Ilia Shumailov

Google (United States) — United States

Computer Science · Physical Sciences

16h-index1.4kcitations114works12.22yr avg
OpenAlex↗ORCID↗Google Scholar↗Lab page↗

Accepting Students?

No reports yet. Know if this professor is taking students?

Research Topics

Adversarial Robustness in Machine Learning(42), Anomaly Detection Techniques and Applications(19), Topic Modeling(15), Privacy-Preserving Technologies in Data(12), Advanced Malware Detection Techniques(12)

Publications114 total

Iterative Deployment Improves Planning Skills in LLMs
ArXiv.org·2025Open Access
Iterative Deployment Improves Planning Skills in LLMs
arXiv (Cornell University)·2025Open Access
Cascading Adversarial Bias from Injection to Distillation in Language Models
2025Open Access
Extracting alignment data in open models
arXiv (Cornell University)·2025Open Access
When Vision Fails: Text Attacks Against ViT and OCR
2025Open Access
The Attacker Moves Second: Stronger Adaptive Attacks Bypass Defenses Against Llm Jailbreaks and Prompt Injections
arXiv (Cornell University)·2025Open Access
SynthID-Image: Image watermarking at internet scale
arXiv (Cornell University)·2025Open Access
Reasoning Introduces New Poisoning Attacks Yet Makes Them More Complicated
arXiv (Cornell University)·2025Open Access
Beyond Laplace and Gaussian: Exploring the Generalized Gaussian Mechanism for Private Machine Learning
arXiv (Cornell University)·2025Open Access
Cascading Adversarial Bias from Injection to Distillation in Language Models
arXiv (Cornell University)·2025Open Access
Machine Learning Models Have a Supply Chain Problem
arXiv (Cornell University)·2025Open Access
Exploring the limits of strong membership inference attacks on large language models
ArXiv.org·2025Open Access
Architectural Backdoors for Within-Batch Data Stealing and Model Inference Manipulation
arXiv (Cornell University)·2025Open Access
Lessons from Defending Gemini Against Indirect Prompt Injections
arXiv (Cornell University)·2025Open Access
Architectural Neural Backdoors from First Principles
2025· 2 cited
Watermarking Needs Input Repetition Masking
arXiv (Cornell University)·2025Open Access
Locking Machine Learning Models into Hardware
2025· 1 cited
SEA: Shareable and Explainable Attribution for Query-Based Black-Box Attacks
2025
Inexact Unlearning Needs More Careful Evaluations to Avoid a False Sense of Privacy
2025· 5 cited
Author Correction: AI models collapse when trained on recursively generated data
Nature·2025· 2 citedOpen Access
page 1 of 6Next →

Frequent Co-authors

Nicolas Papernot(23), Ross Anderson(22), Yiren Zhao(17), Robert Mullins(11), Zakhar Shumaylov(4), Yarin Gal(4), Franziska Boenisch(4), Adam Dziedzic(3), Roei Schuster(3), Ali Shahin Shamsabadi(3), Nicholas Boucher(3), Jamie Hayes(3), Xitong Gao(3), Anvith Thudi(3), Itay Yona(3), Yiren Zhao(2), Po-Sen Huang(2), Robert Stanforth(2), Borja Balle(2), Pushmeet Kohli(2)