Experimental Data for Persistent Billable State: Denial-of-Wallet Attacks and Defenses in Tool-Calling LLM Agents

This dataset contains experimental results supporting the study “Persistent Billable State: Denial-of-Wallet Attacks and Defenses in Tool-Calling LLM Agents.” It summarizes attack and benign scenarios in tool-calling LLM agents, including experiment counts, Cost Amplification Factor (CAF), agent turns, and estimated costs. The data were generated through controlled experiments and organized in CSV format for analysis and verification.

No personal information, credentials, or sensitive provider data are included.