BackdoorSim: An Educational into Remote Administration Tools
LoVerst is a backdoor generator and backdoor generating tools.
Input-aware Dynamic Backdoor Attack (NeurIPS 2020)
[ICLR2025] Detecting Backdoor Samples in Contrastive Language Image Pretraining
Official implementation of the NeurIPS 24 paper "Mitigating Backdoor Attack by Injecting Proactive Defensive Backdoor"
[TIFS 2026] Trigger without Trace: Towards Stealthy Backdoor Attack on Text-to-Image Diffusion Models
A repository for documenting and exploring the backdoor attacks, featuring a curated collection of research groups, conferences, journals, competitions, papers, reports, and resource.
A benchmark for catching backdoors in neural networks — and finding out which detectors actually work. Implants known backdoors into a model zoo, runs published detectors against it blind, scores them all on identical metrics.
Dual-Embedding Guided Backdoor Attacks on CLIP (Based on BadCLIP, CVPR 2024)
Training-time data poisoning that silently suppresses web search in LLM security agents. Attack evades 5 detectors. 98% detection recall via behavioral monitoring.
[AAAI'26] FRBAT: Conditionally-Visible Physical Backdoor Attack via Fluorescence
A comprehensive framework simulating integrity (Backdoor) and privacy (Gradient Inversion) attacks in Federated Learning, featuring a Machine Unlearning defense strategy
This report details a critical supply chain vulnerability embedded in the system application `com.spreadtrum.sgps
Backdoor data-poisoning attack on a text classifier + a clustering defense (Python, scikit-learn).
A data-poisoning backdoor in an MNIST classifier: normal accuracy on clean input, attacker-chosen output on any input carrying the trigger
3DGS + StealthAttack on RTX 5070 Ti (Blackwell/sm_120): reproducible Docker env, ASR eval harness, verified ASR=pass poison, and poison-detection team scaffold (2026 Hansung competition, private).
Can a useful KV-cache Cartridge hide a working backdoor trigger, and can a defender catch it? Toy-scale model organism: joint training hides it, the trigger over-fires a lexical zone, and an activation monitor catches it in the act but never at rest. Qwen3-4B, single GPU.
Data-free backdoor detection in federated learning via reverse engineering (H100 experiments).
Independent Class-wise Coherence Score (ICS) for detecting adversarial data poisoning in Federated Learning — Generative AI course project
To associate your repository with the backdoor-attack topic, visit your repo's landing page and select "manage topics."