Unifying Evaluation of Machine Learning Safety Monitors

TL;DR

This paper proposes a unified framework with three safety-oriented metrics for evaluating machine learning safety monitors across diverse applications, ensuring consistent and system-aligned assessments.

Contribution

It introduces a formalized set of metrics and evaluation procedures that unify diverse existing methods for assessing ML safety monitors.

Findings

Metrics effectively compare different monitors across tasks

Evaluation choices significantly influence perceived monitor performance

Formal safety assumptions align evaluations with system requirements

Abstract

With the increasing use of Machine Learning (ML) in critical autonomous systems, runtime monitors have been developed to detect prediction errors and keep the system in a safe state during operations. Monitors have been proposed for different applications involving diverse perception tasks and ML models, and specific evaluation procedures and metrics are used for different contexts. This paper introduces three unified safety-oriented metrics, representing the safety benefits of the monitor (Safety Gain), the remaining safety gaps after using it (Residual Hazard), and its negative impact on the system's performance (Availability Cost). To compute these metrics, one requires to define two return functions, representing how a given ML prediction will impact expected future rewards and hazards. Three use-cases (classification, drone landing, and autonomous driving) are used to demonstrate how metrics from the literature can be expressed in terms of the proposed metrics. Experimental results on these examples show how different evaluation choices impact the perceived performance of a monitor. As our formalism requires us to formulate explicit safety assumptions, it allows us to ensure that the evaluation conducted matches the high-level system requirements.