Senne Batsleer is doing an Artificial Intelligence Master-after-Master degree at KULeuven. He is writing his thesis on the detection of fake webshops in the .be domain using machine learning. This concerns webshops that do not deliver sold goods or send fake products for example. 'I do this based on registration data and the content of the website, which are automatically recognised from the webshops' html code,' says Senne.
For the specialists: Senne does this using Positive Unlabeled learning or PU-learning for short. This means that we work with positively labelled data (websites that we know are fake webshops) and with unlabelled data (websites that we did not evaluate personally and of which we do not know whether they are fake webshops or not). ‘The method is not used very often and has never been applied to detect fake webshops', says Senne, who is positive about his research results. ‘It seems to work, but there are no guarantees.’
Before starting, Senne analysed the available documentation. Senne looked at how we dealt with fake web shops and how he wanted to change that. He knew DNS Belgium, but was surprised to see that we do so much in the field of cyber security. ‘I didn't know they looked for fake web shops so actively.'
‘Because everyone here is always on hand to provide assistance, I really learned to ask for help.'
Vladyslav Kozynin and Khava Djabrailova are both graduate students in programming at HOGent, who are doing their internship at DNS Belgium. Vladyslav studied the code of our various monitoring systems to improve them. Khava developed a user interface for a crawler created by DNS Belgium which collects public domain name data.’
All three agree this is not an ideal time for an internship in a company, as internships also have to be done from home. Neither of the three interns had any problems with that though. ‘I only went to the office on the first day but I have a lot of contact with my internship supervisor Maarten, who helps me a lot when I have questions', says Senne.
Vladyslav also experiences the efficient communication as a positive aspect of his internship. ‘I hear from friends that often there is very little communication with their internship places or supervisors. At DNS Belgium I found myself in a very warm and friendly team that is always willing to help.’
‘Because everyone here is always on hand to provide assistance, I really learned to ask for help,’ says Khava. For her it was very nice to work in a team, even remotely. ‘During our studies we mainly did individual assignments, whereas now we really worked in a team,' she says. 'Daily scrums, where everyone talks about what they have done, were completely new to me. It was very interesting to learn to work like that. In the beginning I didn't know what to say, but it's much better now.’
In addition to scrumming our students also gained some technical knowledge fortunately. Both Vlad and Khava concluded that their programming knowledge was up to date, but that HOGent did not talk much about networks. ‘What we learned at school is not enough for the work I do here', says Vlad.
On the other hand we also learn from our interns. We had already set up an internal study group to build more knowledge about artificial intelligence and machine learning. 'In terms of machine learning, DNS Belgium already had a detection system, but I may have been able to contribute something to the available knowledge,' says Senne. 'PU learning was certainly new to us', says Maarten Bosteels. 'We already had some experience with supervised learning and unsupervised learning. PU learning is kind of in between. In other words, it is definitely interesting to work with a KULeuven student and their supervisors on the university side.'
'I am developing my user interface with React and apparently that's the first time it's been used here,' says Khava. Maarten was interested in using it and I knew how it worked. Win-win in other words.