Scholarlink

<p>Internet-enabled intelligent devices have become deeply rooted in both consumer and industrial systems, significantly increasing a wide range of applications such as smart homes and healthcare systems to automated technology. This rapid growth has resulted in massive traffic complexity, giving rise to highly advanced and ever-adapting cyber threats that exploit individuals’ lifestyles. Existing detection mechanisms require proactive improvisation to identify these emerging attack strategies. However, there is an urgent demand for pioneering, flexible systems that are effective in securing devices from the widespread occurrence of cyber-attacks. Motivated by this challenge, this paper introduces the novel combination of multi-ladder gated network units (ML-GNU) and the Improved Frilled Lizard Optimization (IFLO) model to recognize distinct types of multiple cyber-threats. These methods are utilized to create an intelligent framework structured to defend against both known and unknown threat patterns. The proposed ML-GNUs are used for differentiating diverse attack patterns, and the improved frilled lizard technique is employed to tune the model’s hyper-parameters, reducing computational overhead and enhancing detection performance. The efficiency of the proposed model is assessed utilising both simulated real-time data traffic patterns and the CIC-IDS-2017 benchmark dataset, and standard performance metrics like accuracy, precision, recall, specificity, Matthews correlation coefficient (MCC), and F1-score are analyzed and measured. To validate the excellence of the recommended hybrid approach, the analyzed metrics are compared to other prevailing methods. Simulation results demonstrate the model’s strong efficiency over existing approaches, attaining an average accuracy of 99.9%, precision of 99.56%, recall of 99.5%, and F1-score of 98.7%. The observations indicate that the proposed method substantially increases system security by accurately recognizing new threats and adapting to dynamic attack strategies.</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 novel cluster of improved frilled lizard optimization and multi-ladder gated networks for the detection of cyber-attacks in computer networks

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年6月14日

查看全文
  • Vandana Dharmapuri,
  • Sushama Rani Dutta

摘要

Internet-enabled intelligent devices have become deeply rooted in both consumer and industrial systems, significantly increasing a wide range of applications such as smart homes and healthcare systems to automated technology. This rapid growth has resulted in massive traffic complexity, giving rise to highly advanced and ever-adapting cyber threats that exploit individuals’ lifestyles. Existing detection mechanisms require proactive improvisation to identify these emerging attack strategies. However, there is an urgent demand for pioneering, flexible systems that are effective in securing devices from the widespread occurrence of cyber-attacks. Motivated by this challenge, this paper introduces the novel combination of multi-ladder gated network units (ML-GNU) and the Improved Frilled Lizard Optimization (IFLO) model to recognize distinct types of multiple cyber-threats. These methods are utilized to create an intelligent framework structured to defend against both known and unknown threat patterns. The proposed ML-GNUs are used for differentiating diverse attack patterns, and the improved frilled lizard technique is employed to tune the model’s hyper-parameters, reducing computational overhead and enhancing detection performance. The efficiency of the proposed model is assessed utilising both simulated real-time data traffic patterns and the CIC-IDS-2017 benchmark dataset, and standard performance metrics like accuracy, precision, recall, specificity, Matthews correlation coefficient (MCC), and F1-score are analyzed and measured. To validate the excellence of the recommended hybrid approach, the analyzed metrics are compared to other prevailing methods. Simulation results demonstrate the model’s strong efficiency over existing approaches, attaining an average accuracy of 99.9%, precision of 99.56%, recall of 99.5%, and F1-score of 98.7%. The observations indicate that the proposed method substantially increases system security by accurately recognizing new threats and adapting to dynamic attack strategies.