Institutional Repository of Vilnius University
Vartotojo elgesio pokyčiais paremtas įsilaužimų nustatymas /
Abstract
dc:descriptionAfter the system breach followed by user compromise, the user’s behavior changes and deviates from normal. Defined user’s behavior can be used as an identifier for particular user. Goal of this paper is to evaluate if artificial neural network (ANN) can be used to detect intrusion based on changes in user’s behavior. To achieve this goal, processes data of two users are collected and refined to represent user behavior. And then used for ANN to learn upon. Main emphasis of this paper is on the impact that data manipulations has on the end results of ANN. Experiments showed that not using any arguments, and only taking the process name gives be-st results when combined with not grouping processes at all. This combination showed consistent above 90 % success rate on ANN guessing which user’s data is presented. To imitate attack, commands usually used during the attack has been injected into the data set. Recording success results with this modified data and comparing them with normal day behavior showed, that ANN success rate drops drastically when the attack happens, which in turn can be treated as alert for the intrusion.Fast breach detection allows to minimize or neutralize the damages of an attacker.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2021
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Žeižys, Paulius,
- Contributors dc:contributor
-
- Radavičius, Irmantas
Rights
dc:rights- Statement dc:rights
-
- info:eu-repo/semantics/openAccess
- Language dc:language
- lit
Identifiers
dc:identifier.*- Repository record dc:identifier
- https://repository.vu.lt/VU:ELABAETD107116676&prefLang=en_US