Abstract:Anomaly detection (AD) is a crucial process often required in industrial settings. Anomalies can signal underlying issues within a system, prompting further investigation. Industrial processes aim to streamline operations as much as possible, encompassing the production of the final product, making AD an essential mean to reach this goal.Conventional anomaly detection methodologies typically classify observations as either normal or anomalous without providing insight into the reasons behind these classifications.Consequently, in light of the emergence of Industry 5.0, a more desirable approach involves providing interpretable outcomes, enabling users to understand the rationale behind the results.This paper presents the first industrial application of ExIFFI, a recently developed approach focused on the production of fast and efficient explanations for the Extended Isolation Forest (EIF) Anomaly detection method. ExIFFI is tested on two publicly available industrial datasets demonstrating superior effectiveness in explanations and computational efficiency with the respect to other state-of-the-art explainable AD models.

What problem does this paper attempt to address?

The paper primarily aims to address the issue of Anomaly Detection (AD) in industrial environments, with a particular focus on improving the interpretability of anomaly detection models. Specifically, the paper proposes a new method called ExIFFI, which is a feature importance evaluation algorithm designed specifically for the Extended Isolation Forest (EIF) anomaly detection method. Through ExIFFI, researchers can quickly and efficiently explain how the EIF model makes decisions, thereby helping users understand the logic behind the model's identification of anomalies. Key points in the paper include: - **Background and Motivation**: In the context of the Industrial Internet of Things (IIoT), anomaly detection is crucial for timely identification of system failures. Traditional anomaly detection methods often only provide a classification result of normal or abnormal without explaining the reasons, which limits their value in practical applications. Therefore, with the development of Industry 5.0, researchers are increasingly focusing on the interpretability of models. - **Proposed Method**: The ExIFFI algorithm is designed to provide a quick explanation of the EIF model's decision-making process. It determines the importance of features by evaluating each feature's contribution to the imbalance degree of anomaly samples. - **Experimental Validation**: To validate the effectiveness of ExIFFI, researchers conducted experiments on two public industrial datasets: the Tennessee Eastman Process (TEP) and the Packaging Industry Anomaly Detection (PIADE) datasets. The results show that ExIFFI can not only accurately identify anomalies but also its explanation results align with known causes of anomalies. - **Comparison with Other Methods**: ExIFFI was compared with existing explainable AI methods such as DIFFI, AcME-AD, and KernelSHAP. Experiments indicate that ExIFFI excels in terms of explanation quality and computational efficiency, particularly outperforming other methods in time efficiency. - **Conclusion and Future Work**: ExIFFI provides effective and easy-to-understand explanations for anomaly detection in industrial environments. Future research may explore how to apply ExIFFI to resource-constrained devices to enhance their anomaly detection capabilities. In summary, by proposing the ExIFFI method, the paper addresses the issue of model interpretability in industrial anomaly detection, providing valuable tools and technical support for industrial practice.

Interpretable Data-driven Anomaly Detection in Industrial Processes with ExIFFI

Enhancing Interpretability and Generalizability in Extended Isolation Forests

Enabling Efficient and Flexible Interpretability of Data-driven Anomaly Detection in Industrial Processes with AcME-AD

Interpretable Anomaly Detection with DIFFI: Depth-based Isolation Forest Feature Importance

Adaptable and Interpretable Framework for Anomaly Detection in SCADA-based industrial systems

AcME-AD: Accelerated Model Explanations for Anomaly Detection

Information Fingerprint for Secure Industrial Big Data Analytics.

Self-Supervised Iterative Refinement for Anomaly Detection in Industrial Quality Control

Interpreting Outliers in Time Series Data through Decoding Autoencoder

An Echo State Network-Based Light Framework for Online Anomaly Detection: An Approach to Using AI at the Edge

Isolation Forests and Deep Autoencoders for Industrial Screw Tightening Anomaly Detection

Neuro-symbolic Empowered Denoising Diffusion Probabilistic Models for Real-time Anomaly Detection in Industry 4.0

Interactive Explainable Anomaly Detection for Industrial Settings

Anomaly Detection for Industrial Big Data

Data-Efficient and Interpretable Tabular Anomaly Detection

XAI-IoT: An Explainable AI Framework for Enhancing Anomaly Detection in IoT Systems

An Energy-efficient and Trustworthy Unsupervised Anomaly Detection Framework (EATU) for IIoT

Anomaly-based Intrusion Detection in Industrial Data with SVM and Random Forests

Artificial immunity based distributed and fast anomaly detection for Industrial Internet of Things

EIFDAA: Evaluation of an IDS with function-discarding adversarial attacks in the IIoT