Authorize-on-Demand: Dynamic Authorization with Legality-Aware Intellectual Property Protection for VLMs

Researchers propose AoD-IP, a new framework for protecting intellectual property in vision-language models through dynamic authorization and legality-aware assessment. The system allows flexible, user-controlled authorization that can adapt to changing deployment scenarios while preventing unauthorized use of valuable AI models.

• →AoD-IP introduces dynamic authorization for VLMs that can be controlled on-demand rather than static training-time definitions.
• →The framework includes a dual-path inference mechanism that predicts both input legality and task-specific outputs.
• →The system enables proactive IP protection by confining model deployment to authorized domains and preventing unauthorized transfers.
• →Experimental results show the framework maintains strong performance in authorized domains while reliably detecting unauthorized use.
• →The approach provides greater flexibility and extensibility compared to existing static-domain IP protection methods.