Primjena algoritama strojnog učenja u svrhu detekcije netipičnih vrijednosti mrežnih podataka

master's thesis

• Title (english): Application of machine learning algorithms for detection of atypical network data values
• Author: Andreja Vodopija
• Mentor: Mihaela Vranić (mentor)
• Committee member: Mihaela Vranić (član povjerenstva)
• Committee member: Boris Vrdoljak (član povjerenstva)
• Committee member: Damir Pintar (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2021-07-15, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

Razvojem Interneta je došlo do potrebe za zaštitom u virtualnom svijetu jer su korisnici interneta svakodnevno izloženi zlonamjernim napadima. Analizom mrežnog prometa mogu se pronaći netipične vrijednosti u mrežnim paketima koje najčešće upućuju na neku vrstu napada. Pronalazak tih netipičnih vrijednosti se provodi detekcijom anomalija koja se može implementirati raznim algoritmima strojnog učenja. Nakon uvoda u strojno učenje i algoritme, analizirat će se rezultati algoritama logističke regresije, stroja potpornih vektora, slučajnih šuma, stabla odlučivanja, Gaussovog naivnog Bayesa, umjetnih neuronskih mreža i algoritma k srednjih vrijednosti i k najbližih susjeda. Za svaki od algoritama se pronalazi onaj sa najboljim hiperparametrima te se rezultati grafički prikazuju i objašnjavaju. Za učenje i testiranje modela koristi se NSL-KDD skup podataka koji je prethodno pretprocesiran.

Abstract (english)

With the development of the Internet, a need for protection in the virtual world arose. Internet users are exposed to malicious attacks on daily basis and analysis of network traffic can find atypical values in network packets that most often indicate a malicious attack. The discovery of these atypical values is carried out by anomaly detection that can be implemented by various machine learning algorithms. After an introduction to machine learning and algorithms, the results of logistic regression, support vector machine, random forests, decision tree, Gaussian naive Bayes, artificial neural networks and k means and k nearest neighbors algorithm will be analyzed. For each of the algorithms, a model with the best hyperparameters is found, and results are graphically displayed and explained. For training and testing, the model uses an NSL-KDD data set that is previously preprocessed.

• Keywords: detekcija anomalija; strojno učenje; NSL-KDD; pretprocesiranje podataka; dos; probe; r2l; u2r
• Keywords (english): anomaly detection; machine learning; NSL-KDD; data preprocessing; dos; probe; r2l; u2r
• Language: croatian
• URN:NBN: urn:nbn:hr:168:214485
• Study programme: Title: Computing; Study programme type: university; Study level: graduate; Academic / professional title: magistar/magistra inženjer/inženjerka računarstva (magistar/magistra inženjer/inženjerka računarstva)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Closed access
• Created on: 2023-01-13 14:12:14