Toward Supervised Anomaly Detection

By Nico Görnitz, Marius Kloft, Konrad Rieck, Ulf Brefeld. Journal of Artificial Intelligence Research, 2014

418 citations recorded by the source corpus.

Anomaly detection is being regarded as an unsupervised learning task as anomalies stem from adversarial or unlikely events with unknown distributions. However, the predictive performance of purely unsupervised anomaly detection often fails to match the required detection rates in many tasks and there exists a need for labeled data to guide the model generation. Our first contribution shows that classical semi-supervised approaches, originating from a supervised classifier, are inappropriate and hardly detect new and unknown anomalies. We argue that semi-supervised anomaly detection needs to ground on the unsupervised learning paradigm and devise a novel algorithm that meets this requirement. Although being intrinsically non-convex, we further show that the optimization problem has a convex equivalent under relatively mild assumptions. Additionally, we propose an active learning strategy to automatically filter candidates for labeling. In an empirical study on network intrusion detection data, we observe that the proposed learning methodology requires much less labeled data than the state-of-the-art, while achieving higher detection accuracies.

Subfields: Artificial Intelligence, Computer Networks and Communications

View Paper: https://www.semanticscholar.org/paper/f5a951b9596be0df5ad7ede180b405c9e97a65c9

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

Citations in the corpus

SPot-the-Difference Self-Supervised Pre-training for Anomaly Detection and Segmentation (2022, methodology): “In such cases, one can improve over 1-class methods with a standard 2-class model [12,18,21,27] by incorporating the anomalous data in training, which is in fact a well-established practice in commercial visual inspection AI services [1,2].”
Progress in Outlier Detection Techniques: A Survey (2019, background): “[200] proposed another work where an active learning strategy is applied for anomaly detection.”
Generalized Out-of-Distribution Detection: A Survey (2021, neutral): “Fully supervised AD is studied in [335].”
Unsupervised real-time anomaly detection for streaming data (2017, background): “Techniques that require data labels, such as supervised classification-based methods [42], are typically unsuitable for real-time anomaly detection and continuous learning.”
SSD: A Unified Framework for Self-Supervised Outlier Detection (2021, background): “What if access to a fraction of OOD data or training data labels is available? How do we move past a detector based on unlabeled data and design a framework which can take advantage of such information? Though access to outliers during training is a strong assumption, it may be feasible to obtain a ”
Confident Learning: Estimating Uncertainty in Dataset Labels (2019, methodology): “First, Cỹ,y∗ is useful for supervised anomaly detection (Görnitz et al., 2013) – if an example has low (near-uniform) probabilities across classes, it is not counted so that Cỹ,y∗ is robust to examples from an alien class not in the dataset.”
A Deep Neural Network for Unsupervised Anomaly Detection and Diagnosis in Multivariate Time Series Data (2018, background): “To build a system which can automatically detect and diagnose anomalies, one main problem is that few or even no anomaly label is available in the historical data, which makes the supervised algorithms (G¨ornitz et al. 2013) infea-a sible.”
A Unifying Review of Deep and Shallow Anomaly Detection (2020, methodology): “The adoption of semisupervised [144], [229], [419] or multisphere [145], [155], [316] model extensions give successful examples for such a transfer.”
Deep Learning for Medical Anomaly Detection – A Survey (2020, result): “accurate medical anomaly detection framework measures should be taken to mitigate such hindrances. Considering these challenges, medical anomaly detection is often posed as a supervised learning task [8], [1], where a supervision signal is presented for the model to learn to discriminate normal from”
Deep Semi-Supervised Anomaly Detection (2019, methodology): “A few works (Wang et al., 2005; Liu & Zheng, 2006; Görnitz et al., 2013) have investigated the general semi-supervised AD setting where one also utilizes labeled anomalies, however existing deep approaches are domain or data-type specific (Ergen et al.”
Unsupervised Anomaly Detection With LSTM Neural Networks (2017, methodology): “Remark 9: In the semisupervised framework, we have the following optimization problem for our SVDD-based algorithms [35]:”