Scholarlink

<p>Connected devices improve life quality and generate a large amount of data which require computation and transfer; the need for security becomes a major concern with networks evolving in complexity and scale. Analyzing network packets for identifying deviations from the standard behavior is called anomaly detection. Deep learning tools have emerged as a promising alternative over classical machine learning approaches due to their proficiency in feature modeling, appraise detection rate, and mirror cognitive development. The review outlines the role of deep learning techniques in feature extraction and classification of outliers across application domains. A detailed analysis of standard datasets and methodologies of latest experimental research is performed. Scholarly contributions for the past decade are reflected in terms of techniques employed, performance metrics, datasets, challenges faced, and application domains. Eventually, the paper explores future research directions, emphasizing the need for hybrid models, advanced hardware, and improved interpretability.</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 comprehensive survey on techniques, challenges, evaluation metrics and applications of deep learning models for anomaly detection

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年7月15日

查看全文
  • Mudita Kohli,
  • Indu Chhabra

摘要

Connected devices improve life quality and generate a large amount of data which require computation and transfer; the need for security becomes a major concern with networks evolving in complexity and scale. Analyzing network packets for identifying deviations from the standard behavior is called anomaly detection. Deep learning tools have emerged as a promising alternative over classical machine learning approaches due to their proficiency in feature modeling, appraise detection rate, and mirror cognitive development. The review outlines the role of deep learning techniques in feature extraction and classification of outliers across application domains. A detailed analysis of standard datasets and methodologies of latest experimental research is performed. Scholarly contributions for the past decade are reflected in terms of techniques employed, performance metrics, datasets, challenges faced, and application domains. Eventually, the paper explores future research directions, emphasizing the need for hybrid models, advanced hardware, and improved interpretability.