Detection and classification of DDoS attack using artificial neural network

مقدم أطروحة جامعية

al-Musayidin, Muhammad Abd Allah

مشرف أطروحة جامعية

al-Kasasibah, Muhammad Sharari Zamil

أعضاء اللجنة

al-Abbadi, Muhammad Ali Husayn
al-Hasanat, Ahmad Bashir
al-Sarayirah, Jafar Muhammad

الجامعة

جامعة مؤتة

الكلية

كلية تكنولوجيا المعلومات

دولة الجامعة

الأردن

الدرجة العلمية

ماجستير

تاريخ الدرجة العلمية

2015

الملخص الإنجليزي

Distributed denial of service (DDoS) attacks are considered an ongoing challenge for users and organizations.

The security engineer works to maintain a service at all times by dealing with intruder attacks.

Intrusion-detection systems (IDS) are one of the solutions used to detect and classify any abnormal behavior.

An IDS system must be constantly updated with all the latest techniques to deal with intruder attacks in order to preserve service availability.

In this thesis, we study the effects of DDoS attacks in both the network layer and application layer, including most modern DDoS attacks such as (SIDDOS and HTTP Flood) attacks.

We have also created a system to collect a dataset from a controlled environment using a network simulator.

The dataset was generated through the following stages: data collecting, data preprocessing and classification.

Unlike other datasets, the proposed dataset includes 2,160,668 records with 28 attributes and with no duplicate records.

The proposed dataset includes four types of attacks, organized as follows: (Smurf, UDP-Flood, HTTP-Flood, and SIDDOS).

Multilayer Perceptron (MLP), Naïve Bayes and Random Forest algorithms were used for training and testing on the proposed dataset to evaluate the dataset models.

The MLP classifier achieved the highest accuracy rate (98.63%) for detecting and classifying DDoS attacks with the longest time for building the training model; the Random Forest classifier achieved 98.01% for detecting and classifying DDoS attacks; the Naïve Bayes achieved 96.91% for detecting and classifying DDoS attacks, and therefore the Naïve Bayes classifier achieved the fastest time for building the training model.

التخصصات الرئيسية

تكنولوجيا المعلومات وعلم الحاسوب

عدد الصفحات

78

قائمة المحتويات

• APA
• MLA
• AMA

لغة النص

الإنجليزية

نوع البيانات

رسائل جامعية

رقم السجل

BIM-729778