12 citations · 23 across the 25 of their papers we have counts for
25 papers
HW-Router: Hardware-Aware Routing for Scalable Multi-LLM Serving
Ahasan Kabir, Jiaqi Xue, Mengxin Zheng +1
Modern large language model (LLM) serving platforms deploy multiple models across different GPUs, requiring routers to direct incoming queries to appropriate LLMs. However, existin…
INFRAMIND: Infrastructure-Aware Multi-Agent Orchestration
Ahasan Kabir, Jiaqi Xue, Mengxin Zheng +1
Existing multi-agent LLM orchestration methods, ranging from brute-force ensembles to learned routers, select models and topologies based on task and model features. However, these…
RobPI: Robust Private Inference against Malicious Client
Jiaqi Xue, Mengxin Zheng, Qian Lou
The increased deployment of machine learning inference in various applications has sparked privacy concerns. In response, private inference (PI) protocols have been created to allo…
R2-Router: A New Paradigm for LLM Routing with Reasoning
Jiaqi Xue, Qian Lou, Jiarong Xing +1
As LLMs proliferate with diverse capabilities and costs, LLM routing has emerged by learning to predict each LLM's quality and cost for a given query, then selecting the one with h…
PRO: Enabling Precise and Robust Text Watermark for Open-Source LLMs
Jiaqi Xue, Yifei Zhao, Mansour Al Ghanim +4
Text watermarking for large language models (LLMs) enables model owners to verify text origin and protect intellectual property. While watermarking methods for closed-source LLMs a…
DictPFL: Efficient and Private Federated Learning on Encrypted Gradients
Jiaqi Xue, Mayank Kumar, Yuzhang Shang +5
Federated Learning (FL) enables collaborative model training across institutions without sharing raw data. However, gradient sharing still risks privacy leakage, such as gradient i…
Securing Transformer-based AI Execution via Unified TEEs and Crypto-protected Accelerators
Jiaqi Xue, Yifei Zhao, Mengxin Zheng +3
Recent advances in Transformer models, e.g., large language models (LLMs), have brought tremendous breakthroughs in various artificial intelligence (AI) tasks, leading to their wid…
SoK: Can Fully Homomorphic Encryption Support General AI Computation? A Functional and Cost Analysis
Jiaqi Xue, Xin Xin, Wei Zhang +11
Artificial intelligence (AI) increasingly powers sensitive applications in domains such as healthcare and finance, relying on both linear operations (e.g., matrix multiplications i…
TFHE-Coder: Evaluating LLM-agentic Fully Homomorphic Encryption Code Generation
Mayank Kumar, Jiaqi Xue, Mengxin Zheng +1
Fully Homomorphic Encryption over the torus (TFHE) enables computation on encrypted data without decryption, making it a cornerstone of secure and confidential computing. Despite i…
CipherPrune: Efficient and Scalable Private Transformer Inference
Yancheng Zhang, Jiaqi Xue, Mengxin Zheng +4
Private Transformer inference using cryptographic protocols offers promising solutions for privacy-preserving machine learning; however, it still faces significant runtime overhead…
Evaluating the Robustness and Accuracy of Text Watermarking Under Real-World Cross-Lingual Manipulations
Mansour Al Ghanim, Jiaqi Xue, Rochana Prih Hastuti +3
We present a study to benchmark representative watermarking methods in cross-lingual settings. The current literature mainly focuses on the evaluation of watermarking methods for t…
BadFair: Backdoored Fairness Attacks with Group-conditioned Triggers
Jiaqi Xue, Qian Lou, Mengxin Zheng
Attacking fairness is crucial because compromised models can introduce biased outcomes, undermining trust and amplifying inequalities in sensitive applications like hiring, healthc…
DataSeal: Ensuring the Verifiability of Private Computation on Encrypted Data
Muhammad Husni Santriaji, Jiaqi Xue, Qian Lou +1
Fully Homomorphic Encryption (FHE) allows computations to be performed directly on encrypted data without needing to decrypt it first. This "encryption-in-use" feature is crucial f…
CryptoTrain: Fast Secure Training on Encrypted Dataset
Jiaqi Xue, Yancheng Zhang, Yanshan Wang +3
Secure training, while protecting the confidentiality of both data and model weights, typically incurs significant training overhead. Traditional Fully Homomorphic Encryption (FHE)…
BoostCom: Towards Efficient Universal Fully Homomorphic Encryption by Boosting the Word-wise Comparisons
Ardhi Wiratama Baskara Yudha, Jiaqi Xue, Qian Lou +2
Fully Homomorphic Encryption (FHE) allows for the execution of computations on encrypted data without the need to decrypt it first, offering significant potential for privacy-prese…
BadRAG: Identifying Vulnerabilities in Retrieval Augmented Generation of Large Language Models
Jiaqi Xue, Mengxin Zheng, Yebowen Hu +3
Retrieval-Augmented Generation (RAG) enhances Large Language Models (LLMs) by retrieving relevant information from external knowledge bases to provide more accurate, contextually i…
CR-UTP: Certified Robustness against Universal Text Perturbations on Large Language Models
Qian Lou, Xin Liang, Jiaqi Xue +3
It is imperative to ensure the stability of every prediction made by a language model; that is, a language's prediction should remain consistent despite minor input variations, lik…
TrojFair: Trojan Fairness Attacks
Mengxin Zheng, Jiaqi Xue, Yi Sheng +3
Deep learning models have been incorporated into high-stakes sectors, including healthcare diagnosis, loan approvals, and candidate recruitment, among others. Consequently, any bia…
TrojFSP: Trojan Insertion in Few-shot Prompt Tuning
Mengxin Zheng, Jiaqi Xue, Xun Chen +3
Prompt tuning is one of the most effective solutions to adapting a fixed pre-trained language model (PLM) for various downstream tasks, especially with only a few input samples. Ho…
TrojLLM: A Black-box Trojan Prompt Attack on Large Language Models
Jiaqi Xue, Mengxin Zheng, Ting Hua +4
Large Language Models (LLMs) are progressively being utilized as machine learning services and interface tools for various applications. However, the security implications of LLMs,…
vFHE: Verifiable Fully Homomorphic Encryption with Blind Hash
Qian Lou, Muhammad Santriaji, Ardhi Wiratama Baskara Yudha +2
Fully homomorphic encryption (FHE) is a powerful encryption technique that allows for computation to be performed on ciphertext without the need for decryption. FHE will thus enabl…
SSL-Cleanse: Trojan Detection and Mitigation in Self-Supervised Learning
Mengxin Zheng, Jiaqi Xue, Zihao Wang +4
Self-supervised learning (SSL) is a prevalent approach for encoding data representations. Using a pre-trained SSL image encoder and subsequently training a downstream classifier, i…
ESTAS: Effective and Stable Trojan Attacks in Self-supervised Encoders with One Target Unlabelled Sample
Jiaqi Xue, Qian Lou
Emerging self-supervised learning (SSL) has become a popular image representation encoding method to obviate the reliance on labeled data and learn rich representations from large-…
Audit and Improve Robustness of Private Neural Networks on Encrypted Data
Jiaqi Xue, Lei Xu, Lin Chen +3
Performing neural network inference on encrypted data without decryption is one popular method to enable privacy-preserving neural networks (PNet) as a service. Compared with regul…