Abstract—

From the last decade, the use of internet and its growth is continuously increasing. Similarly, numbers of services are coming out along with the internet and it is being used for providing facilities to human beings. Wireless sensor have been used for various application such as fire safety, military application, petroleum industry, security system, monitoring and environmental condition and many more. WSN node exposes itself to various security related attacks due to low battery power supply, low bandwidth support, data transmission over multi hop node, dependency on intermediate or other nodes, distributed in nature and self-organization. The WSN attacks observe in all layers of OSI model. Wireless sensor nodes has various issues because of that, it experiences number problem related to its functionalities and some malfunction due to attacks. It is require to build defence and network monitoring system for identifying attacks and prevent them. Intrusion detection system (IDS) plays an important role to detect threads inside the system and generate the alert related to the attack. In this work, supervised classification models for intrusion detection are built using such as Random Forest classifier, Support Vector Machine, Decision Tree Classifier, LGBM Classifier, Extra Tree Classifier, Gradient Boosting Classifier, Ada Boost Classifier, K Nearest Neighbour Classifier, MLP Classifier, Gaussian Naive Bayes Classifier and Logistic Regression Classifier. The NSLKDD, i.e. Modified version of the KDD99 Data Set on which we checks these algorithms. Experimental results how the highest accuracy relative to other classification systems in the support vector machine.

中文翻译：

---

无线传感器网络中入侵检测异常类型的监督机器学习分类算法

摘要—

从过去的十年开始，互联网的使用及其增长一直在不断增长。类似地，大量服务与互联网一起出现，并被用于为人类提供设施。无线传感器已用于各种应用，例如消防安全，军事应用，石油工业，安全系统，监视和环境状况等等。由于电池电量低，带宽支持低，多跳节点上的数据传输，对中间节点或其他节点的依赖，自然分布和自组织，WSN节点使其自身遭受各种与安全相关的攻击。WSN攻击在OSI模型的所有层中均会观察到。因此，无线传感器节点存在各种问题，它会遇到与其功能相关的数字问题以及由于攻击而导致的某些故障。需要建立防御和网络监控系统来识别和预防攻击。入侵检测系统（IDS）在检测系统内部的线程并生成与攻击有关的警报方面起着重要作用。在这项工作中，使用随机森林分类器，支持向量机，决策树分类器，LGBM分类器，额外树分类器，梯度提升分类器，Ada Boost分类器，K最近邻居分类器，MLP分类器，高斯朴素贝叶斯分类器和逻辑回归分类器。NSLKDD，即我们在其上检查这些算法的KDD99数据集的修改版本。实验结果表明，相对于支持向量机中的其他分类系统，其准确性最高。