Quantitative Sensitivity Analyses of C2PA-Signed Video Signals to Unauthorized Manipulations

In recent years, AI-powered audio and video (AV) processing tools have become so advanced that it is increasingly difficult to distinguish original content from that generated by artificial intelligence. This has resulted in extraordinary challenges regarding media trust. AI-based AV processing can alter the truth, whether through unintentional misinformation or the deliberate deception of viewers. Consequently, there is a growing need for standards to regulate the provenance and authenticity of digital media from both the public and private sectors. To establish the necessary guidelines for these standards, the Coalition for Content Provenance and Authenticity (C2PA) organized a symposium that brought together policymakers, media industry leaders, and end-user organizations to discuss methods for protecting media content in the era of fake news and disinformation. This paper measures the effectiveness of AV material authentication methods. Subsequent processing of the AV material was performed in a Non-Linear Editing (NLE) system to make specific adjustments. Measurements were conducted to verify if, and to what extent, these changes caused the loss of the efficiency certificate (Content Credentials).