Whose Agent Are You? Multi-Layer Fingerprinting and Attribution of Autonomous Web Agents

The paper presents a multi-layer fingerprinting framework for distinguishing AI web agents from human users and traditional crawlers, addressing privacy and security challenges posed by autonomous content scraping. By analyzing six frameworks (AutoGen, Browser Use, Claude, Gemini, Operator, and Skyvern), the authors identify structural differences in HTTP request assembly and browser interactions, achieving a 97% accuracy rate using a decision tree classifier. This approach offers a robust method for enforcing web security policies and protecting content against unauthorized scraping by leveraging cross-layer tracking techniques.