Paper 'How does Watermarking Affect Visual Language Models in Document Understanding?' accepted to Conference on Language Modeling (COLM) 2025.
Paper 'Secure Embedding Aggregation for Cross-Silo Federated Representation Learning' accepted to IEEE Transactions on Information Forensics & Security.
Paper 'PriRoAgg: Achieving Robust Model Aggregation with Minimum Privacy Leakage for Federated Learning' accepted to IEEE Transactions on Information Forensics & Security.
Paper 'Resolving the Efficiency-Utility Dilemma of Threshold Linearly Homomorphic Encryption via Message-Space Adapter' accepted to Crypto 2025.
Paper 'SecurePay: Enabling Secure and Fast Payment Processing for Platform Economy' accepted to IWQoS 2025.
Paper 'TUNI: A Textual Unimodal Detector for Identity Inference in CLIP Models' accepted to NAACL 2025 Workshop on Privacy in NLP (Oral).
Paper 'DeDe: Detecting Backdoor Samples for SSL Encoders via Decoders' accepted to CVPR 2025.
Research Experience
Leads the S3IC lab, working on a wide range of research topics to improve the security, scalability, and trustworthiness of distributed computing frameworks, with a focus on applications of machine learning, artificial intelligence, and blockchain.
Background
Research interests include safety, privacy, and security of large language models; secure multi-party computation for AI; blockchain scalability and privacy.
Miscellany
Looking for strongly motivated PhD and Master students, Undergraduate students, Post-docs, Research assistants, and Visitors/interns to join the lab. Interested applicants please email your CV, transcript, and any related publications to songzeli@seu.edu.cn or songzeli8824@outlook.com.