Scholarlink

<p>As technology advances in the data-driven era, the number of systems connected to the internet and the volume of data generated are rapidly increasing. Consequently, network security is becoming increasingly significant. Internet of Things (IoT) incorporates intelligence into internet-connected devices to communicate with one another, exchange data, make decisions, trigger actions, and provide wonderful services to us. Further, IoT domai ns present numerous challenges in terms of energy, latency, bandwidth, reliability, security, in-network data processing, and latency that must be appropriately considered, necessitating specific Software Defined Networks (SDN) paradigm enhancements. Due to its extreme adaptability and programmability, SDN has the potential to revolutionize the management of various IoT environments. This paper introduces a novel Intrusion Detection System (IDS) based on a deep learning model, which has been shown to improve the effectiveness and performance of IDS in various domains. The steps involved in this process are (a) The first step is to collect data from popular repositories such as InSDN, UNSW-NB15, Bot-IoT, and ToN-IoT. (b) Once the raw data is obtained, it undergoes preprocessing, including data transmission, cleaning, reduction, and discretization. (c) Following the preprocessing step, to extract certain features from the preprocessed data, an AutoEncoder(AE) is used. (d) Lastly, a Ridge classifier analyzes the extracted features for intrusion detection. The proposed system’s effectiveness was evaluated against existing models using various datasets, including InSDN, UNSW-NB15, BoT-IoT, and ToN-IoT, as well as platforms such as Windows (Windows 7, Windows 10), Linux (Disk Activity, Process Scheduling, Memory Activity), Network, and IoT devices (Fridge, Garage Door, GPS Tracker, Modbus, Motion Light, Thermostat, Weather). Evaluation metrics such as precision, recall, accuracy, F1-score, sensitivity, and specificity were employed. The recorded accuracy for each dataset, in the respective order mentioned above, was as follows: 0.9862, 0.9748, 0.9336, 0.9962, 0.9997, 0.8207, 0.84, 0.8417, 0.8393, 0.7677, 0.8842, 0.8139, 0.9784, 0.8348, 0.911, 0.9152.</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

Enhancing intrusion detection: protocol-based security using a hybrid RIDGE classifier on InSDN, UNSW-NB15, BoT-IoT, and ToN-IoT datasets

  • 期刊论文

  • 发表日期：

    2025年9月3日

查看全文
  • Anand Nemalikanti,
  • Sreenija Kaki,
  • Rami Reddy Ambati,
  • Raveendra Babu Ponnuru

摘要

As technology advances in the data-driven era, the number of systems connected to the internet and the volume of data generated are rapidly increasing. Consequently, network security is becoming increasingly significant. Internet of Things (IoT) incorporates intelligence into internet-connected devices to communicate with one another, exchange data, make decisions, trigger actions, and provide wonderful services to us. Further, IoT domai ns present numerous challenges in terms of energy, latency, bandwidth, reliability, security, in-network data processing, and latency that must be appropriately considered, necessitating specific Software Defined Networks (SDN) paradigm enhancements. Due to its extreme adaptability and programmability, SDN has the potential to revolutionize the management of various IoT environments. This paper introduces a novel Intrusion Detection System (IDS) based on a deep learning model, which has been shown to improve the effectiveness and performance of IDS in various domains. The steps involved in this process are (a) The first step is to collect data from popular repositories such as InSDN, UNSW-NB15, Bot-IoT, and ToN-IoT. (b) Once the raw data is obtained, it undergoes preprocessing, including data transmission, cleaning, reduction, and discretization. (c) Following the preprocessing step, to extract certain features from the preprocessed data, an AutoEncoder(AE) is used. (d) Lastly, a Ridge classifier analyzes the extracted features for intrusion detection. The proposed system’s effectiveness was evaluated against existing models using various datasets, including InSDN, UNSW-NB15, BoT-IoT, and ToN-IoT, as well as platforms such as Windows (Windows 7, Windows 10), Linux (Disk Activity, Process Scheduling, Memory Activity), Network, and IoT devices (Fridge, Garage Door, GPS Tracker, Modbus, Motion Light, Thermostat, Weather). Evaluation metrics such as precision, recall, accuracy, F1-score, sensitivity, and specificity were employed. The recorded accuracy for each dataset, in the respective order mentioned above, was as follows: 0.9862, 0.9748, 0.9336, 0.9962, 0.9997, 0.8207, 0.84, 0.8417, 0.8393, 0.7677, 0.8842, 0.8139, 0.9784, 0.8348, 0.911, 0.9152.