Utjecaj homomorfne kriptografije na sustav prepoznavanja slika zasnovan na strojnom učenju

master's thesis

• Title (english): The influence of homomorphic cryptography on machine learning tool for image recognition
• Author: Mária Krajčí
• Mentor: Marin Golub (mentor)
• Committee member: Marin Golub (predsjednik povjerenstva)
• Committee member: Ante Đerek (član povjerenstva)
• Committee member: Vlado Sruk (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2023-09-20, Croatia

Abstract

U ovom radu proučavan je utjecaj homomorfne enkripcije na učinkovitost modela strojnog učenja. Korišten je pristup kao i u projektu CryptoNets kako bi se istražila učinkovitost na različitim skupovima podataka. Rezultati su pokazali da primjena homomorfne enkripcije nije značajno utjecala na rezultate modela. Iako je primijećeno blago smanjenje točnosti, rezultati su ostali zadovoljavajući. Iz dobivenih rezultata zaključuje se da je moguće upotrijebiti homomorfnu enkripciju u modelima strojnog učenja uz zadržavanje zadovoljavajuće učinkovitosti. Daljnje istraživanje treba biti usmjereno na optimizaciju algoritama homomorfne enkripcije radi postizanja bolje ravnoteže između sigurnosti i učinkovitosti te na ubrzavanje rada algoritma.

Abstract (english)

In this study, we investigated the impact of homomorphic encryption on the performance of machine learning models. The approach used was similar to the CryptoNets project, aiming to evaluate performance across different datasets. The results demonstrated that the application of homomorphic encryption did not significantly affect the model's performance. Although a slight decrease in accuracy was observed, the results remained satisfactory. Therefore, this work highlights the potential of utilizing homomorphic encryption in machine learning models while maintaining satisfactory performance. Further research should focus on optimizing homomorphic encryption algorithms to achieve a better balance between security and performance, as well as improving the computational efficiency of the algorithm.

• Keywords: homomorfna enkripcija; CryptoNets; neuronska mreža
• Keywords (english): homomorphic encryption; CryptoNets; neural network
• Language: croatian
• URN:NBN: urn:nbn:hr:168:338847
• 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: 2023-10-11 09:23:49