Harshit Kumar

Research Areas. Deep Learning, Malware Detection, Hardware Security


I am a fifth-year Ph.D. candidate in Electrical and Computer Engineering at Georgia Tech, advised by Prof. Saibal Mukhopadhyay. My current research is in AI-driven system modeling, with a focus on malware detection and discrete dynamical systems (e.g., forest fire).

In cybersecurity, I’ve led research on Hardware-based Malware Detectors, addressing AI modeling challenges arising from limitations in behavioral profiling of malware. More recently, in the Science of AI program (sponsored by Office of Naval Research), my work focuses on adapting AI training and evaluation methods for stochastic dynamical systems.

Rooted in first principles, my problem-solving approach has enabled me to adapt and innovate in various fields. During internships in research labs at New York University, Intel, and Whiterabbit.ai, I’ve tackled diverse challenges from semiconductor supply chain security to ransomware detection and medical imaging. My research ethos is centered on addressing fundamental problems with the potential for significant real-world impact.

Some specific things I’ve worked on, by topic:

  • Deep Learning
    • DNN Architectures (Transformer, Convolutional, Recurrent) for spatio-temporal prediction
    • Evaluation strategies for stochastic dynamical systems (e.g., forest fire evolution)
  • Application of Machine Learning in Malware Detection
    • Uncertainty Estimation for detecting concept drift
    • Statistical Models for learning malicious behavior in a time-series under noisy labels
  • Medical Imaging
    • Transformer-based DNN architecture for breast cancer detection in 3D mammograms
  • Hardware Security
    • Logic Locking for semiconductor supply chain security
    • Camouflaging for supply chain security of superconducting circuits (used in quantum computers)
  • Signal Processing
    • De-noising algorithms for single-channel speech enhancement