Drevesa za napovedno razvrščanje v kibernetski varnosti : diplomsko delo visokošolskega študijskega programa Informacijska varnost

Uršič, Črt

| | SLO | ENG | Cookies and privacy

Bigger font | Smaller font

First page > Show document

Show document

Title:	Drevesa za napovedno razvrščanje v kibernetski varnosti : diplomsko delo visokošolskega študijskega programa Informacijska varnost
Authors:	ID Uršič, Črt (Author) ID Mihelič, Anže (Mentor) More about this mentor...
Files:	VS_Ursic_Crt_2024.pdf (1,95 MB) MD5: 72A53C3966E3DA712B26E0AF3B035C43
Language:	Slovenian
Work type:	Bachelor thesis/paper
Typology:	2.11 - Undergraduate Thesis
Organization:	FVV - Faculty of Criminal Justice and Security
Abstract:	Glavni cilj raziskave je oceniti, kako učinkovito lahko algoritem dreves za napovedno razvrščanje identificira in klasificira različne vrste kibernetskih napadov. Strojno učenje, ki je osrednji del sodobnih IT-okolij, vključno s sistemi za zaznavanje in preprečevanje napadov, se nenehno razvija in izboljšuje. V tem kontekstu smo izbrali algoritem dreves za napovedno razvrščanje (PCT) zaradi njegove predhodne uspešnosti v večrazrednih klasifikacijskih nalogah. Za primerjavo učinkovitosti PCT smo vzpostavili konkurenčno skupino enajstih drugih algoritmov, katerih rezultati služijo kot referenčna osnova za oceno. Da bi zagotovili natančnost in relevantnost rezultatov, smo razvili lastno metodologijo priprave podatkov in modelov. Prvi korak naših meritev je zajemal odstranjevanje neveljavnih in manjkajočih vrednosti v podatkovni zbirki. Nadaljevali smo z določanjem optimalnega števila lastnosti, izbirali smo samo tiste, ki imajo ključen vpliv na končni izid. V zaključni fazi smo nastavili parametre algoritmov in njihove vrednosti, ki smo jih optimizirali z uporabo mrežnega iskanja. Vse te strategije so omogočile zmanjšanje potrebnih računalniških virov, skrajšale čas učenja in procesiranja podatkov ter izboljšale končne rezultate. Ta pristop nam omogoča boljši vpogled v dejansko učinkovitost algoritmov. Rezultati naše študije kažejo, da čeprav izbrani algoritem PCT ni izstopal kot najboljši med primerjanimi algoritmi, kaže obetaven potencial za nadaljnje raziskave in praktično uporabo v sektorju kibernetske varnosti. Diplomsko delo tako ne le opredeljuje učinkovitost PCT, ampak tudi predlaga možne smeri za nadaljnji razvoj in izboljšave v tehnologijah zaznavanja kibernetskih groženj.
Keywords:	drevesa za napovedno razvrščanje, kibernetska varnost, strojno učenje, diplomske naloge
Place of publishing:	Ljubljana
Place of performance:	Ljubljana
Publisher:	Č. Uršič
Year of publishing:	2024
Year of performance:	2024
Number of pages:	IX f., [49] str.
PID:	20.500.12556/DKUM-89406
UDC:	004.056(043.2)
COBISS.SI-ID:	203474947
Publication date in DKUM:	02.08.2024
Views:	433
Downloads:	55
Metadata:
Categories:	FVV
:	Copy citation

Average score:	(0 votes)
Your score:	Voting is allowed only for logged in users.
Share:

Hover the mouse pointer over a document title to show the abstract or click on the title to get all document metadata.

Licences

License:	CC BY-NC-ND 4.0, Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International

Link:	http://creativecommons.org/licenses/by-nc-nd/4.0/
Description:	The most restrictive Creative Commons license. This only allows people to download and share the work for no commercial gain and for no other purposes.
Licensing start date:	08.07.2024

Secondary language

Language:	English
Title:	Predictive clustering trees in cybersecurity
Abstract:	The main goal of the study is to assess how effectively a predictive clustering tree algorithm can identify and classify different types of cyber attacks. Machine learning, which is a central part of modern IT environments, including systems for detecting and preventing attacks, is continually evolving and improving. In this context, we selected the predictive clustering tree (PCT) algorithm because of its prior success in multi-class classification tasks. To compare the effectiveness of the PCT, we established a competitive group of eleven other algorithms, whose results serve as a reference basis for assessment. To ensure the accuracy and relevance of the results, we developed our own methodology for preparing data and models. The first step in our measurements involved removing invalid and missing values from the data set. We then proceeded to determine the optimal number of features, selecting only those that have a key impact on the final outcome. In the final phase, we set the parameters of the algorithms and their values, which we optimized using grid search. All these strategies enabled the reduction of required computing resources, shortened the learning and data processing time, and improved the final results. This approach allows us a better insight into the actual effectiveness of the algorithms. The results of our study show that although the selected PCT algorithm did not stand out as the best among the compared algorithms, it shows promising potential for further research and practical application in the cyber security sector. The thesis not only defines the effectiveness of PCT but also suggests possible directions for further development and improvements in cyber threat detection technologies.
Keywords:	predictive clustering trees, cybersecurity, machine learning

Comments

Leave comment

You must log in to leave a comment.

Comments (0)

0 - 0 / 0

There are no comments!

Back