Povzetek
Namen prispevka: Ob neprestanem izboljševanju informacijske infrastrukture so za njeno zaščito potrebni tudi novi pristopi k varnosti in razvoj novih tehnik zaznavanja kibernetskih groženj. Med te tehnologije sodijo tudi nevronske mreže, ki se že dolgo uporabljajo na različnih področjih, kot so medicina, logistika in biologija. Namen prispevka je prepoznati in predstaviti njihovo uporabnost na področju kibernetske varnosti. Metode: Izveden je bil sistematični pregled literature, s katerim so bile prepoznane pozitivne in negativne lastnosti nevronskih mrež kot aktualnega pristopa strojnega učenja za zaznavo kibernetskih napadov. Izvedena je bila tudi primerjava uporabe nevronskih mrež s konvencionalnimi sistemi. Ugotovitve: Rezultati eksperimentov so večinoma v korist nevronskim mrežam, saj je proces hitrejši, natančnejši in z manj lažnimi alarmi kot konvencionalni sistemi, ki običajno delujejo na principu statične analize. Vendar pa so nevronske mreže zaradi njihovega načina delovanja pogosto nepredvidljive in so najbolj učinkovite šele v kombinaciji s konvencionalnimi sistemi. V obstoječi literaturi primanjkujejo predvsem testiranja teh sistemov v realnih situacijah, izven kontroliranih umetnih okolij. Omejitve/uporabnost raziskave V pregled literature so bili vključeni znanstveni prispevki, objavljeni v letih od 2017 do 2019 in indeksirani v bazah Web of Science in Scopus. Izvirnost/pomembnost prispevka: Prispevek s povzetkom osnovnih tehničnih principov delovanja nevronskih mrež predstavlja začetno točko za strokovnjake na področju kibernetske varnosti, ki z njimi še niso seznanjeni. Prispevek povzema trenutno stanje na področju uporabe nevronskih mrež v kibernetski varnosti in potencialne smeri razvoja v prihodnosti. Prispevek predstavlja enega prvih sistematičnih pregledov literature na področju uporabe nevronskih mrež v kibernetski varnosti, ki se je znanstvenoraziskovalno razcvetelo predvsem v zadnjih treh letih.
Ključne besede
nevronske mreže;strojno učenje;umetna inteligenca;sistem za zaznavanje vdorov;škodljiva programska oprema;kibernetska varnost;
Podatki
Jezik: |
Slovenski jezik |
Leto izida: |
2020 |
Tipologija: |
1.02 - Pregledni znanstveni članek |
Organizacija: |
UM FVV - Fakulteta za varnostne vede |
UDK: |
004.056+004.032.26 |
COBISS: |
30703875
|
ISSN: |
1580-0253 |
Matična publikacija: |
Varstvoslovje
|
Št. ogledov: |
523 |
Št. prenosov: |
22 |
Ocena: |
0 (0 glasov) |
Metapodatki: |
|
Ostali podatki
Sekundarni jezik: |
Angleški jezik |
Sekundarni naslov: |
Use of neural networks in cybersecurity |
Sekundarni povzetek: |
Purpose:
In addition to the continuous improvement of the information infrastructure, new approaches to security and the development of new techniques for detecting cyber threats are needed to protect it. These technologies also include neural networks which have long been used in various fields such as medicine, logistics and biology. The purpose of this paper is to identify and present their applicability in the field of cybersecurity.
Design/Methods/Approach:
A systematic review of the literature was performed to identify the positive and negative properties of neural networks as a current approach to machine learning for the detection of cyberattacks. A comparison of the use of neural networks with conventional systems was also done.
Findings:
The results of the experiments are mostly in favour of neural networks, as the process is faster, more accurate and with fewer false positives than conventional systems which are typically based on static analysis. However, neural networks are often unpredictable due to their mode of operation and are most effective only in combination with conventional systems. The existing literature predominantly lacks testing these systems in real-world situations outside of controlled artificial environments.
Research Limitations / Implications:
The review of the literature included scientific papers published in the years from 2017 to 2019 and indexed in the Web of Science and Scopus databases.
Practical Implications:
The paper recapitulates the basic technical principles of neural networks and is a starting point for cybersecurity experts who are not yet familiar with them. The paper summarizes the current situation in the use of neural networks in cybersecurity and potential directions for future development.
Originality/Value:
The paper presents one of the first systematic reviews of the literature in the field of the use of neural networks in cybersecurity which has flourished scientifically in the last three years.
|
Sekundarne ključne besede: |
neural networks;machine learning;artificial intelligence;intrucion detection systems;malware;cybersecurity; |
Vrsta dela (COBISS): |
Znanstveno delo |
Strani: |
str. 197-210 |
Letnik: |
ǂLetn. ǂ22 |
Zvezek: |
ǂšt. ǂ2 |
Čas izdaje: |
2020 |
ID: |
12055963 |