6 papers
PrivacyScalpel: Enhancing LLM Privacy via Interpretable Feature Intervention with Sparse Autoencoders
Ahmed Frikha, Muhammad Reza Ar Razi, Krishna Kanth Nakka +3
Large Language Models (LLMs) have demonstrated remarkable capabilities in natural language processing but also pose significant privacy risks by memorizing and leaking Personally I…
PII-Scope: A Comprehensive Study on Training Data PII Extraction Attacks in LLMs
Krishna Kanth Nakka, Ahmed Frikha, Ricardo Mendes +2
In this work, we introduce PII-Scope, a comprehensive benchmark designed to evaluate state-of-the-art methodologies for PII extraction attacks targeting LLMs across diverse threat…
Robust Skin Color Driven Privacy Preserving Face Recognition via Function Secret Sharing
Dong Han, Yufan Jiang, Yong Li +2
In this work, we leverage the pure skin color patch from the face image as the additional information to train an auxiliary skin color feature extractor and face recognition model…
ObfuscaTune: Obfuscated Offsite Fine-tuning and Inference of Proprietary LLMs on Private Datasets
Ahmed Frikha, Nassim Walha, Ricardo Mendes +3
This work addresses the timely yet underexplored problem of performing inference and finetuning of a proprietary LLM owned by a model provider entity on the confidential/private da…
IncogniText: Privacy-enhancing Conditional Text Anonymization via LLM-based Private Attribute Randomization
Ahmed Frikha, Nassim Walha, Krishna Kanth Nakka +3
In this work, we address the problem of text anonymization where the goal is to prevent adversaries from correctly inferring private attributes of the author, while keeping the tex…
PII-Compass: Guiding LLM training data extraction prompts towards the target PII via grounding
Krishna Kanth Nakka, Ahmed Frikha, Ricardo Mendes +2
The latest and most impactful advances in large models stem from their increased size. Unfortunately, this translates into an improved memorization capacity, raising data privacy c…