Scholarlink

<p>The detection of anomalous patterns in smart home environments represents a major challenge because traditional modelling approaches experience difficulties distinguishing between standard behaviour and abnormal patterns, resulting in incorrect alerting or undetected incidents. Anomaly detection requires exact identification to improve system security and enhance both efficiency and user experience. The novel framework Anomaly-Enhanced Activity Recognition Iteration (AEARI) applies user feedback to perform successive improvement of anomaly detection capabilities. AEARI operates with two machine learning components: activity detection relies on Random Forest (RF) because it efficiently recognises user operations, and One-Class SVM (OCSVM) executes anomaly detection by detecting variations from standard user conduct. The system obtains feedback from users and uses these inputs to adapt behaviour detection capabilities while it evolves into a more accurate system through time. The adaptive learning method decreases false positives in the model while improving its capability to detect previously unseen behaviours, thus creating a dependable anomaly detection system. The experimental analysis using smart home data demonstrates AEARI has a performance rate of 98% for both accuracy and precision along with recall and F1-score according to metric scores, whereas it outperforms support vector machine (SVM), k-mean algorithm (K-Means) and Probabilistic Neural Network<b> (</b>PNN)-based models. The system benefits from AEARI through its ability to decrease false positives by 15%, thereby improving system reliability. This framework includes a real-time adaptive model refinement feature that provides both instant scalability and adaptability, which makes it an ideal modern security solution for smart home applications. AEARI introduces a standard of advanced anomaly detection technology through its user-empowered continuous learning process for smart homes. The source code for the proposed framework can be accessed at <a href="https://gitlab.com/khalidsophee/aeari-framework.git">https://gitlab.com/khalidsophee/aeari-framework.git</a></p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A feedback-driven anomaly detection framework for smart homes: an anomaly-enhanced activity recognition iteration (AEARI) approach

  • 期刊论文

  • 发表日期：

    2025年11月6日

查看全文
  • Khalid Aziz,
  • Sakshi Dua,
  • Prabal Gupta

摘要

The detection of anomalous patterns in smart home environments represents a major challenge because traditional modelling approaches experience difficulties distinguishing between standard behaviour and abnormal patterns, resulting in incorrect alerting or undetected incidents. Anomaly detection requires exact identification to improve system security and enhance both efficiency and user experience. The novel framework Anomaly-Enhanced Activity Recognition Iteration (AEARI) applies user feedback to perform successive improvement of anomaly detection capabilities. AEARI operates with two machine learning components: activity detection relies on Random Forest (RF) because it efficiently recognises user operations, and One-Class SVM (OCSVM) executes anomaly detection by detecting variations from standard user conduct. The system obtains feedback from users and uses these inputs to adapt behaviour detection capabilities while it evolves into a more accurate system through time. The adaptive learning method decreases false positives in the model while improving its capability to detect previously unseen behaviours, thus creating a dependable anomaly detection system. The experimental analysis using smart home data demonstrates AEARI has a performance rate of 98% for both accuracy and precision along with recall and F1-score according to metric scores, whereas it outperforms support vector machine (SVM), k-mean algorithm (K-Means) and Probabilistic Neural Network (PNN)-based models. The system benefits from AEARI through its ability to decrease false positives by 15%, thereby improving system reliability. This framework includes a real-time adaptive model refinement feature that provides both instant scalability and adaptability, which makes it an ideal modern security solution for smart home applications. AEARI introduces a standard of advanced anomaly detection technology through its user-empowered continuous learning process for smart homes. The source code for the proposed framework can be accessed at https://gitlab.com/khalidsophee/aeari-framework.git