16 December 2024 in IEEE Transactions on Information Forensics and Security
Well-trained deep learning (DL) models are widely recognized as valuable intellectual property (IP) and have been extensively adopted. However, concerns regarding IP infringement emerge when these models are either privately sold to end-users or publicly released online. Unauthorized activities, such as redistributing privately purchased models or exploiting restricted open-source models for commercial gain, pose a significant threat to the interests of model owners. In this paper, we introduce DEEPREG, a trustworthy and privacy-friendly regulatory framework designed to address IP infringement within the realm of DL models, thereby nurturing a healthier development ecosystem. DEEPREG enables a designated third-party regulator to extract the fingerprint of the original model within a Trusted Execution Environment, as well as to verify suspect models utilizing solely the predicted label without probability. Specifically, we leverage the uniqueness of feature extractors in DL models to craft multiple synthetic inputs for a selected real input. The real input, along with its synthetic inputs, establishes a one-to-many relationship, thereby creating a unique fingerprint for the original model. Furthermore, we propose two distinct methods for suspect detection and piracy judgment. These methods analyze the responses from the model API upon feeding the fingerprint, ensuring a high level of confidence while preventing malicious accusations. Experimental results demonstrate that DEEPREG achieves 100% detection accuracy for pirated models, with zero false positives for irrelevant models.
12 January 2026 in IEEE Transactions on Pattern Analysis and Machine Intelligence
As model modification techniques are increasingly employed to obtain high-performing machine learning models at reduced costs, identifying lineage relationships—i.e., whether one model is derived from another—has garnered significant research interest. However, existing approaches are largely empirical, lack theoretical grounding, and are often ineffective against high-impact modifications. Furthermore, none have addressed the measurement of lineage closeness, which quantifies the degree of modification between models. In this paper, we reformulate the model lineage determination problem as a question of whether two models’ parameters reside within the same local optimum of the loss landscape. Based on this formulation, we propose a simple yet effective method for lineage determination. We further explore the impact of various modification techniques on models’ decision boundaries using visualization techniques, and observe that changes in decision boundaries serve as an accurate metric for lineage closeness. Leveraging this insight, we propose a task-agnostic and modification-type-agnostic method to measure lineage closeness by computing the mean adversarial distance from data points to decision boundaries and the matching rate of data point predictions. To reduce computational overhead, we design an efficient sampling strategy for data point selection. Extensive experiments demonstrate that our approach achieves 100% accuracy in model lineage determination and provides precise, quantitative measurements of lineage closeness across a wide range of modification scenarios.