Scholarlink

<p>Security has become a primary issue in IoT (Internet of Things) because of its complex environment, handling multiple devices, several connection methods, and transmission of a huge volume of data. Traditional techniques don’t manage datasets or learn features optimally, which degrades attack detection accuracy. To solve the issues, an effective Intrusion Detection System (IDS) is proposed for detecting security risks in IoT systems. The framework comprises three phases, namely data pre-processing, feature selection, and classification. Pre-processing is applied to the data gathered from the public repository. The best features are then selected from the pre-processed dataset by inputting the pre-processed data into the Walrus Optimization Algorithm (WOA). It is followed by data classification using Bidirectional Gated Recurrent Unit-Convolutional Neural Network (BiGRU-CNN). This framework can classify data into five categories, such as normal (benign), DoS (Denial of Service), Shellcode (U2R), Fuzzers (R2L), and Probe attacks. The effectiveness of the proposed framework is evaluated based on metrics such as intrusion detection rate (IDR %) and network accuracy. The results are highly satisfactory, leading to network accuracy and IDR as 98.20% and 98.15%, respectively.</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

Intrusion detection system with walrus optimization algorithm (WOA) and BiGRU-CNN for securing IoT systems

  • 期刊论文

  • 发表日期：

    2025年4月29日

查看全文
  • Ahilan Appathurai,
  • P. Bachan,
  • Ruchi Kaushik,
  • A. Tamizhselvi,
  • Rajshree Jodha,
  • Kulwant Singh

摘要

Security has become a primary issue in IoT (Internet of Things) because of its complex environment, handling multiple devices, several connection methods, and transmission of a huge volume of data. Traditional techniques don’t manage datasets or learn features optimally, which degrades attack detection accuracy. To solve the issues, an effective Intrusion Detection System (IDS) is proposed for detecting security risks in IoT systems. The framework comprises three phases, namely data pre-processing, feature selection, and classification. Pre-processing is applied to the data gathered from the public repository. The best features are then selected from the pre-processed dataset by inputting the pre-processed data into the Walrus Optimization Algorithm (WOA). It is followed by data classification using Bidirectional Gated Recurrent Unit-Convolutional Neural Network (BiGRU-CNN). This framework can classify data into five categories, such as normal (benign), DoS (Denial of Service), Shellcode (U2R), Fuzzers (R2L), and Probe attacks. The effectiveness of the proposed framework is evaluated based on metrics such as intrusion detection rate (IDR %) and network accuracy. The results are highly satisfactory, leading to network accuracy and IDR as 98.20% and 98.15%, respectively.