InFlow: Robust outlier detection utilizing Normalizing Flows

TL;DR

This paper introduces InFlow, an enhanced normalizing flow model with attention mechanisms, that effectively detects outliers and adversarial inputs without needing outlier training data, achieving state-of-the-art results.

Contribution

The paper proposes extending normalizing flows with attention mechanisms to improve outlier detection, addressing their overconfidence issue without requiring outlier data for training.

Findings

Achieves state-of-the-art OOD detection performance

Effectively detects adversarial attacks

Does not require outlier data for training

Abstract

Normalizing flows are prominent deep generative models that provide tractable probability distributions and efficient density estimation. However, they are well known to fail while detecting Out-of-Distribution (OOD) inputs as they directly encode the local features of the input representations in their latent space. In this paper, we solve this overconfidence issue of normalizing flows by demonstrating that flows, if extended by an attention mechanism, can reliably detect outliers including adversarial attacks. Our approach does not require outlier data for training and we showcase the efficiency of our method for OOD detection by reporting state-of-the-art performance in diverse experimental settings. Code available at https://github.com/ComputationalRadiationPhysics/InFlow .