About Me

Welcome to my homepage! My name is Pratik and I am a PhD Candidate at the Computer Science Department at Stony Brook University. I am a member of the Ethos Security and Privacy Lab directed by Amir Rahmati.

My research focuses on making reliable and fair machine learning accessible to all! Towards this effort, I am currently working on developing efficient methods for training certifiably/provably robust neural networks. This research is being conducted in collaboration with Kevin Eykholt of IBM Research.

During my PhD studies, I had the opportunity to contribute towards the developement of a robust biometric system as an Applied Scientist Intern at Amazon (Summer "20 & '21). Prior to joining the PhD program, I received a MS in Computer Science degree in 2018 from Stony Brook University. My MS Thesis was on generating temporal action proposals in long untrimmed videos. I received my Bachelors degree in Electronics Engineering in 2016 from Sardar Vallabhbhai National Institute of Technology, Surat, India.

When I am not working, I like to read, kayak and play video games!

---

Recent News

Oct '22	Will be serving as a reviewer for CVPR 2023.
Oct '22	Won the Best Overall Poster award at the Graduate Research Day held in my department!
Oct '22	Our preliminary work on the feasibility of compressing certifiably robust neural networks was accepted at NeurIPS 2022 workshop on Trustworthy and Socially Responsible ML!
Sept '22	Our paper on accelerating the process of training certifiably robust neural networks was accepted at NeurIPS 2022!
Sept '22	My advisor Amir Rahmati wins the Meta Towards Trustworthy Products in AR, VR, and Smart Devices and Security Research grant!
Aug '22	I will be presenting our paper on accelerating adversarial training at USENIX Security 2022 in Boston, MA.
May '22	My collaborator Farhan Ahmed will be virtually presenting the system we developed for efficiently evaluating robutness of ML systems at IEEE S&P 2022.
Mar '22	Will be serving as a reviewer for ECCV 2022 and NeurIPS 2022.

Publications

• On the Feasibility of Compressing Certifiably Robust Neural Networks
  Pratik Vaishnavi, Veena Krish, Farhan Ahmed, Kevin Eykholt, Amir Rahmati
  NeurIPS 2022, Workshop on Trustworthy and Socially Responsible ML
  venue

• Accelerating Certified Robustness Training via Knowledge Transfer
  Pratik Vaishnavi, Kevin Eykholt, Amir Rahmati
  NeurIPS 2022
  PDF code venue

• Ares: A System-Oriented Wargame Framework for Adversarial ML
  Farhan Ahmed, Pratik Vaishnavi, Kevin Eykholt, Amir Rahmati
  IEEE S&P 2022, Workshop on Deep Learning Security
  PDF code venue

• Transferring Adversarial Robustness Through Robust Representation Matching
  Pratik Vaishnavi, Kevin Eykholt, Amir Rahmati
  USENIX Security 2022
  PDF code video venue

• Can Attention Masks Improve Adversarial Robustness?
  Pratik Vaishnavi, Tianji Cong, Kevin Eykholt, Atul Prakash, Amir Rahmati
  AAAI 2020, Workshop on Engineering Dependable and Secure ML Systems
  PDF venue

• Robust Pose Recognition Using Deep Learning
  Aparna Mohanty, Alfaz Ahmed, Trishita Goswami, Arpita Das, Pratik Vaishnavi, Rajiv Ranjan Sahay
  CVIP 2016
  PDF venue

• Nrityabodha: Towards Understanding Indian Classical Dance Using a Deep Learning Approach
  Aparna Mohanty, Pratik Vaishnavi, Prerana Jana, Anubhab Majumdar, Alfaz Ahmed, Trishita Goswami, Rajiv Ranjan Sahay
  Signal Processing: Image Communication 2016
  PDF venue

---

Hosted on GitHub Pages — Theme by orderedlist