Deep Anomaly Detection with Deviation Networks

By Guansong Pang, Chunhua Shen, Anton van den Hengel. Knowledge Discovery and Data Mining, 2019

468 citations recorded by the source corpus.

Although deep learning has been applied to successfully address many data mining problems, relatively limited work has been done on deep learning for anomaly detection. Existing deep anomaly detection methods, which focus on learning new feature representations to enable downstream anomaly detection methods, perform indirect optimization of anomaly scores, leading to data-inefficient learning and suboptimal anomaly scoring. Also, they are typically designed as unsupervised learning due to the lack of large-scale labeled anomaly data. As a result, they are difficult to leverage prior knowledge (e.g., a few labeled anomalies) when such information is available as in many real-world anomaly detection applications. This paper introduces a novel anomaly detection framework and its instantiation to address these problems. Instead of representation learning, our method fulfills an end-to-end learning of anomaly scores by a neural deviation learning, in which we leverage a few (e.g., multiple to dozens) labeled anomalies and a prior probability to enforce statistically significant deviations of the anomaly scores of anomalies from that of normal data objects in the upper tail. Extensive results show that our method can be trained substantially more data-efficiently and achieves significantly better anomaly scoring than state-of-the-art competing methods.

Subfields: Artificial Intelligence, Computer Networks and Communications, Signal Processing

View Paper: https://www.semanticscholar.org/paper/f24dc0f3c1cc1da6c4fd12e1a65f55a6d82bb8ed

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Networks and Communications

In Signal Processing

Citations in the corpus

A Comprehensive Survey on Graph Anomaly Detection With Deep Learning (2021, background): “com/PanpanZheng/OCAN DevNet [131] Python Tensorflow Static Graph https://github.”
Deep Learning for Anomaly Detection (2020, methodology): “00% Tabular [114, 115] Dropout prediction KDD Cup 15 35,091 27 0.”
HRN: A Holistic Approach to One Class Learning (2020, background): “Some even use anomalies in the training data to build multi-class classiﬁers [76, 36, 62].”
Anomaly Detection on Attributed Networks via Contrastive Self-Supervised Learning (2021, unknown): “With the rocketing growth of deep learning for anomaly detection [12], [19], [20], [21], researchers also present deep neural networks-based methods to solve the anomaly detection problem on attributed networks.”
COVID-19 Screening on Chest X-ray Images Using Deep Learning based Anomaly Detection (2020, methodology): “To address the data imbalance problem existing between COVID-19 and non-COVID-19 cases, we propose a X-raybased COVID-19 screening model, inspired by [11] , to promote the imbalance binary classification task through the anomaly detection task.”
ADBench: Anomaly Detection Benchmark (2022, methodology): “…in real-world applications, e.g., a few anomalies identiﬁed by domain experts or human-in-the-loop techniques like active learning [5, 7, 78, 189] Notably, there is a group of semi-supervised AD algorithms [127, 128, 130, 131, 152, 168, 205] that have not been covered by existing benchmarks.”
Deep Learning for Time Series Anomaly Detection: A Survey (2022, unknown): “Various GNN architectures exist, such as Graph Convolution Networks (GCN) [103], which aggregate one-step neighbours, and Graph Attention Networks (GAT) [173], which use attention functions to compute different weights for each neighbour.”
A Unifying Review of Deep and Shallow Anomaly Detection (2020, methodology): “performance improvements on complex data [60], [144], [332], [336], [337], [538].”
Deep Isolation Forest for Anomaly Detection (2022, methodology): “Recent studies [10], [11], [12], [13] devise deep anomaly detection models based on the representation learning, while some other meth-ods use concepts like reconstruction [14] or deviation [15] in the deep learning framework as scoring function.”