Matching Pairs: Attributing Fine-Tuned Models to their Pre-Trained Large Language Models
- Myles Foley
- Ambrish Rawat
- et al.
- 2023
- ACL 2023
Bio
Ambrish Rawat is a Research Scientist in the AI Security & Privacy team at IBM. His research interests are at the cross-sections of security, privacy and Artificial Intelligence (AI). Most recently, he has worked on Privacy Enhancing Technologies (PETs) like Federated Learning and Differential Privacy. He is passionate about building trustworthy AI systems with security and privacy guarantees within the regulatory demands of GDPR as well as EU AI and Digital Acts.
He holds a Master of Philosophy in Machine Learning and Machine Intelligence from the University of Cambridge, UK, and a Master of Technology in Mathematics and Computing from the Indian Institute of Technology, Delhi (IIT Delhi). He joined IBM in 2016 and has since been leading and contributing to numerous efforts in AI and ML at the Dublin Research Lab.
His work has been published at top AI conferences and he's an active contributor to open source software projects. He have been recognised as Master Inventor at IBM for his contributions to IBM patent portfolio and has also received Research Division Award and several Outstanding Technical Accomplishment Awards for the contributions to the vast array of cutting-edge research at IBM.
Publications
Machine Learning Platform for Extreme Scale Computing on Compressed IoT Data
- 2022
- Big Data 2022
Federated Continual Learning with Differentially Private Data Sharing
- 2022
- NeurIPS 2022
The Devil is in the GAN: Backdoor Attacks and Defenses in Deep Generative Models
- 2022
- ESORICS 2022
Robust Learning Protocol for Federated Tumor Segmentation Challenge
- 2022
- MICCAI 2022
Challenges and Pitfalls of Bayesian Unlearning
- Ambrish Rawat
- James Requeima
- et al.
- 2022
- ICML 2022
Federated Unlearning: How to Efficiently Erase a Client in FL?
- Anisa Halimi
- Swanand Ravindra Kadhe
- et al.
- 2022
- ICML 2022
Towards an Accountable and Reproducible Federated Learning: A FactSheets Approach
- Nathalie Baracaldo Angel
- Ali Anwar
- et al.
- 2022
- arXiv
Bandit Limited Discrepancy Search and Application to Machine Learning Pipeline Optimization
- Akihiro Kishimoto
- Djallel Bouneffouf
- et al.
- 2022
- AAAI 2022
Searching for Machine Learning Pipelines Using a Context-Free Grammar
- Radu Marinescu
- Akihiro Kishimoto
- et al.
- 2021
- AAAI 2021
Patents
Learning Input Preprocessing To Harden Machine Learning Models
- 19 Jun 2023
- US
- 11681796
Automated Generation Of A Machine Learning Pipeline
- 10 Apr 2023
- US
- 11625632
Automated Decision Making For Neural Architecture Search
- 30 Jan 2023
- US
- 11568249
Adding Adversarial Robustness To Trained Machine Learning Models
- 16 May 2022
- US
- 11334671
Providing Adversarial Protection For Electronic Screen Displays
- 28 Mar 2022
- US
- 11288408
Protecting A Machine Learning Model
- 14 Jun 2021
- US
- 11036857
Providing Adversarial Protection Of Speech In Audio Signals
- 18 Jan 2021
- US
- 10896664
Projects
Incremental Machine Learning for Extreme Scale Computing
AutoML for incremental machine learning algorithms for big time-series data.
Blog posts
Researchers develop defenses against deep learning hack attacks
ReleaseAmbrish Rawat, Killian Levacher, and Mathieu Sinn7 minute read- Adversarial Robustness and Privacy
- Data and AI Security
- Generative AI
- Security
- Trustworthy AI
Top collaborators
GZ
Giulio Zizzo
Research Scientist
PR
Parikshit Ram
Research Staff Member
MP
Mark Purcell
Research Manager - Security, AI Security & PETs
ST
