNN) for image segmentation on Remote Sensing iSAID dataset. We propose a novel loss function that helps a smaller CNN model to perform equally to a 5x larger ViT model.
Google Sites
Report abuse