Scholarlink

The exponential expansion of the Internet of Things (IoT) has significantly amplified the attack surface for botnet-enabled Distributed Denial of Service (DDoS) attacks, exploiting the intrinsic limitations of IoT devices namely constrained processing power, memory, and security architecture. Despite the proliferation of Machine Learning (ML) and Deep Learning (DL) methods for IoT botnet detection, existing solutions suffer from critical shortcomings: dependence on narrowly scoped datasets, limited cross-platform generalizability, inadequate support for real-time detection, and poor resilience against zero-day and adversarial threats. Including classical classifiers, deep architecture, hybrid models, and ensemble frameworks, this review offers a complete and technically grounded taxonomy of ML and DL-based detection methods. Also, it systematically analyzes detection accuracy, computational overhead, interpretability, and real-world deployment feasibility across benchmark datasets such as N-BaIoT, BoT-IoT, and UNSW-NB15 etc. A key contribution of this work lies in identifying persistent gaps such as the lack of adaptive, explainable, and resource-aware solutions and in proposing future research directions.

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 Review of Machine Learning and Deep Learning Approaches for Botnet Detection in IoT Environments

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Rahul Modak,
  • Santanu Phadikar,
  • Koushik Majumder,
  • Anurag Dasgupta

摘要

The exponential expansion of the Internet of Things (IoT) has significantly amplified the attack surface for botnet-enabled Distributed Denial of Service (DDoS) attacks, exploiting the intrinsic limitations of IoT devices namely constrained processing power, memory, and security architecture. Despite the proliferation of Machine Learning (ML) and Deep Learning (DL) methods for IoT botnet detection, existing solutions suffer from critical shortcomings: dependence on narrowly scoped datasets, limited cross-platform generalizability, inadequate support for real-time detection, and poor resilience against zero-day and adversarial threats. Including classical classifiers, deep architecture, hybrid models, and ensemble frameworks, this review offers a complete and technically grounded taxonomy of ML and DL-based detection methods. Also, it systematically analyzes detection accuracy, computational overhead, interpretability, and real-world deployment feasibility across benchmark datasets such as N-BaIoT, BoT-IoT, and UNSW-NB15 etc. A key contribution of this work lies in identifying persistent gaps such as the lack of adaptive, explainable, and resource-aware solutions and in proposing future research directions.