# Device-Bind Key-Storageless Hardware AI Model IP Protection: A PUF and Permute-Diffusion Encryption-Enabled Approach

arXiv: https://arxiv.org/abs/2212.11133 · HTML: https://mlsanity.com/paper/2212.11133/

## Key facts

- Cited by 3 papers in the map (data through 14 Oct 2026).
- Placed in Watermarking generated content and models, one of 12 directions in LLM safety and security.
- Builds on 9 of its 32 references that are in the map.
- First posted on arXiv 21 Dec 2022.

## Placed in

- LLM safety and security / [Watermarking generated content and models](https://mlsanity.com/safety/watermarking-generated-content-and-models/)

## Builds on (most cited first, 9 of 9)

- [Fine-Pruning: Defending Against Backdooring Attacks on Deep Neural Networks](https://mlsanity.com/paper/1805.12185/) (2018)
- [Turning Your Weakness Into a Strength: Watermarking Deep Neural Networks by Backdooring](https://mlsanity.com/paper/1802.04633/) (2018)
- [Embedding Watermarks into Deep Neural Networks](https://mlsanity.com/paper/1701.04082/) (2017)
- [How to Prove Your Model Belongs to You: A Blind-Watermark based Framework to Protect Intellectual Property of DNN](https://mlsanity.com/paper/1903.01743/) (2019)
- [Encryption-then-Compression Systems using Grayscale-based Image Encryption for JPEG Images](https://mlsanity.com/paper/1811.00236/) (2018)
- [Intellectual Property Protection for Deep Learning Models: Taxonomy, Methods, Attacks, and Evaluations](https://mlsanity.com/paper/2011.13564/) (2020)
- [VeriML: Enabling Integrity Assurances and Fair Payments for Machine Learning as a Service](https://mlsanity.com/paper/1909.06961/) (2019)
- [GNN4IP: Graph Neural Network for Hardware Intellectual Property Piracy Detection](https://mlsanity.com/paper/2107.09130/) (2021)
- [Fully Convolutional Network with Multi-Step Reinforcement Learning for Image Processing](https://mlsanity.com/paper/1811.04323/) (2018)

## Built on by (most cited first, 3 of 3)

- [Turn Passive to Active: A Survey on Active Intellectual Property Protection of Deep Learning Models](https://mlsanity.com/paper/2310.09822/) (2023)
- [Attack on a PUF-based Secure Binary Neural Network](https://mlsanity.com/paper/2510.24422/) (2025)
- [A PUF-Based Approach for Copy Protection of Intellectual Property in Neural Network Models](https://mlsanity.com/paper/2603.10753/) (2026)

## Abstract (arXiv)

Machine learning as a service (MLaaS) framework provides intelligent services or well-trained artificial intelligence (AI) models for local devices. However, in the process of model transmission and deployment, there are security issues, i.e. AI model leakage due to the unreliable transmission environments and illegal abuse at local devices without permission. Although existing works study the intellectual property (IP) protection of AI models, they mainly focus on the watermark-based and encryption-based methods and have the following problems: (i) The watermark-based methods only provide passive verification afterward rather than active protection. (ii) Encryption-based methods are low efficiency in computation and low security in key storage. (iii) The existing methods are not device-bind without the ability to avoid illegal abuse of AI models. To deal with these problems, we propose a device-bind and key-storageless hardware AI model IP protection mechanism. First, a physical unclonable function (PUF) and permute-diffusion encryption-based AI model protection framework is proposed, including the PUF-based secret key generation and the geometric-value transformation-based weights encryption. Second, we design a PUF-based key generation protocol, where delay-based Anderson PUF is adopted to generate the derive-bind secret key. Besides, convolutional coding and convolutional interleaving technologies are combined to improve the stability of PUF-based key generation and reconstruction. Third, a permute and diffusion-based intelligent model weights encryption/decryption method is proposed to achieve effective IP protection, where chaos theory is utilized to convert the PUF-based secret key to encryption/decryption keys. Finally, experimental evaluation demonstrates the effectiveness of the proposed intelligent model IP protection mechanism.