Detection of Injection Attacks in Web Applications

Mehmet Serhan Erçin; Esra N. Yolaçan

Article Details

Detection of Injection Attacks in Web Applications

Authors:

Mehmet Serhan Erçin | Esra N. Yolaçan

Publisher:

Eskişehir Osmangazi University

Main Topics:

Bilgisayar Yazılımı | Yazılım Mühendisliği (Diğer)

Journal:

Eskişehir Türk Dünyası Uygulama ve Araştırma Merkezi Bilişim Dergisi Vol.5 Issue 1, pp.1-11

Document Type:

Article

Publication Date:

6/2024

Document Language:

Turkish

Document ID (Org No.):

2001310028

pISSN:

2687-606X

Keywords:

Abstract:

 The attacks that we can bring together on Injection subject whose frequency of occurence is increasing day by day are more preferable for attackers due to their destructive effects and easy applicability. Today, web applications and linked framework structures are great services that we use frequently and that touch our lives at many points. Therefore, it keeps the interest of cyber attackers alive and motivates them to discover new methods. Due to the changing and developing services and applications via emerging technologies that general topics often needs reconsidering as sub topics and new solutions should be found accordingly. Generalization is an intense need in injection type attacks unlike some other intrusion methods. Hereby, the generalization ability of machine learning methods despite signature-based systems will provide significant advantages in the detection of injections. Therefore, machine learning methods which have much more ability of making generalization rather than signature-based systems will provide significant advantages in the detection of injections. This study is focused on input-driven web attacks, differentiating benign and malicious traffic, time performance on processes and accurate classification of outputs. With Random Forest and Decision Tree classifiers, 94.54 and 94.61 hit rates were obtained, and 15 and 12 sec. learning time performance was measured.

 The attacks that we can bring together on Injection subject whose frequency of occurence is increasing day by day are more preferable for attackers due to their destructive effects and easy applicability. Today, web applications and linked framework structures are great services that we use frequently and that touch our lives at many points. Therefore, it keeps the interest of cyber attackers alive and motivates them to discover new methods. Due to the changing and developing services and applications via emerging technologies that general topics often needs reconsidering as sub topics and new solutions should be found accordingly. Generalization is an intense need in injection type attacks unlike some other intrusion methods. Hereby, the generalization ability of machine learning methods despite signature-based systems will provide significant advantages in the detection of injections. Therefore, machine learning methods which have much more ability of making generalization rather than signature-based systems will provide significant advantages in the detection of injections. This study is focused on input-driven web attacks, differentiating benign and malicious traffic, time performance on processes and accurate classification of outputs. With Random Forest and Decision Tree classifiers, 94.54 and 94.61 hit rates were obtained, and 15 and 12 sec. learning time performance was measured.

 The attacks that we can bring together on Injection subject whose frequency of occurence is increasing day by day are more preferable for attackers due to their destructive effects and easy applicability. Today, web applications and linked framework structures are great services that we use frequently and that touch our lives at many points. Therefore, it keeps the interest of cyber attackers alive and motivates them to discover new methods. Due to the changing and developing services and applications via emerging technologies that general topics often needs reconsidering as sub topics and new solutions should be found accordingly. Generalization is an intense need in injection type attacks unlike some other intrusion methods. Hereby, the generalization ability of machine learning methods despite signature-based systems will provide significant advantages in the detection of injections. Therefore, machine learning methods which have much more ability of making generalization rather than signature-based systems will provide significant advantages in the detection of injections. This study is focused on input-driven web attacks, differentiating benign and malicious traffic, time performance on processes and accurate classification of outputs. With Random Forest and Decision Tree classifiers, 94.54 and 94.61 hit rates were obtained, and 15 and 12 sec. learning time performance was measured.

 Enjeksiyon üst başlığında toplayabileceğimiz saldırılar, yıkıcı etkilerinden ve kolay uygulanabilirliklerinden dolayı saldırganlar tarafından daha çok tercih edilmekte, rastlanma sıklıkları her geçen gün artmaktadır. Günümüzde, web uygulamaları ve bağlantılı çerçeve yapıları, sıklıkla kullandığımız ve hayatımıza pek çok noktada dokunan, büyük hizmetlerdir. Bu yüzden siber saldırganların ilgisini sürekli canlı tutmakta ve yeni yöntemler keşfetmeye motive etmektedir. Sızma tespiti ve önlenmesi üzerine literatürde pek çok çalışma bulunmaktadır. Genel başlıklarda değerlendirilen bu çözümlerin, değişen ve gelişen uygulamalardan dolayı, alt başlıklarda ve ayrıntılı değerlendirilmesi ve buna uygun yeni çözümlerin bulunması gerekmektedir. Enjeksiyon tipi saldırılarda, girdilerin içerisindeki hedef sistem rezerve kelimeleri hariç tutulursa, kullanılan diğer harf ve rakamsal kombinasyonların sayısı sınırsızdır. Bu nedenle imza tabanlı sistemler yerine makine öğrenmesi yöntemlerinin genelleştirme performansı enjeksiyonların tespitinde önemli avantajlar sağlayacaktır. Bu çalışmada özellikle web enjeksiyon saldırılarına ilişkin saldırının doğru tespit edilmesinin yanı sıra, zamansal performans ve çıktıların sınıflandırılması da esas alınmaktadır. Rassal Orman ve Karar Ağacı sınıflandırıcılarında 94,54 ve 94,61 isabet oranları elde edilmiş, 15 ve 12 sn. öğrenme süreleri performansı ölçülmüştür.

Article Details

Detection of Injection Attacks in Web Applications

Authors:

Mehmet Serhan Erçin | Esra N. Yolaçan

Publisher:

Eskişehir Osmangazi University

Main Topics:

Bilgisayar Yazılımı | Yazılım Mühendisliği (Diğer)

Journal:

Eskişehir Türk Dünyası Uygulama ve Araştırma Merkezi Bilişim Dergisi Vol.5 Issue 1, pp.1-11

Document Type:

Article

Publication Date:

6/2024

Document Language:

Turkish

Document ID (Org No.):

2001310028

pISSN:

2687-606X

Injection

Machine Learning

SQLi

XSS

CMDi

CVEEnjeksiyon

Makine Öğrenmesi

SQLi

XSS

CMDi

CVE

Issues

Article Details

Authors: Mehmet Serhan Erçin | Esra N. Yolaçan

Publisher: Eskişehir Osmangazi University

Main Topics: Bilgisayar Yazılımı | Yazılım Mühendisliği (Diğer)

Journal: Eskişehir Türk Dünyası Uygulama ve Araştırma Merkezi Bilişim Dergisi Vol.5 Issue 1, pp.1-11

Document Type: Article

Publication Date: 6/2024

Document Language: Turkish

Document ID (Org No.): 2001310028

pISSN: 2687-606X

Issues

Authors:

Mehmet Serhan Erçin | Esra N. Yolaçan

Publisher:

Eskişehir Osmangazi University

Main Topics:

Bilgisayar Yazılımı | Yazılım Mühendisliği (Diğer)

Journal:

Eskişehir Türk Dünyası Uygulama ve Araştırma Merkezi Bilişim Dergisi Vol.5 Issue 1, pp.1-11

Document Type:

Article

Publication Date:

6/2024

Document Language:

Turkish

Document ID (Org No.):

2001310028

pISSN:

2687-606X