Skip to content

@HKU-TASR HKU Trustworthy AI and Systems Research (TASR) Lab

Change the repository type filter

All

    Repositories list

    4 repositories

    • Sanitizer

      Public
      Sanitizer is a server-side method that ensures client-embedded backdoors can only be used for contribution demonstration in federated learning but not be triggered on natural queries in harmful ways.
      watermarkfederated-learningharmfulbackdoor-attacks
      Python
      •
      MIT License
      •0•2•0•0•Updated Dec 8, 2024Dec 8, 2024

    • Geminio

      Public
      Geminio is a VLM-powered gradient inversion attack in federated learning (FL). It allows the adversary (the FL server) to describe the data of value and reconstruct the victim client's private data matching the description.
      vlmfederated-learningai-privacygradient-inversion
      Python
      •
      MIT License
      •0•5•0•0•Updated Dec 6, 2024Dec 6, 2024

    • AnywhereDoor

      Public
      AnywhereDoor is a multi-target backdoor attack tailored for object detection. Once implanted, it enables adversaries to specify different attack types (object vanishing, fabrication, or misclassification) and configurations (untargeted or targeted with specific classes) to dynamically control detection behavior.
      object-detectionbackdoor-attacks
      Jupyter Notebook
      •
      MIT License
      •0•2•0•0•Updated Dec 6, 2024Dec 6, 2024

    • Imperio

      Public
      [IJCAI 2024] Imperio is an LLM-powered backdoor attack. It allows the adversary to issue language-guided instructions to control the victim model's prediction for arbitrary targets.
      ai-securitybackdoor-attacksllm
      Python
      •
      MIT License
      •4•42•1•0•Updated Apr 17, 2024Apr 17, 2024