Optimizirana arhitektura klasifikatora temeljenog na umjetnim neuronskim mrežama u domeni implementacijskih napada na kriptografske uređaje

master's thesis

• Title (english): Optimized classification architecture based on artificial neural networks in the domain of implementation attacks on cryptographic devices
• Author: Juraj Juričić
• Mentor: Domagoj Jakobović (mentor)
• Committee member: Domagoj Jakobović (predsjednik povjerenstva)
• Committee member: Marin Golub (član povjerenstva)
• Committee member: Vlado Sruk (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2019-07-04, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

U diplomskom radu opisan je postupak optimizacije arhitekture klasifikatora temeljenog na umjetnim neuronskim mrežama u domeni implementacijskih napada na kriptografske uređaje koristeći tehnike evolucijskog računarstva. Objašnjen je implementacijski napad diferencijalnom analizom potrošnje električne energije kriptografskog uređaja te primjena metoda strojnog učenja u istome.

Implementirani su algoritam NEAT te dvije varijante genetskog algoritma - generacijski genetski algoritam i genetski algoritam sa segregacijom po vrstama. Dobiveni rezultati uspoređeni su s rezultatima prikazanim u Picek i dr. Algoritam NEAT se nije pokazao uspješnim na problemima klasifikacije, dok su obje varijante genetskog algoritma davale rezultate usporedive s prijašnjim radovima u području.

Abstract (english)

The thesis describes approaches for optimising the architectures of artificial neural network classifiers in the domain of side-channel cryptographic attacks using evolutionary computation techniques. The utilisation of machine learning methods in the differential power analysis attack is explained, and the attack itself is described.

NEAT and two variations of the genetic algorithm - a generational genetic algorithm and genetic algorithm with species segregation - are implemented and their performance is benchmarked. The results are compared with the results logged in Picek et al. NEAT's performance at classification is subpar at best, while using both of the genetic algorithms yielded results comparable to previous research in the field.

• Keywords: kriptografija; DPA; SCA; strojno učenje; evolucijsko računarstvo; neuroevolucija; genetski algoritam
• Keywords (english): Cryptography; DPA; SCA; Machine Learning; Evolutionary Computation; Neuroevolution; Genetic Algorithm
• Language: croatian
• URN:NBN: urn:nbn:hr:168:445405
• 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: 2020-01-07 18:50:30