Harshvardhan

I'm

About Me

I am Harshvardhan Takawale, a first year computer science PhD student at the University of Maryland College Park working at iCoSMoS Lab with Prof. Nirupam Roy.

In the past, I have worked as a Lead Researcher at Silence Laboratories Singapore. I finished my B.E at BITS, Pilani in August 2020.


Research interests: Ambient Computing, Intelligent Acoustics, Ubiquitous Healthcare, Security, Applied Machine Learning.

Recent Updates

10-23 : Our paper "Scribe: Simultaneous Voice and Handwriting Interface" is accepted at IMWUT 2024.

04-23 : I served on the Artifact Evaluation committee of ACM MobiSys 2023.

04-23 : Our paper "Structure Assisted Spectrum Sensing for Low-power Acoustic Event Detection" is accepted at IASA CPS IoT Week 2023.

02-23 : Our poster "Natural Voice Interface for the Next Generation of Smart Spaces" is accepted at ACM HotMobile 2022.

03-22 : I served on the Replicability committee of WiSec 2022.

03-22 : Received Dean's Fellowship at UMDCS

03-22 : I'll be joining Prof. Nirupam Roy at University of Maryland Computer Science (UMDCS) as a Ph.D student in Fall 2022.

12-21 : Our paper "A Seamless Second-Factor Verification as a Mobility-Service for Future Commute Stations" is accepted at ACM HotMobile 2022

Resume

Here is my short resume. The PDF can also be found (last updated - 11/05/23) here.

Education

Ph.D. in Computer Science

2022 - present

University of Maryland, College Park

At UMD, I work with Prof. Nirupam Roy at the iCoSMoS lab.

Bachelor of Engineering Electrical & Electronics

2016 - 2020

Birla Institute of Technology & Science, Pilani

At BITS Pilani I graduated with a Bachelor in Engineering with a major in Electrical and Electronics and minor in Data Science.

Professional Experience

Lead Researcher

Jul 2020 - Present

Silence Laboratories & SUTD, Singapore

Lead Research and implementation of the the following projects

  • Co-location based Seamless Authentication for Mobility Services
    • Designed a robust system aimed at verifying co-presence of two or multi-party systems using correlated observations of a) acoustic features and b) time series encoding of events.
    • Achieved authentication accuracy of >92\% across different use cases of second factor authentication.
    • Enhanced robustness to fit adverse and noisy scenarios of application use cases.
  • Landmark extraction from infrastructure for indoor mobility claim verification
    • Developed a proof-of-visit and proof-of-route module using fusion of mobile sensor based human activity detection engine and WiFi channel.
    • Added gait based continuous authentication module for fraud detection.
    • Operational in SUTD for tracking campus guards for more than a year.
  • Real Time Gesture Tracking
    • Developed a robust and real-time pipeline to do 3D tracking of human gestures performed using a mobile device.
    • Verified its robustness with a <2% false-positive rate in a multi-user study.

Security Research Intern

Jan 2020 - Jul 2020

Cyber Security Research Centre @ NTU, Singapore

  • Title: Lightweight Malware Detection for Embedded Systems (Undergraduate thesis)
  • Created an malware detection model using hardware performance counters as features for advanced driver-assistance system (ADAS). Measured changes in 6 registers - cycles, instructions, cache-references, cache-misses, branches, branch-misses. Achieved an accuracy of 92% with an F1-score of 0.9149. Awarded the highest grade for the thesis.

Research Intern

Jun 2019 - Sep 2019

Electrical Engineering & Computer Sciences @ UC Berkeley, Singapore

  • Project Title: Malware Detection on Highly Imbalanced Data through Sequence Modeling(undergraduate thesis)
  • Performed dynamic analysis on mobile application activity sequences for the purpose of malware detection on highly imbalanced dataset. Used the state-of-the-art language representation model BERT, to create a sequential model and achieved an F1 score of 0.919 with just 0.5% of the examples being malware in the dataset.

Mitacs Summer Scholar

May 2019 - Aug 2019

School of Computer Science @ University of Windsor, Canada

  • Project Title: Integrating Continuous Authentication into the Personal Health Record Applications
  • Developed a protocol to authenticate users based on their interaction with the phone using anomaly in inertial sensor data. Analysed across 6 different classification models and achieved 95–97% accuracy for each, when tested using tenfold cross validation.

Contact

Location:

IRB 3245
Brendan Iribe Center for Computer Science and Engineering
University of Maryland, College Park

Schedule a meeting:

Click Here

Call:

+1 (240)-854-8549

-->