Rule extraction in unsupervised anomaly detection for model explainability: Application to OneClass SVM

Barbado, Alberto; Corcho, Óscar; Benjamins, Richard

doi:10.1016/j.eswa.2021.116100

Cited by 37 publications

(5 citation statements)

References 21 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…These requirements aim to check the complexity of interaction between end-users and the suggested explainable approaches and overview the end-user expectations. The operational aspect covers the interaction with the end-user, the trade-off between model explanation and prediction and the user awareness needed to use the explanatory power of the methodology effectively (Barbado et al. , 2022).…”

Section: Fact Sheets Analysis: Benchmarking Of Two Methodsmentioning

confidence: 99%

When to choose ranked area integrals versus integrated gradient for explainable artificial intelligence – a comparison of algorithms

Singh

Konovalova

Kar

2022

BIJ

View full text Add to dashboard Cite

PurposeExplainable artificial intelligence (XAI) has importance in several industrial applications. The study aims to provide a comparison of two important methods used for explainable AI algorithms.Design/methodology/approachIn this study multiple criteria has been used to compare between explainable Ranked Area Integrals (xRAI) and integrated gradient (IG) methods for the explainability of AI algorithms, based on a multimethod phase-wise analysis research design.FindingsThe theoretical part includes the comparison of frameworks of two methods. In contrast, the methods have been compared across five dimensions like functional, operational, usability, safety and validation, from a practical point of view.Research limitations/implicationsA comparison has been made by combining criteria from theoretical and practical points of view, which demonstrates tradeoffs in terms of choices for the user.Originality/valueOur results show that the xRAI method performs better from a theoretical point of view. However, the IG method shows a good result with both model accuracy and prediction quality.

show abstract

Section: Fact Sheets Analysis: Benchmarking Of Two Methodsmentioning

confidence: 99%

When to choose ranked area integrals versus integrated gradient for explainable artificial intelligence – a comparison of algorithms

Singh

Konovalova

Kar

2022

BIJ

View full text Add to dashboard Cite

show abstract

“…A second category of anomaly explanation methods groups approaches that additionally associate the responsible features with the values they take, as for instance [2]. The latter can for instance be identified by rules expressed as conjunction of predicates, where explanations take a disjunctive normal form.…”

Section: Anomaly Explanationmentioning

confidence: 99%

PANDA: Human-in-the-Loop Anomaly Detection and Explanation

Smits

Lesot

Tchaghe

et al. 2022

Information Processing and Management of Uncertainty in Knowledge-Based Systems

View full text Add to dashboard Cite

The paper addresses the tasks of anomaly detection and explanation simultaneously, in the human-in-the-loop paradigm integrating the end-user expertise: it first proposes to exploit two complementary data representations to identify anomalies, namely the description induced by the raw features and the description induced by a user-defined vocabulary. These representations respectively lead to identify so-called data-driven and knowledge-driven anomalies. The paper then proposes to confront these two sets of instances so as to improve the detection step and to dispose of tools towards anomaly explanations. It distinguishes and discusses three cases, underlining how the two description spaces can benefit from one another, in terms of accuracy and interpretability.

show abstract

“…Recently, there are a few studies discussing interpretation methods on unsupervised learning or anomaly detection. First, some studies develop global interpretation through model distillation [49] and rule extraction [5] as well as interpretation for clustering models [25], which are beyond the scope of this study. As for local interpretation of anomaly detection, in addition to COIN [31] and CADE [58] which are used as baselines, most of other works simply modify the existing supervised methods to unsupervised learning.…”

Section: Related Workmentioning

confidence: 99%

DeepAID: Interpreting and Improving Deep Learning-based Anomaly Detection in Security Applications

Han

Wang

Chen

et al. 2021

Proceedings of the 2021 ACM SIGSAC Conference on Computer and Communications Security

View full text Add to dashboard Cite

Unsupervised Deep Learning (DL) techniques have been widely used in various security-related anomaly detection applications, owing to the great promise of being able to detect unforeseen threats and the superior performance provided by Deep Neural Networks (DNN). However, the lack of interpretability creates key barriers to the adoption of DL models in practice. Unfortunately, existing interpretation approaches are proposed for supervised learning models and/or non-security domains, which are unadaptable for unsupervised DL models and fail to satisfy special requirements in security domains.In this paper, we propose DeepAID, a general framework aiming to (1) interpret DL-based anomaly detection systems in security domains, and (2) improve the practicality of these systems based on the interpretations. We first propose a novel interpretation method for unsupervised DNNs by formulating and solving well-designed optimization problems with special constraints for security domains. Then, we provide several applications based on our Interpreter as well as a model-based extension Distiller to improve security systems by solving domain-specific problems. We apply DeepAID over three types of security-related anomaly detection systems and extensively evaluate our Interpreter with representative prior works. Experimental results show that DeepAID can provide high-quality interpretations for unsupervised DL models while meeting several special requirements in security domains. We also provide several use cases to show that DeepAID can help security operators to understand model decisions, diagnose system mistakes, give feedback to models, and reduce false positives.

show abstract

Rule extraction in unsupervised anomaly detection for model explainability: Application to OneClass SVM

Cited by 37 publications

References 21 publications

When to choose ranked area integrals versus integrated gradient for explainable artificial intelligence – a comparison of algorithms

When to choose ranked area integrals versus integrated gradient for explainable artificial intelligence – a comparison of algorithms

PANDA: Human-in-the-Loop Anomaly Detection and Explanation

DeepAID: Interpreting and Improving Deep Learning-based Anomaly Detection in Security Applications

Contact Info

Product

Resources

About