Alert button
Picture for Ilia Shumailov

Ilia Shumailov

Alert button

Architectural Neural Backdoors from First Principles

Feb 10, 2024
Harry Langford, Ilia Shumailov, Yiren Zhao, Robert Mullins, Nicolas Papernot

Viaarxiv icon

Buffer Overflow in Mixture of Experts

Feb 08, 2024
Jamie Hayes, Ilia Shumailov, Itay Yona

Viaarxiv icon

Revisiting Block-based Quantisation: What is Important for Sub-8-bit LLM Inference?

Oct 21, 2023
Cheng Zhang, Jianyi Cheng, Ilia Shumailov, George A. Constantinides, Yiren Zhao

Figure 1 for Revisiting Block-based Quantisation: What is Important for Sub-8-bit LLM Inference?
Figure 2 for Revisiting Block-based Quantisation: What is Important for Sub-8-bit LLM Inference?
Figure 3 for Revisiting Block-based Quantisation: What is Important for Sub-8-bit LLM Inference?
Figure 4 for Revisiting Block-based Quantisation: What is Important for Sub-8-bit LLM Inference?
Viaarxiv icon

Beyond Labeling Oracles: What does it mean to steal ML models?

Oct 03, 2023
Avital Shafran, Ilia Shumailov, Murat A. Erdogdu, Nicolas Papernot

Viaarxiv icon

Human-Producible Adversarial Examples

Sep 30, 2023
David Khachaturov, Yue Gao, Ilia Shumailov, Robert Mullins, Ross Anderson, Kassem Fawaz

Figure 1 for Human-Producible Adversarial Examples
Figure 2 for Human-Producible Adversarial Examples
Figure 3 for Human-Producible Adversarial Examples
Figure 4 for Human-Producible Adversarial Examples
Viaarxiv icon

SEA: Shareable and Explainable Attribution for Query-based Black-box Attacks

Aug 23, 2023
Yue Gao, Ilia Shumailov, Kassem Fawaz

Figure 1 for SEA: Shareable and Explainable Attribution for Query-based Black-box Attacks
Figure 2 for SEA: Shareable and Explainable Attribution for Query-based Black-box Attacks
Figure 3 for SEA: Shareable and Explainable Attribution for Query-based Black-box Attacks
Figure 4 for SEA: Shareable and Explainable Attribution for Query-based Black-box Attacks
Viaarxiv icon

LLM Censorship: A Machine Learning Challenge or a Computer Security Problem?

Jul 20, 2023
David Glukhov, Ilia Shumailov, Yarin Gal, Nicolas Papernot, Vardan Papyan

Figure 1 for LLM Censorship: A Machine Learning Challenge or a Computer Security Problem?
Figure 2 for LLM Censorship: A Machine Learning Challenge or a Computer Security Problem?
Viaarxiv icon

Gradients Look Alike: Sensitivity is Often Overestimated in DP-SGD

Jul 01, 2023
Anvith Thudi, Hengrui Jia, Casey Meehan, Ilia Shumailov, Nicolas Papernot

Figure 1 for Gradients Look Alike: Sensitivity is Often Overestimated in DP-SGD
Figure 2 for Gradients Look Alike: Sensitivity is Often Overestimated in DP-SGD
Figure 3 for Gradients Look Alike: Sensitivity is Often Overestimated in DP-SGD
Figure 4 for Gradients Look Alike: Sensitivity is Often Overestimated in DP-SGD
Viaarxiv icon

Machine Learning needs its own Randomness Standard: Randomised Smoothing and PRNG-based attacks

Jun 24, 2023
Pranav Dahiya, Ilia Shumailov, Ross Anderson

Figure 1 for Machine Learning needs its own Randomness Standard: Randomised Smoothing and PRNG-based attacks
Figure 2 for Machine Learning needs its own Randomness Standard: Randomised Smoothing and PRNG-based attacks
Figure 3 for Machine Learning needs its own Randomness Standard: Randomised Smoothing and PRNG-based attacks
Figure 4 for Machine Learning needs its own Randomness Standard: Randomised Smoothing and PRNG-based attacks
Viaarxiv icon