Deep Weakly-supervised Anomaly Detection (2019-10-30T00:00:00.000000Z)

TL;DR

Empirical results on 12 real-world datasets show that PReNet significantly outperforms nine competing methods in detecting seen and unseen anomalies, and theoretically and empirically justify the robustness of the model w.r.t. anomaly contamination in the unlabeled data.

Abstract

Recent semi-supervised anomaly detection methods that are trained using small labeled anomaly examples and large unlabeled data (mostly normal data) have shown largely improved performance over unsupervised methods. However, these methods often focus on fitting abnormalities illustrated by the given anomaly examples only (i.e., seen anomalies), and consequently they fail to generalize to those that are not, i.e., new types/classes of anomaly unseen during training. To detect both seen and unseen anomalies, we introduce a novel deep weakly-supervised approach, namely Pairwise Relation prediction Network (PReNet), that learns pairwise relation features and anomaly scores by predicting the relation of any two randomly sampled training instances, in which the pairwise relation can be anomaly-anomaly, anomaly-unlabeled, or unlabeled-unlabeled. Since unlabeled instances are mostly normal, the relation prediction enforces a joint learning of anomaly-anomaly, anomaly-normal, and normal-normal pairwise discriminative patterns, respectively. PReNet can then detect any seen/unseen abnormalities that fit the learned pairwise abnormal patterns, or deviate from the normal patterns. Further, this pairwise approach also seamlessly and significantly augments the training anomaly data. Empirical results on 12 real-world datasets show that PReNet significantly outperforms nine competing methods in detecting seen and unseen anomalies. We also theoretically and empirically justify the robustness of our model w.r.t. anomaly contamination in the unlabeled data. The code is available at https://github.com/mala-lab/PReNet.

Authors

Guansong Pang

5 papers

Chunhua Shen

21 papers

A. van den Hengel

2 papers

TL;DR

Abstract

Authors

References72 items

Weakly Supervised Anomaly Detection: A Survey

Estimating the Contamination Factor's Distribution in Unsupervised Anomaly Detection

Transferring the Contamination Factor between Anomaly Detection Domains by Shape Similarity

Weakly Supervised Anomaly Detection Based on Two-Step Cyclic Iterative PU Learning Strategy

Catching Both Gray and Black Swans: Open-set Supervised Anomaly Detection*

Explainable Deep Few-shot Anomaly Detection with Deviation Networks

Towards Total Recall in Industrial Anomaly Detection

Deep Clustering based Fair Outlier Detection

Feature Encoding With Autoencoders for Weakly Supervised Anomaly Detection

Few-shot Network Anomaly Detection via Cross-network Meta-learning

Deep One-Class Classification via Interpolated Gaussian Descriptor

Learning and Evaluating Representations for Deep One-class Classification

PANDA: Adapting Pretrained Features for Anomaly Detection and Segmentation

Toward Deep Supervised Anomaly Detection: Reinforcement Learning from Partially Labeled Anomaly Data

Deep Learning for Anomaly Detection

Explainable Deep One-Class Classification

Class Prior Estimation in Active Positive and Unlabeled Learning

Comparison of novelty detection methods for multispectral images in rover-based planetary exploration missions

PUMAD: PU Metric learning for anomaly detection

Learning Memory-Guided Normality for Anomaly Detection

Training

Self-Trained Deep Ordinal Regression for End-to-End Video Anomaly Detection

DROCC: Deep Robust One-Class Classification

Unsupervised Representation Learning by Predicting Random Distances

Deep Anomaly Detection with Deviation Networks

Deep Semi-Supervised Anomaly Detection

Generalizing from a Few Examples

Memorizing Normality to Detect Anomaly: Memory-Augmented Deep Autoencoder for Unsupervised Anomaly Detection

MAD-GAN: Multivariate Anomaly Detection for Time Series Data with Generative Adversarial Networks

A Deep Neural Network for Unsupervised Anomaly Detection and Diagnosis in Multivariate Time Series Data

Learning from positive and unlabeled data: a survey

Adversarially Learned Anomaly Detection

Semi-Supervised Anomaly Detection with an Application to Water Analytics

PU Learning in Payload-based Web Anomaly Detection

Deep One-Class Classification

XGBOD: Improving Supervised Outlier Detection with Unsupervised Representation Learning

Learning Representations of Ultrahigh-dimensional Data for Random Distance-based Outlier Detection

Anomaly Detection with Partially Observed Anomalies

One-Class Adversarial Nets for Fraud Detection

Deep Autoencoding Gaussian Mixture Model for Unsupervised Anomaly Detection

Learning to Compare: Relation Network for Few-Shot Learning

A Semisupervised Approach to the Detection and Characterization of Outliers in Categorical Data

Unsupervised Anomaly Detection with Generative Adversarial Networks to Guide Marker Discovery

Prototypical Networks for Few-shot Learning

Efficient Training for Positive Unlabeled Learning

Matching Networks for One Shot Learning

UNSW-NB15: a comprehensive data set for network intrusion detection systems (UNSW-NB15 network data set)

Guilt by association: large scale malware detection by mining file-relation graphs

Robust linear regression: A review and comparison

Toward Supervised Anomaly Detection

Subsampling for efficient and effective unsupervised outlier detection ensembles

Outlier Analysis

FRaC: a feature-modeling approach for semi-supervised and unsupervised anomaly detection

Isolation-Based Anomaly Detection

Outlier Detection Using Nonconvex Penalized Regression

Signaling Potential Adverse Drug Reactions from Administrative Health Databases

Anomaly detection: A survey

SNARE: a link analytic system for graph labeling and risk detection

A New Local Distance-Based Outlier Detection Approach for Scattered Real-World Data

Learning classifiers from only positive and unlabeled data

Mining Unexpected Temporal Associations: Applications in Detecting Adverse Drug Reactions

Wilcoxon Signed Rank Test

One-shot learning of object categories

Learning to Classify Texts Using Positive and Unlabeled Data

Deep anomaly detection with self-supervised learning and adversarial training

Unseen Anomaly Detection on Networks via Multi-Hypersphere Learning

Learning Prototype Representations Across Few-Shot Tasks for Event Detection

Longbing Cao, and Anton van den Hengel. 2021. Deep Learning for Anomaly Detection: A Review

Activation anomaly analysis

Outlier Detection with Autoencoder Ensembles

k-Zero Day Safety: A Network Security Metric for Measuring the Risk of Unknown Vulnerabilities

Area under the Precision-Recall Curve: Point Estimates and Confidence Intervals

Field of Study

Journal Information

Name

Venue Information