SynGhost: Invisible and Universal Task-agnostic Backdoor Attack via Syntactic Transfer
cs.CR, cs.AI, cs.CL
Submitted: 2024-02-29
Updated: 2026-09-14
Comments: 17 pages, 16 figures, 12 tables, accepted at NAACL 2025 Findings
Code: https://github.com/Zhou-CyberSecurity-AI/SynGhost
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Terminology
Sources
- GPT-4 Technical Report
- Backdoor Attacks and Countermeasures in Natural Language Processing Models: A Comprehensive Security Review
- BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
- The Philosopher's Stone: Trojaning Plugins of Large Language Models
- UOR: Universal Backdoor Attacks on Pre-trained Language Models
- DeBERTa: Decoding-enhanced BERT with Disentangled Attention
- ALBERT: A Lite BERT for Self-supervised Learning of Language Representations
- ChatGPT as an Attack Tool: Stealthy Textual Backdoor Attack via Blackbox Generative Model Trigger
- RoBERTa: A Robustly Optimized BERT Pretraining Approach
- Backdoor Attacks on Dense Retrieval via Public and Unintentional Triggers
- UMAP: Uniform Manifold Approximation and Projection for Dimension Reduction
- Hidden Killer: Invisible Textual Backdoor Attacks with Syntactic Trigger
- Turn the Combination Lock: Learnable Textual Backdoor Attacks via Word Substitution
- Learning to Poison Large Language Models for Downstream Manipulation
- Frequency Principle: Fourier Analysis Sheds Light on Deep Neural Networks
- Universal Vulnerabilities in Large Language Models: Backdoor Attacks for In-context Learning
- Be Careful about Poisoned Word Embeddings: Exploring the Vulnerability of the Embedding Layers in NLP Models
Related papers
- SoK: AI-Augmented Binary Reversing
- Relaxed Sender Anonymity for CBDC Interbank Settlement: A Zero-Knowledge Approach on Permissioned EVM
- Calibration-Family Overfit: Why Trusted Sabotage Monitors Don't Transfer Across Lineages
- Efficient Fuzzy PSI under One-Sided Assumptions
- Sealing the Audit-Runtime Gap for LLM Skills
- Token Composition: A Graph Based on EVM Logs