Efficient and Private: Memorisation under differentially private parameter-efficient fine-tuning in language models
Ma, Olivia, Passerat-Palmbach, Jonathan, Usynin, Dmitrii
–arXiv.org Artificial Intelligence
Fine-tuning large language models (LLMs) for specific tasks introduces privacy risks, as models may inadvertently memorise and leak sensitive training data. While Differential Privacy (DP) offers a solution to mitigate these risks, it introduces significant computational and performance trade-offs, particularly with standard fine-tuning approaches. Previous work has primarily focused on full-parameter updates, which are computationally intensive and may not fully leverage DPs potential in large models. In this work, we address these shortcomings by investigating Parameter-Efficient Fine-Tuning (PEFT) methods under DP constraints. We show that PEFT methods achieve comparable performance to standard fine-tuning while requiring fewer parameters and significantly reducing privacy leakage. Furthermore, we incorporate a data poisoning experiment involving intentional mislabelling to assess model memorisation and directly measure privacy risks. Our findings indicate that PEFT methods not only provide a promising alternative but also serve as a complementary approach for privacy-preserving, resource-efficient fine-tuning of LLMs.
arXiv.org Artificial Intelligence
Nov-24-2024
- Country:
- North America
- United States
- Massachusetts > Plymouth County
- Hanover (0.04)
- California
- Los Angeles County > Los Angeles (0.14)
- Santa Clara County > Santa Clara (0.04)
- Massachusetts > Plymouth County
- Canada > Ontario
- Toronto (0.04)
- United States
- Europe
- Romania > Sud - Muntenia Development Region
- Giurgiu County > Giurgiu (0.04)
- Germany > Bavaria
- Upper Bavaria > Munich (0.04)
- Denmark > Capital Region
- Copenhagen (0.04)
- Romania > Sud - Muntenia Development Region
- Asia
- Myanmar > Tanintharyi Region
- Dawei (0.04)
- Middle East
- Jordan (0.04)
- UAE > Abu Dhabi Emirate
- Abu Dhabi (0.04)
- Myanmar > Tanintharyi Region
- North America
- Genre:
- Research Report > New Finding (1.00)
- Industry:
- Information Technology > Security & Privacy (1.00)
- Technology: