深度学习 异常检测 Deep Learning for Anomaly Detection: A Comprehensive Survey
This post summarizes a comprehensive survey paper on deep learning for anomaly detection — “Deep Learning for Anomaly Detection: A Review” [1], discussing challenges, methods and opportunities in this direction.
Anomaly detection, a.k.a. outlier detection, has been an active research area for several decades, due to its broad applications in a large number of key domains such as risk management, compliance, security, financial surveillance, health and medical risk, and AI safety. Although it is a problem widely studied in various communities including data mining, machine learning, computer vision and statistics, there are still some unique problem complexities and challenges that require advanced approaches. In recent years, deep learning enabled anomaly detection has emerged as a critical direction towards addressing these challenges. However, there is a lack of systematic review and discussion of the research progress in this direction. We aim to present a comprehensive review of this direction to discuss the main challenges, a large number of state-of-the-art methods, how they address the challenges, as well as future opportunities.
Largely Unsolved Challenges in Anomaly Detection
Although anomaly detection is a lasting active research area for years, there are still a number of largely unsolved challenges due to some unique and complex nature of anomalies, e.g., unknownness (they remain unknown until actually occur), heterogeneity (different anomalies demonstrate completely different abnormal characteristics), rareness (anomalies are rarely occurred data instances), diverse form of anomalies (point anomaly, contextual anomaly, and group anomaly).
One of the most challenging issues is the difficulty to achieve high anomaly detection recall rate (Challenge #1). Since anomalies are highly rare and heterogeneous, it is difficult to identify all of the anomalies. Many normal instances are wrongly reported as anomalies while true yet sophisticated anomalies are missed.
Anomaly detection in high-dimensional and/or not-independent data (Challenge #2) is also a significant challenge. Anomalies often exhibit evident abnormal characteristics in a low-dimensional space yet become hidden and unnoticeable in a high-dimensional space. High-dimensional anomaly detection has been a long-standing problem. Subspace/feature selection-based methods may be a straightforward solution. However, identifying intricate (e.g., high-order, nonlinear and heterogeneous) feature interactions and couplings may be essential in high-dimensional data yet remains a major challenge for anomaly detection.
Due to the difficulty and cost of collecting large-scale labeled anomaly data, it is important to have data-efficient learning of normality/abnormality (Challenge #3). wo major challenges are how to learn expressive normality/abnormality representations with a small amount of labeled anomaly data and how to learn detection models that are generalized to novel anomalies uncovered by the given labeled anomaly data.
Many weakly/semi-supervised anomaly detection methods assume the given labeled training data is clean, which can be highly vulnerable to noisy instances that are mistakenly labeled as an opposite class label. One main challenge here is how to develop noise-resilient anomaly detection (Challenge #4).
Most of existing methods are for point anomalies, which cannot be used for conditional anomaly and group anomaly since they exhibit completely different behaviors from point anomalies. One main challenge here is to incorporate the concept of conditional/group anomalies into anomaly measures/models for the detection of those complex anomalies (Challenge #5).
In many critical domains there may be some major risks if anomaly detection models are directly used as black-box models. For example, the rare data instances reported as anomalies may lead to possible algorithmic bias against the minority groups presented in the data, such as under-represented groups in fraud detection and crime detection systems. An effective approach to mitigate this type of risk is to have anomaly explanation (Challenge #6) algorithms that provide straightforward clues about why a specific data instance is identified as anomaly. Providing such explanation can be as important as detection accuracy in some applications. To derive anomaly explanation from specific detection methods is still a largely unsolved problem, especially for complex models. Developing inherently interpretable anomaly detection models is also crucial, but it remains a main challenge to well balance the model’s interpretability and effectiveness.
Addressing the Challenges with Deep Anomaly Detection
In a nutshell deep anomaly detection aims at learning feature representations or anomaly scores via neural networks for the sake of anomaly detection. In recent years, a large number of deep anomaly detection methods have been introduced, demonstrating significantly better performance than conventional anomaly detection on addressing challenging detection problems in a variety of real-world applications. We systematically review the current deep anomaly detection methods and their capabilities in addressing the aforementioned challenges.
To have a thorough understanding of the area, we introduce a hierarchical taxonomy to classify existing deep anomaly detection methods into three main categories and 11 fine-grained categories from the modeling perspective. An overview of the taxonomy of the methods, together with the challenges they address, is shown in Figure 1. Specifically, deep anomaly detection consists of three conceptual paradigms — Deep Learning for Feature Extraction, Learning Feature Representations of Normality, and End-to-end Anomaly Score Learning.
Figure 1. A Hierarchical Taxonomy of Current Deep Anomaly Detection Techniques. The detection challenges that each category of methods can address are also presented.
In the Deep Learning for Feature Extraction framework, deep learning and anomaly detection are fully separated in the first main category, so deep learning techniques are used as some independent feature extractors only. The two modules are dependent on each other in some form in the second main category — Learning Feature Representations of Normality, with an objective of learning expressive representations of normality. This category of methods can be further divided into two subcategories based on whether traditional anomaly measures are incorporated into their objective functions. These two subcategories encompass seven fine-grained categories of methods, with each category taking a different approach to formulate its objective function. The two modules are fully unified in the third main category — End-to-end Anomaly Score Learning, in which the methods are dedicated to learning anomaly scores via neural networks in an end-to-end fashion. These methods are further grouped into four categories based on the formulation of neural network-enabled anomaly scoring.
For each category of methods, we review detailed methodology and algorithms, covering their key intuitions, objective functions, underlying assumptions, advantages and disadvantages, and discuss how they address the aforementioned challenges. The full details are difficult to demonstrate here. See the full paper below for detail.
Future Opportunities
Through such a review, we identify some exciting opportunities. Some of them are described as follows.
Exploring Anomaly-supervisory Signals
Informative supervisory signals are the key for deep anomaly detection to learn expressive representations of normality/abnormality or anomaly scores and reduce false positives. While a wide range of unsupervised or self-supervised supervisory signals have been explored, to learn the representations, a key issue for these formulations is that their objective functions are generic but not optimized specifically for anomaly detection. Current anomaly measure-dependent feature learning approaches help address this issue by imposing constraints derived from traditional anomaly measures. However, these constraints can have some inherent limitations, e.g., implicit assumptions in the anomaly measures. It is critical to explore new sources of anomaly-supervisory signals that lie beyond the widely-used formulations such as data reconstruction and GANs, and have weak assumptions on the anomaly distribution. Another possibility is to develop domain-driven anomaly detection by leveraging domain knowledge such as application-specific knowledge of anomaly and/or expert rules as the supervision source.
Deep Weakly-supervised Anomaly Detection
Deep weakly-supervised anomaly detection aims at leveraging deep neural networks to learn anomaly-informed detection models with some weakly-supervised anomaly signals, e.g.,, partially/inexactly/inaccurately labeled anomaly data. This labeled data provides important knowledge of anomaly and can be a major driving force to lift detection recall rates. One exciting opportunity is to utilize a small number of accurate labeled anomaly examples to enhance detection models as they are often available in real-world applications, e.g., some intrusions/frauds from deployed detection systems/end-users and verified by human experts. However, since anomalies can be highly heterogeneous, there can be unknown/novel anomalies that lie beyond the span set of the given anomaly examples. Thus, one important direction here is unknown anomaly detection, in which we aim to build detection models that are generalized from the limited labeled anomalies to unknown anomalies.
To detect anomalies that belong to the same classes of the given anomaly examples can be as important as the detection of novel/unknown anomalies. Thus, another important direction is to develop data-efficient anomaly detection or few-shot anomaly detection, in which we aim at learning highly expressive representations of the known anomaly classes given only limited anomaly examples. It should be noted that the limited anomaly examples may come from different anomaly classes and thus exhibit completely different manifold/class features. This scenarios is fundamentally different from the general few-shot learning, in which the limited examples are class-specific and assumed to share the same manifold/class structure.
Large-scale Normality Learning
Large-scale unsupervised/self-supervised representation learning has gained tremendous success in enabling downstream learning tasks. This is particular important for learning tasks, in which it is difficult to obtain sufficient labeled data, such as anomaly detection. The goal is to learn transferable pre-trained representation models from large-scale unlabeled data in an unsupervised/self-supervised mode and fine-tune detection models in a semi-supervised mode. Self-supervised classification-based anomaly detection methods may provide some initial sources of supervision for the normality learning. However, precautions must be taken to ensure that (i) the unlabeled data is free of anomaly contamination and/or (ii) the representation learning methods are robust w.r.t. possible anomaly contamination. This is because most methods implicitly assume that the training data is clean and does not contain any noise/anomaly instances. This robustness is important in both the pre-trained modeling and the fine-tuning stage. Additionally, anomalies and datasets in different domains vary significantly, so the large-scale normality learning may need to be domain/application-specific.
If you find the summarization of the survey paper interesting and helpful, you can read the full paper for detail.
[1] Guansong Pang, Chunhua Shen, Longbing Cao, Anton van den Hengel. “Deep Learning for Anomaly Detection: A Review”. 2020. arXiv preprint: 2007.02500.
欢迎转载，转载请保留页面地址。帮助到你的请点个推荐。
