A Policy Driven AI-Assisted PoW Framework

TL;DR

This paper presents an AI-assisted PoW framework that adaptively adjusts puzzle difficulty based on traffic features to differentiate trustworthy from untrustworthy network connections, improving cyberdefense.

Contribution

It introduces a novel modular framework that uses AI to dynamically generate puzzles, enhancing PoW systems' ability to throttle malicious traffic.

Findings

Effectively throttles untrustworthy traffic

Adjusts puzzle difficulty based on traffic features

Improves differentiation between legitimate and malicious requests

Abstract

Proof of Work (PoW) based cyberdefense systems require incoming network requests to expend effort solving an arbitrary mathematical puzzle. Current state of the art is unable to differentiate between trustworthy and untrustworthy connections, requiring all to solve complex puzzles. In this paper, we introduce an Artificial Intelligence (AI)-assisted PoW framework that utilizes IP traffic based features to inform an adaptive issuer which can then generate puzzles with varying hardness. The modular framework uses these capabilities to ensure that untrustworthy clients solve harder puzzles thereby incurring longer latency than authentic requests to receive a response from the server. Our preliminary findings reveal our approach effectively throttles untrustworthy traffic.