SwitchNet: protecting neural networks by structure obfuscation and switch-controlled inference
Authors:
- Yuling CaiYCYuling Cai
- Guozhu MengGMGuozhu Meng
- Yinzhi CaoYCYinzhi Cao
- Guangdong BaiGBGuangdong Bai
Open Access
- 27 April 2026
- journal article
- Published by Springer Nature in Cybersecurity
- Vol. 9 (1)
Abstract
Training deep learning models requires substantial financial and human resources, so once deployed in untrusted environments, these models immediately attract the attention of attackers who seek to steal and misuse them. Traditional model protection methods are ineffective in addressing model accuracy, performance, and proactive defense. To this end, we present an active defensive approach SwitchNet by obfuscating model structure and proposing a switch-controlled mechanism to manage model inference. Specifically, SwitchNet learns the weight distribution of the original model and then constructs confusion layers that are strategically inserted into the original model for structure obfuscation. Each of the model layers is equipped with a switch , which is controlled by a switching policy network. We train this policy network with an adaptive pattern as a “secret key” that can accurately control the switch states, and thereby the model inference process. We conduct a comprehensive theoretical analysis of the perturbation boundary and certify that SwitchNet maintains high robustness under $$\ell _\infty$$ perturbations, with certified accuracy exceeding 80% at $$\epsilon = 0.0048$$ (CROWN). In addition, we perform extensive experiments on both classical convolutional networks and Vision Transformers. The results show that SwitchNet effectively preserves model accuracy for legitimate users (with only a 0.35% drop), while reducing the accuracy for unauthorized users to near-random guessing. Compared to the state-of-the-art, our approach reduces inference and construction overhead by 20.89% and 12.08%, respectively. Furthermore, SwitchNet proves to be stealthy and resilient against various attacks aimed at detecting or compromising the protection mechanism.
Keywords
This publication has 34 references indexed in Scilit:
- Re-Thinking Model Inversion Attacks Against Deep Neural Networks2023
- Planning-oriented Autonomous Driving2023
- MedAttacker: Exploring Black-Box Adversarial Attacks on Risk Prediction Models in Healthcare2022
- LightFace: A Hybrid Deep Face Recognition Framework2020
- Privacy-Preserving Deep Learning on Machine Learning as a Service—a Comprehensive SurveyIEEE Access, 2020
- Hardware-Assisted Intellectual Property Protection of Deep Learning Models2020
- User authentication on mobile devices: Approaches, threats and trendsComputer Networks, 2020
- Deep learning-based intelligent face recognition in IoT-cloud environmentComputer Communications, 2020
- Knockoff Nets: Stealing Functionality of Black-Box Models2019
- Boosting Adversarial Attacks with Momentum2018