Yulian Sun
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Yulian Sun has a total of 7 co-authors. Here are the top co-authors by number of shared publications:
- Li Duan
Huawei Technologies Düsseldorf
4 shared publications - Yong Li
Huawei Technologies Düsseldorf
4 shared publications - Asja Fischer
Ruhr University Bochum
1 shared publication - Derui Zhu
Technical University of Munich, Munich, Germany
1 shared publication - Ricardo Mendes
CISUC and Department of Informatics Engineering, University of Coimbra, Coimbra, Portugal
1 shared publication - Vedant Bonde
Huawei Technologies Düsseldorf
1 shared publication
Latest Publications
- 20 June 2025
- Published by Springer Nature
- p. 391-414
Obfuscation for Deep Neural Networks Against Model Extraction: Attack Taxonomy and Defense Optimization
- 20 June 2025
- Published by Springer Nature
- p. 391-414
Obfuscation for Deep Neural Networks Against Model Extraction: Attack Taxonomy and Defense Optimization
- 1 April 2025
- Published by Springer Nature
- p. 29-38
Short Paper: PSY: Posterior Sampling Based Privacy Enhancer in Large Language Models
- 1 April 2025
- Published by Springer Nature
- p. 29-38
Short Paper: PSY: Posterior Sampling Based Privacy Enhancer in Large Language Models
- Open Access
PSY: Posterior Sampling Based Privacy Enhancer in Large Language Models
Privacy vulnerabilities in LLMs, such as leakage from memorization, have been constantly identified, and various mitigation proposals have been proposed. LoRA is usually used in fine-tuning LLMs and a good entry point to insert privacy-enhancing modules. In this ongoing research, we introduce PSY, a Posterior Sampling based PrivacY enhancer that can be used in LoRA. We propose a simple yet effective realization of PSY using posterior sampling, which effectively prevents privacy leakage from intermediate information and, in turn, preserves the privacy of data owners. We evaluate LoRA extended with PSY against state-of-the-art membership inference and data extraction attacks. The experiments are executed on three different LLM architectures fine-tuned on three datasets with LoRA. In contrast to the commonly used differential privacy method, we find that our proposed modification consistently reduces the attack success rate. Meanwhile, our method has almost no negative impact on model fine-tuning or final performance. Most importantly, PSY reveals a promising path toward privacy enhancement with latent space extensions.
Open AccessPSY: Posterior Sampling Based Privacy Enhancer in Large Language Models
Privacy vulnerabilities in LLMs, such as leakage from memorization, have been constantly identified, and various mitigation proposals have been proposed. LoRA is usually used in fine-tuning LLMs and a good entry point to insert privacy-enhancing modules. In this ongoing research, we introduce PSY, a Posterior Sampling based PrivacY enhancer that can be used in LoRA. We propose a simple yet effective realization of PSY using posterior sampling, which effectively prevents privacy leakage from intermediate information and, in turn, preserves the privacy of data owners. We evaluate LoRA extended with PSY against state-of-the-art membership inference and data extraction attacks. The experiments are executed on three different LLM architectures fine-tuned on three datasets with LoRA. In contrast to the commonly used differential privacy method, we find that our proposed modification consistently reduces the attack success rate. Meanwhile, our method has almost no negative impact on model fine-tuning or final performance. Most importantly, PSY reveals a promising path toward privacy enhancement with latent space extensions.
- 6 September 2024
- Published by Springer Nature
- p. 390-409
Exploiting Internal Randomness for Privacy in Vertical Federated Learning
- 6 September 2024
- Published by Springer Nature
- p. 390-409
Exploiting Internal Randomness for Privacy in Vertical Federated Learning