Double-Adversarial Activation Anomaly Detection: Adversarial Autoencoders are Anomaly Generators

Abstract

Anomaly detection is a challenging task for machine learning algorithms due to the inherent class imbalance. It is costly and time-demanding to manually analyse the observed data, thus usually only few known anomalies if any are available. Inspired by generative models and the analysis of the hidden activations of neural networks, we introduce a novel unsupervised anomaly detection method called DA3D. Here, we use adversarial autoencoders to generate anomalous counterexamples based on the normal data only. These artificial anomalies used during training allow the detection of real, yet unseen anomalies. With our novel generative approach, we transform the unsupervised task of anomaly detection to a supervised one, which is more tractable by machine learning and especially deep learning methods. DA3D surpasses the performance of state-of-the-art anomaly detection methods in a purely data-driven way, where no domain knowledge is required.

Accepted at IJCNN 2022

References in corpus (9)

• Fashion-MNIST: a Novel Image Dataset for Benchmarking Machine Learning Algorithms
• Deep Learning for Anomaly Detection: A Review
• A Unifying Review of Deep and Shallow Anomaly Detection
• A Comprehensive Survey on Graph Anomaly Detection with Deep Learning
• MAD-GAN: Multivariate Anomaly Detection for Time Series Data with Generative Adversarial Networks
• Anomaly Detection with Adversarial Dual Autoencoders
• Unsupervised Anomaly Detection with Adversarial Mirrored AutoEncoders
• ${\text{A}}^3$: Activation Anomaly Analysis
• The Complexity of Adversarially Robust Proper Learning of Halfspaces with Agnostic Noise

Cited by in corpus (2)

• Semisupervised Anomaly Detection using Support Vector Regression with Quantum Kernel
• QUACK: Quantum Aligned Centroid Kernel