BAPLe++: Cross-Model Transferability and Prompt-Level Defense of Backdoors in Medical Foundation Models

Summary

BAPLe++ studies whether backdoors implanted through prompt learning in medical foundation models (Med-FMs) transfer to other Med-FMs, measured with Clean Accuracy Difference (CAD) and Backdoor Accuracy Difference (BAD). The authors also propose DePLe, which uses the infected prompt as a fixed anchor and learns a new prompt on a small clean set to reduce backdoor accuracy. Across four Med-FMs and six chest X-ray and histopathology datasets, backdoors transferred asymmetrically and in a dataset-dependent manner, and DePLe reduced backdoor accuracy more consistently than four training-free input transformations.

Mitigation

DePLe: learn a new prompt on a small clean set, using the infected prompt as a fixed anchor, so the model keeps its clean task while diverging from the compromised embedding region, without retraining the backbone.

Related items

• HighCVE-2026-123818: LMDeploy is a toolkit for compressing, deploying, and serving large language models. From 0.12.3 until 0.15.0…Similar attack · NVD/CVE Database
• HighGHSA-vvr2-h2jp-838m: Code Graph RAG: MCP paginated read_file path traversal / arbitrary file readSimilar attack · GitHub Advisory Database
• InfoSuppression Is Not Erasure: Mechanistic Analysis and Reactivation Boundaries of Subspace Projection for LoRA Backdoor DefenseSimilar attack · Arxiv (cs.CR + cs.CL + cs.LG)
• MediumAttackers Crash Claude's 'Office Party' Flaw to Run Malicious Code on LinuxSimilar attack · Dark Reading
• InfoSay Shibboleth: Fingerprinting the LLMs and Harnesses Behind Agentic AttacksSimilar attack · Arxiv (cs.CR + cs.CL + cs.LG)