NORMA eResearch @NCI Library

A Deep Learning Visual Content Based Recommender System to Defend Adversarial Attacks

-, Komal (2022) A Deep Learning Visual Content Based Recommender System to Defend Adversarial Attacks. Masters thesis, Dublin, National College of Ireland.

[thumbnail of Master of Science]
PDF (Master of Science)
Download (2MB) | Preview
[thumbnail of Configuration manual]
PDF (Configuration manual)
Download (2MB) | Preview


Recommender System is a type of machine learning technique which is used to produce significant recommendations to a group of users based on their preferences in the past. Deep Neural Networks have proven to be a wonderful fit while being used to deploy recommender systems, however the challenge with deep neural networks is that they are vulnerable to Adversarial Attacks, according to recent studies. The main goal of this research is to demonstrate adversarial attack defenses for visual content-based recommender system using deep learning. This study consisted of the examination of vulnerability of visual content-based recommender system against different targeted adversarial attacks using state-of-the-art white-box adversarial attack techniques and different Adversarial Training defense mechanisms to make our recommender system more robust against these executed attacks. A DeepFashion dataset used in this study which is a combination of 800,000 labelled images of clothes. For evaluation success rate metric was used in this research. Results of our experiments showed that from Fast Gradient Sign Method, Projected Gradient Descent and Carlini & Wagner methods, PGD with 128 iterations and CW attacks had the highest success rate. And traditional Adversarial Training defense method made system more robust compared to Curriculum Adversarial Training method. This proposed study helped in understanding the positive impact of defense mechanism on the adversarial attacked model and encouraged to train our recommender systems against these attacks in advance.

Item Type: Thesis (Masters)
Subjects: Q Science > QA Mathematics > Electronic computers. Computer science
T Technology > T Technology (General) > Information Technology > Electronic computers. Computer science
Q Science > QA Mathematics > Computer software > Computer Security
T Technology > T Technology (General) > Information Technology > Computer software > Computer Security
H Social Sciences > HF Commerce > Marketing > Consumer Behaviour
Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Machine learning
Divisions: School of Computing > Master of Science in Data Analytics
Depositing User: Tamara Malone
Date Deposited: 21 Feb 2023 15:28
Last Modified: 02 Mar 2023 09:43

Actions (login required)

View Item View Item