BadBone: Backdoor Attacks Against Backbone Models in Visual Prompt Learning

Summary

BadBone is a backdoor attack against prompt learning that uses bi-level optimization to compromise a backbone model, so only downstream tasks employing prompt learning inherit the vulnerability. Experiments on three models and three datasets show the targeted and untargeted backdoored models reach high attack performance while keeping utility on pre-training and downstream tasks. The authors report that six model-level defenses, including Neural Cleanse, ABS, MNTD, NAD, CLP and D-BR, are largely ineffective against these models.

Topics

• Adversarial machine learning

Related items

• MediumFake OpenAI, SpaceX events used to steal Microsoft, Google accountsSimilar attack · BleepingComputer
• HighGHSA-64vx-6h2c-rjh7: Banks: Unsafe importlib.import_module of attacker-controlled Tool.import_path in CompletionExtension allows RCESimilar attack · GitHub Advisory Database
• MediumHacking on Autopilot: Inside ARTEX Cybercrime CampaignsSimilar attack · Check Point Research
• MediumCVE-2026-122160: myshell-ai AIlice HuggingFace Model Wrapper untrusted model loadingSimilar attack · NVD/CVE Database
• MediumCVE-2026-122159: Edge0-AI Audio8_TTS path traversal via slow_decode_models argumentSimilar attack · NVD/CVE Database