Research Overview

Dr. Rajesh Kumar
Assistant Professor, Department of Computer Science, Bucknell University

Research Contribution Overview

Dr. Rajesh Kumar is an internationally recognized researcher whose work at the intersection of behavioral biometrics, adversarial machine learning, and human-centered computing has redefined how human behavioral signals can secure, inform, and humanize digital systems. His central research vision is that patterns of human interaction, such as typing rhythm, motion dynamics, gaze behavior, and touch gestures, constitute a powerful behavioral fingerprint that can safeguard cyber infrastructure, preserve academic and social integrity, and enhance the reliability of artificial intelligence systems. Through a decade of rigorous inquiry and innovation, Dr. Kumar has developed a coherent and influential body of work organized into three interrelated thrusts: (1) Behavioral Biometrics for Cybersecurity and Privacy, (2) Behavioral Biometrics for Academic and Social Media Integrity, and (3) Behavioral Biometrics for Enhancing NLP Models for America. Each thrust is grounded in deep technical contributions and directly addresses issues of national importance to the United States.

Thrust 1: Behavioral Biometrics for Cybersecurity and Privacy

Dr. Kumar’s first major research thrust focuses on securing digital systems through continuous and context-aware authentication using behavioral biometrics. Identifying that passwords and static logins are insufficient (CCS’14) in a mobile and cloud-based world, he develops behavioral authentication methods that continuously verify user identity through typing patterns, touch gestures, and motion signals. His early work on Context-Aware Active Authentication (CVPRW’14, Thrust 1) pioneered context-aware user authentication. 

Dr. Kumar’s research also advanced the science of adversarial biometrics. In Treadmill Assisted Gait Spoofing (TAGS) (BTAS’15, Thrust 1; DTRAP’21, Thrust 1) and Dictionary Attack on IMU-based Gait Authentication (AISec/CCS’23, Thrust 1), he systematically modeled and analyzed adversarial threats to wearable sensor-based authentication. These studies revealed how machine-learning-driven adversaries could imitate user motion, prompting the development of new adversarial training countermeasures. 

His follow-up studies, Continuous Authentication of Smartphone Users by Fusing Typing, Swiping, and Phone Movement Patterns (BTAS’16, Thrust 1) and Continuous User Authentication via Unlabeled Phone Movement Patterns (IJCB’17, Thrust 1), refined these techniques into robust multimodal frameworks. These models not only authenticate continuously but also adapt to natural behavioral variations, addressing the long-standing trade-off between usability and security. His later works, including IDeAuth: A Novel Behavioral Biometric-based Implicit DeAuthentication Scheme for Smartphones (PRL’22, Thrust 1), GANTouch: An Attack-Resilient Framework for Touch-based Continuous Authentication System (TBIOM’22, Thrust 1), and iCTGAN: An Attack Mitigation Technique for Random-vector Attack on Accelerometer-based Gait Authentication Systems (IJCB’22, Thrust 1), used generative models and multimodal fusion to defend against evolving attacks.

Dr. Kumar’s research in this domain is widely cited. According to Google Scholar's thematic rankings, he is consistently ranked among the top 10 researchers in continuous authentication using behavioral biometrics.

Collectively, this body of work demonstrates a complete research cycle from threat modeling and vulnerability analysis to the development of robust countermeasures. These innovations directly align with the missions of U.S. cybersecurity agencies and defense organizations, which seek adaptive, privacy-preserving authentication methods to secure national infrastructure against advanced adversarial threats.

Thrust 2: Behavioral Biometrics for Academic and Social Media Integrity

As generative artificial intelligence systems have proliferated, Dr. Kumar recognized that traditional text-based plagiarism detectors were becoming obsolete. His second research thrust, therefore, redefined authorship verification by analyzing how content is produced rather than what it contains. His seminal paper, Keystroke Dynamics Against Academic Dishonesty in the Age of LLMs (IJCB’24b, Thrust 2), demonstrated that typing behavior can reliably distinguish genuine human authorship from AI-assisted writing, earning the Best Paper Award at IEEE IJCB 2024. Extending this work, Detecting LLM-Assisted Academic Dishonesty Using Keystroke Dynamics (TBIOM’25-R1, Thrust 2, under revision) introduced large-scale evaluations under three typing modes (bona fide, transcription, and paraphrasing LLM responses) that quantified subtle timing deviations caused by AI-generated text assistance.

This research expanded internationally with LLM-Assisted Cheating Detection in Korean Language via Keystrokes (IJCB’25, Thrust 2), proving the cultural and linguistic generalizability of behavioral authorship verification. In parallel, Dr. Kumar’s Active Authentication via Korean Keystrokes Under Varying LLM Assistance and Cognitive Contexts (ICMLA’25, Thrust 1) investigated how cognitive loads and LLM intervention impact typing rhythms in turn the performance of active authentication systems, establishing behavioral biometrics as a transparent and ethical indicator of authorship authenticity. This line of work is currently being expanded into code writing, evaluation under deception-based threat models, and low-resourced languages (Vietnamese). According to thematic citation data, Dr. Kumar is ranked among the top ten researchers worldwide in the field of plagiarism detection.

Dr. Kumar has also applied behavioral biometrics to protect the integrity of social and professional media ecosystems. His studies, Spotting Fake Profiles in Social Networks via Keystroke Dynamics (CCNC’24, Thrust 2) and Weak Links in LinkedIn: Enhancing Fake Profile Detection in the Age of LLMs (ASONAM’25, Thrust 2), built scalable behavioral frameworks for identifying fake or automated accounts without relying on invasive content inspection. These models are particularly timely as social media platforms and educational institutions struggle with the authenticity crisis triggered by generative AI.

By moving beyond content-based analysis to behavioral production signals, Dr. Kumar’s work provides a privacy-conscious solution to the growing challenge of AI-generated misinformation and academic dishonesty. His framework establishes a new research frontier in behavioral forensics that strengthens national education systems, social trust, and digital democracy.

Thrust 3: Behavioral Biometrics for Enhancing NLP Models for America

Dr. Kumar’s third major research thrust integrates behavioral signals into natural language processing systems, aiming to make AI models more accurate and human-centered. His study Synthesizing Human Gaze Feedback for Improved NLP Performance (EACL’23, Thrust 3) demonstrated that gaze-based attention data, whether real or synthetically generated, can significantly improve NLP model accuracy in sentiment and sarcasm detection tasks. This innovation introduced a new paradigm in which machine learning models benefit from approximations of human reading behavior, creating explainable systems that align with human cognition.

In A Touchless Typing Approach Using Apple Augmented Reality Kit and Seq2Seq Learning (TAPIA’24, Thrust 3), Dr. Kumar and his students developed an accessibility-focused text entry system that translates head movement and facial gestures into written text using sequence-to-sequence models. This project, which earned Second Prize at the ACM Tapia Student Research Competition in 2024, exemplifies how behavioral modeling can make computing more inclusive for individuals with disabilities.

Furthering this direction, Dr. Kumar introduced G-Loss: Graph-Guided Fine-Tuning of Language Models (LoG’25, Thrust 3) and How Well Do LLMs Imitate Human Writing Style (UEMCON’25, Thrust 3), where he examined the structural and stylistic alignment between human and AI-generated language. These studies bridge behavioral science and machine learning by embedding human behavioral data, such as gaze, head motion, and keystroke timing, directly into language model training pipelines. The result is a new class of behaviorally informed NLP systems that enhance robustness, accuracy, integrity, and accessibility.

This thrust carries immense national importance by promoting equitable access to technology and advancing the U.S. leadership in NLP and ethical AI.  

Impact, Recognition, and Mentorship

Dr. Kumar has authored over 28 peer-reviewed publications, with more than 950 citations on Google Scholar and H- and i10 indices of 13 and 15, respectively. His research has been published in premier venues including ACM CCS, ACM TISSEC/TOPS, IEEE TBIOM, IEEE IJCB, IEEE BTAS, CVPRW, and EACL. According to Web of Science, his work is cited across more than 30 countries, including the United States, Germany, South Korea, India, China, Australia, and the United Kingdom. His research appears in theses, patents, policy reports, and media outlets such as Vice, Business Insider, and Wikipedia.

He has completed 37 verified peer reviews, placing him in the 91st percentile globally, and his peer review-to-publication ratio ranks in the 93rd percentile. He received Best Reviewer Awards at IEEE IJCB in 2021, 2023, 2024, and 2025 and has served on three National Science Foundation review panels evaluating proposals totaling over $20 million in the area of secure and trustworthy cyberspace. His dedication to mentorship is equally strong, having guided more than twenty-five undergraduate researchers, many of whom have co-authored papers, earned national recognition, or entered prestigious graduate programs. His student-led projects have received national awards, including Second Prize at the 2024 ACM Tapia Poster Competition.

Integrated National and Scientific Impact

Across these three interconnected thrusts, Dr. Kumar has built an internationally recognized research program that redefines how behavioral data can be harnessed for the public good. His work in continuous authentication fortifies cybersecurity infrastructure essential for defense, government, and industry. His behavioral integrity research strengthens academic and social media ecosystems against deception, ensuring the authenticity of education and discourse in the era of AI. His behavioral NLP innovations advance inclusivity and fairness in artificial intelligence, supporting the U.S. mission to develop human-centered, trustworthy AI. Combined with his sustained record of publication, mentorship, and global recognition, Dr. Kumar’s research exemplifies depth, originality, and leadership. 

References 

(TBIOM’25, Thrust 2) R. Kumar, A. Mehta, A. Singla, K. Bisht, Y. K. Singla, and R. R. Shah, “Detecting LLM-Assisted Academic Dishonesty Using Keystroke Dynamics,” IEEE Transactions on Biometrics, Behavior, and Identity Science, 2025.

(LoG’25, Thrust 3) A. Sharma, V. Agarwal, R. Kumar, and H. S. Panchal, “G-Loss: Graph-Guided Fine-Tuning of Language Models,” Proceedings of the Learning on Graphs Conference, 2025.

(ASONAM’25, Thrust 2) A. Gulati, R. Kumar, V. Agarwal, and A. Sharma, “Weak Links in LinkedIn: Enhancing Fake Profile Detection in the Age of LLMs,” Proceedings of the IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining (ASONAM), 2025.

(ICMLA’25, Thrust 2) D. H. Roh and R. Kumar, “Active Authentication via Korean Keystrokes Under Varying LLM Assistance and Cognitive Contexts,” IEEE International Conference on Machine Learning and Applications (ICMLA), 2025.

(UEMCON’25, Thrust 3) R. Jemama and R. Kumar, “How Well Do LLMs Imitate Human Writing Style,” IEEE Annual Ubiquitous Computing, Electronics and Mobile Communication Conference (UEMCON), 2025.

(SECRYPT’25, Thrust 1) S. Gupta, R. Kumar, K. Raja, B. Crispo, and C. Maple, “Evaluating a Bimodal User Verification Robustness Against Synthetic Data Attacks,” Proceedings of the International Conference on Security and Cryptography (SECRYPT), 2025.

(IJCB’25, Thrust 2) D. H. Roh, R. Kumar, and A. Ngo, “LLM-Assisted Cheating Detection in Korean Language via Keystrokes,” IEEE International Joint Conference on Biometrics (IJCB), 2025.

(IJCB’24b, Thrust 2) D. Kundu, A. Mehta, R. Kumar, N. Lal, A. Anand, A. Singh, and R. R. Shah, “Keystroke Dynamics Against Academic Dishonesty in the Age of LLMs,” IEEE International Joint Conference on Biometrics (IJCB), 2024.

(IJCB’24a, Thrust 1) A. Ngo, R. Kumar, and P. Cao, “Deep Generative Attacks and Countermeasures for Data-Driven Offline Signature Verification,” IEEE International Joint Conference on Biometrics (IJCB), 2024.

(CCNC’24, Thrust 2) A. Kuruvilla, R. Daley, and R. Kumar, “Spotting Fake Profiles in Social Networks via Keystroke Dynamics,” IEEE Consumer Communications and Networking Conference (CCNC), 2024.

(TAPIA’24, Thrust 3) H. T. Ngo and R. Kumar, “A Touchless Typing Approach Using Apple Augmented Reality Kit and Seq2Seq Learning,” Proceedings of the ACM Richard Tapia Conference on Diversity in Computing, 2023.

(AISec/CCS’23, Thrust 1) R. Kumar, C. Isik, and C. K. Mohan, “Dictionary Attack on IMU-Based Gait Authentication,” Proceedings of the ACM Workshop on Artificial Intelligence and Security (AISec), co-located with ACM Conference on Computer and Communications Security (CCS), 2023.

(EACL’23, Thrust 3) V. Khurana, Y. K. Singla, N. Hollenstein, R. Kumar, and B. Krishnamurthy, “Synthesizing Human Gaze Feedback for Improved NLP Performance,” Proceedings of the 17th Conference of the European Chapter of the Association for Computational Linguistics (EACL), 2023.

(TBIOM’22, Thrust 1) P. Mehrotra, M. Agrawal, R. Kumar, and R. R. Shah, “GANTouch: An Attack-Resilient Framework for Touch-Based Continuous Authentication System,” IEEE Transactions on Biometrics, Behavior, and Identity Science, 2022.

(PRL’22, Thrust 1) S. Gupta, R. Kumar, M. Kacimi, and B. Crispo, “IDeAuth: A Novel Behavioral Biometric-Based Implicit Deauthentication Scheme for Smartphones,” Pattern Recognition Letters, 2022.

(IJCB’22, Thrust 1) J. H. Mo and R. Kumar, “iCTGAN: An Attack Mitigation Technique for Random-Vector Attack on Accelerometer-Based Gait Authentication Systems,” IEEE International Joint Conference on Biometrics (IJCB), 2022.

(DTRAP’21, Thrust 1) R. Kumar, C. Isik, and V. V. Phoha, “Treadmill Assisted Gait Spoofing (TAGS): An Emerging Threat to Wearable Sensor-Based Gait Authentication,” ACM Digital Threats: Research and Practice (DTRAP), 2021.

(IJCB’21, Thrust 1) P. Mehrotra, M. Agrawal, R. Kumar, and R. R. Shah, “Defending Touch-Based Continuous Authentication Systems from Active Adversaries Using Generative Adversarial Networks,” IEEE International Joint Conference on Biometrics (IJCB), 2021.

(ISBA’18, Thrust 1) R. Kumar, P. P. Kundu, and V. V. Phoha, “Continuous Authentication Using One-Class Classifiers and Their Fusion,” IEEE International Conference on Identity, Security and Behavior Analysis (ISBA), 2018.

(IJCB’17, Thrust 1) R. Kumar, P. P. Kundu, D. Shukla, and V. V. Phoha, “Continuous User Authentication via Unlabeled Phone Movement Patterns,” IEEE International Joint Conference on Biometrics (IJCB), 2017.

(BTAS’16, Thrust 1) R. Kumar, V. V. Phoha, and A. Serwadda, “Continuous Authentication of Smartphone Users by Fusing Typing, Swiping, and Phone Movement Patterns,” IEEE International Conference on Biometrics Theory, Applications, and Systems (BTAS), 2016.

(ACM TISSEC/TOPS’16, Thrust 1) A. Serwadda, V. V. Phoha, Z. Wang, R. Kumar, and D. Shukla, “Toward Robotic Robbery on the Touch Screen,” ACM Transactions on Information and System Security (TISSEC/TOPS), 2016.

(BTAS’15, Thrust 1) R. Kumar, V. V. Phoha, and A. Jain, “Treadmill Attack on Gait-Based Authentication Systems,” IEEE International Conference on Biometrics Theory, Applications, and Systems (BTAS), 2015.

(CVPRW’14, Thrust 1) A. Primo, V. V. Phoha, R. Kumar, and A. Serwadda, “Context-Aware Active Authentication Using Smartphone Accelerometer Measurements,” IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), 2014.

(CCS’14, Thrust 1) D. Shukla, R. Kumar, A. Serwadda, and V. V. Phoha, “Beware, Your Hands Reveal Your Secrets,” Proceedings of the ACM Conference on Computer and Communications Security (CCS), 2014.