Semi-supervised learning-based IDS using active learning

Abstract

Due to the increasing number of attacks, IDS systems are becoming more essential to protect networks and devices from malicious attacks. The current study aims to enhance Intrusion Detection System performance by utilizing the least amount of labeled instances.To achieve the objective of this study, a semi-supervised learning approach was used that combines machine learning and active learning method, using the CICIDS 2017 dataset. Our experimental evaluation shows the effectiveness of incorporating active learning into semi-supervised learning for IDS, as the proposed approach significantly outperforms traditional machine learning methods in terms of accuracy, precision, detection rate, and false positive rate.

‫بسبب‬‫الزيادة‬ ‫المستمرة‬ ‫في‬ ‫الهجمات‪،‬‬ ‫أصبحت‬ ‫أنظمة‬ ‫الكشف‬ ‫عن‬ ‫ا‬ ‫ل‬‫تسلل‬‫(‬ ‫‪IDS‬‬‫)‬‫ضرورية‬ ‫لحماية‬ ‫الشبكات‬ ‫واألجهزة‬ ‫من‬‫التهديدات‬ ‫الخبيثة‪.‬‬ ‫تهدف‬ ‫هذه‬ ‫الدراسة‬ ‫إلى‬ ‫تحسين‬ ‫أداء‬ ‫نظام‬ ‫الكشف‬ ‫عن‬ ‫االختراق‬ ‫باستخدام‬ ‫أقل‬ ‫عدد‬ ‫ممكن‬ ‫من‬ ‫العينات‬‫المصنفة‪.‬‬ ‫لتحقيق‬‫هذا‬ ‫الهدف‪،‬‬ ‫تم‬ ‫اعتماد‬ ‫نهج‬ ‫تعلم‬ ‫شبه‬ ‫خاضع‬ ‫لإلشراف‬ ‫يجمع‬ ‫بين‬ ‫التعلم‬ ‫اآللي‬ ‫والتعلم‬ ‫النشط‪،‬‬ ‫وذلك‬ ‫باستخدا‬ ‫م‬ ‫مجموعة‬‫بيانات‬ ‫‪2017‬‬ ‫‪CICIDS‬‬ ‫‪.‬‬ ‫أظهرت‬ ‫نتائج‬ ‫التقييم‬ ‫التجريبي‬ ‫أن‬ ‫دمج‬ ‫التعلم‬ ‫النشط‬ ‫في‬ ‫التعلم‬ ‫شبه‬ ‫الخاضع‬ ‫لإلشراف‬ ‫يعزز‬‫بشكل‬ ‫كبير‬ ‫فعالية‬ ‫نظام‬ ‫الكشف‬ ‫عن‬ ‫االختراق‪،‬‬ ‫متفوقًا‬ ‫على‬ ‫األساليب‬ ‫التقليدية‬ ‫للتعلم‬ ‫اآللي‬ ‫من‬ ‫حيث‬ ‫الدقة‪،‬‬ ‫ال‬ ‫ضبط‪،‬‬ ‫معدل‬‫اإلكتشاف‬ ‫و‬ ‫معدل‬‫االنذارات‬ ‫االيجابي‬ ‫ة‬‫الكاذبة‬ ‫‪.‬‬