Primjena strojnog učenja za prepoznavanje CAPTCHA-e

master's thesis

• Title (english): CAPTCHA recognition with machine learning
• Author: Ana Kovaček
• Mentor: Marija Seder (mentor)
• Committee member: Marija Seder (član povjerenstva)
• Committee member: Ivan Petrović (član povjerenstva)
• Committee member: Juraj Peršić (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2022-07-12, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

CAPTCHA je skraćenica za potpuno automatizirani javni Turingov test koji služi za razlikovanje računala od ljudi. Koristi se u svrhu smanjivanja korištenja web stranica od strane računala. Neka od situacija u kojima se koristi je ograničavanje registracije od strane računala u svrhu sprječavanja stvaranja lažnih računa što smanjuje trošenje resursa usluga i mogućnosti prijevare. Za potrebe izrade rada korišten je skup podataka naziva CIFAR-10. Taj skup podataka najčešće se koristi za treniranje programa strojnog učenja u području računalnog vida. Skup sadrži 60 000 slika koje su podijeljene u 10 klasa. Cilj je s neuronskim mrežama oponašati ljudski mozak na način da traži značajke objekta iz slike. Konvolucijska neuronska mreža izdvaja odnosno identificira značajku određenog objekta na slici što čini jedan od temelja strojnog učenja za otkrivanje ili klasifikaciju objekata. Za odabir slika razvijena je funkcionalnost koja omogućava korisniku da pomoću svoje ruke upravlja mišem te na taj način odabire slike.

Abstract (english)

CAPTCHA stands for fully automated public Turing test to distinguish computers from humans. It is used to reduce the use of websites by computers. One of the situations in which it is used is to restrict registration by computers to prevent the creation of fake accounts which reduces the consumption of service resources and the possibility of fraud. A set of data called CIFAR-10 was used while working on this thesis. This dataset is most commonly used to train machine learning programs in the field of computer vision. The set contains 60,000 images divided into 10 classes. The goal of neural networks is to mimic the human brain by searching for the features of an object from an image. The convolutional neural network extracts or identifies the feature of a particular object in the image, which is one of the foundations of machine learning for the detection or classification of objects. Functionality that has been developed for image selection, allows the user to control the mouse with their hand and thus select images.

• Keywords: CAPTCHA; strojno učenje; računalni vid; neuronska mreža; detekcija objekta; klasifikacija objekta
• Keywords (english): CAPTCHA; machine learning; computer vision; neural network; object detection; object classification
• Language: croatian
• URN:NBN: urn:nbn:hr:168:802727
• 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: 2023-01-13 10:11:40