Protocol Agent: What If Agents Could Use Cryptography In Everyday Life?

TL;DR

This paper explores enabling agents to dynamically create and execute cryptographic protocols for everyday tasks, improving communication efficiency and privacy through a new benchmark and model fine-tuning.

Contribution

It introduces a comprehensive benchmark for agents to recognize, negotiate, implement, compute, and secure cryptographic protocols, along with a dataset-generation approach and evaluation of fine-tuned models.

Findings

Fine-tuned models outperform base models significantly.

The benchmark effectively measures protocol recognition, negotiation, correctness, computation, and security.

Proposed dataset-generation enhances model capabilities.

Abstract

We often assume that agent-to-agent interaction will mirror human conversation. However, agents operate fundamentally differently. What if they could develop communication patterns that are more efficient and better aligned with their capabilities? While cryptographic primitives that could profoundly improve everyday interactions already exist, humans can't use them because they are too complex and the math can't be done in one's head. Examples range from proving your age (or other attributes) without showing your ID, to filing an anonymous report within a group while proving you are a legitimate member, to splitting a dinner bill fairly without revealing salaries. What if agents could create protocols "on the fly" by recognizing which primitive fits an everyday situation, proposing it to an agentic counterpart, persuading them to participate, and then executing the protocol correctly using appropriate computation tools? Protocol Agent frames this problem by introducing a benchmark that spans: (1) cryptographic primitive recognition, (2) negotiation skills, (3) implementation correctness, (4) correct computation and (5) security strength. We evaluate current open-weight and state-of-the-art models on this benchmark, propose a dataset-generation approach to improve these capabilities, and measure the impact of supervised fine-tuning (SFT) on benchmark performance, with tuned models outperforming base models by a wide margin.