5 papers
SEAL-Tag: Self-Tag Evidence Aggregation with Probabilistic Circuits for PII-Safe Retrieval-Augmented Generation
Jin Xie, Songze Li, Guang Cheng
Retrieval-Augmented Generation (RAG) systems introduce a critical vulnerability: contextual leakage, where adversaries exploit instruction-following to exfiltrate Personally Identi…
Odysseus: Jailbreaking Commercial Multimodal LLM-integrated Systems via Dual Steganography
Songze Li, Jiameng Cheng, Yiming Li +2
By integrating language understanding with perceptual modalities such as images, multimodal large language models (MLLMs) constitute a critical substrate for modern AI systems, par…
ReCIT: Reconstructing Full Private Data from Gradient in Parameter-Efficient Fine-Tuning of Large Language Models
Jin Xie, Ruishi He, Songze Li +2
Parameter-efficient fine-tuning (PEFT) has emerged as a practical solution for adapting large language models (LLMs) to custom datasets with significantly reduced computational cos…
TUNI: A Textual Unimodal Detector for Identity Inference in CLIP Models
Songze Li, Ruoxi Cheng, Xiaojun Jia
The widespread usage of large-scale multimodal models like CLIP has heightened concerns about the leakage of PII. Existing methods for identity inference in CLIP models require que…
MSfusion: A Dynamic Model Splitting Approach for Resource-Constrained Machines to Collaboratively Train Larger Models
Jin Xie, Songze Li
Training large models requires a large amount of data, as well as abundant computation resources. While collaborative learning (e.g., federated learning) provides a promising parad…