When Ambiguity Meets Atypicality: Dual-Perspective Test Input Prioritization for DNNs

2026-09-28

AI Summary

This study addresses the blind spots inherent in single-perspective fault detection during deep neural network (DNN) testing by proposing DuFP. Leveraging K-nearest neighbor density estimation for class-conditional density modeling, this method introduces a novel dual-perspective hybrid uncertainty metric that integrates inter-class decision ambiguity with intra-class distributional atypicality to guide test input prioritization. This framework effectively overcomes the limitations of single-perspective approaches, enabling more comprehensive fault prediction. Extensive multi-scenario experiments on both image and text datasets demonstrate that DuFP significantly outperforms existing state-of-the-art methods, efficiently prioritizing the discovery of fault-inducing inputs.

Abstract

While Deep Neural Networks (DNNs) have achieved remarkable progress in cutting-edge domains, their inherent brittleness has become a growing concern. To ensure the reliability and safety of DNN-enabled software, DNN testing has emerged as an indispensable practice. Within this context, test input prioritization is essential for early fault detection and reducing labeling costs. However, it remains challenging to accurately identify failure-inducing inputs. Although decision ambiguity and distributional atypicality are two widely adopted perspectives for characterizing inter-class competition and intra-class typicality respectively, relying on either perspective in isolation inevitably introduces blind spots. In this paper, we propose DuFP (Dual perspective Feature space Prioritization), a KNN density-based test input prioritization approach for DNNs that jointly incorporates both inter-class and intra-class perspectives. The prioritization framework of DuFP is built upon class-conditional density estimation. Based on the estimation results, prediction correctness is characterized by an ambiguity score and an atypicality score, with the former reflecting decision ambiguity and the latter quantifying distributional atypicality. A hybrid uncertainty score is then constructed by integrating both scores to guide the final prioritization. We evaluate DuFP on prioritization and selection tasks across image and text datasets under clean, corrupted, and adversarial scenarios. Experimental results demonstrate that DuFP effectively and efficiently prioritizes fault-inducing inputs and outperforms state-of-the-art approaches.

Problem

• Test Input Prioritization
• Deep Neural Networks
• Decision Ambiguity
• Distributional Atypicality
• Fault Detection

Innovation

• Test Input Prioritization
• Dual-Perspective
• Class-conditional Density Estimation
• Decision Ambiguity
• Distributional Atypicality