Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

On the Vulnerability of Parameter-Level Defenses to Model Merging

About

The training-free integration of expert models via model merging has exposed significant security risks, enabling free-riders to combine specialized models without authorization. Recent works propose parameter-level defenses that employ linear parameter transformations to neutralize this threat. In this paper, we systematically analyze such defenses and reveal that their protected task vectors are inherently small in magnitude. Consequently, the protected weights remain overwhelmingly dominated by the pretrained model. Based on this observation, we designate the pretrained model as a static reference anchor and propose the Anchor-Guided Attack (AGA) to circumvent existing safeguards. Specifically, AGA aligns the protected model with this anchor to recover the transformation matrix analytically. Extensive evaluations validate that AGA consistently bypasses both individual and composite defenses under realistic defense-agnostic scenarios. Furthermore, we provide Anchor-Repulsive Fine-tuning (ARF), a defense method to mitigate the anchor dominance leveraged by AGA. Empirical results confirm that ARF effectively defeats the proposed attack. Our code is available at https://github.com/krumpguo/secure-merge-attack.

Kuangpu Guo, Qingyan Zheng, Jian Liang, Yongcan Yu, Zilei Wang, Ran He, Tieniu Tan• 2026

Related benchmarks

TaskDatasetResultRank
Image ClassificationStanford Cars
Accuracy67.19
705
ClassificationCars
Accuracy82.02
571
Image ClassificationRESISC45
Accuracy85.96
539
Image ClassificationVision Multi-task Suite (SUN397, Cars, RESISC45, EuroSAT, SVHN, GTSRB, MNIST, DTD)
Average Accuracy78.48
171
Image ClassificationSUN397, Cars, RESISC45, EuroSAT, SVHN, GTSRB, MNIST, DTD (test)
Avg Acc82.72
134
Image ClassificationGTSRB
Accuracy67.02
97
Visual Classification8 Vision Tasks (SUN397, Stanford Cars, RESISC45, EuroSAT, SVHN, GTSRB, MNIST, DTD)
Average Accuracy37.18
90
Image ClassificationSVHN
Accuracy81.15
54
Image ClassificationSUN397
Accuracy73.36
40
Image ClassificationDTD
Top-1 Accuracy46.44
26
Showing 10 of 17 rows

Other info

Follow for update