Ju Yong Lee, Innopolis University - on user reputation, measuring trust and data security in social networks

Ju Young Lee— Head of the Laboratory of Network Sciences and Information Technologies at Innopolis University, Associate Professor.

Received her PhD from Syracuse University(USA). Has experience in a university environment and industry (JPMorgan, Sun Microsystems). Professor Lee develops algorithms for determining the activity and reputation of social network users. Her research focuses on distributed computing reputation and node degree estimation in a social network.

Reputation: virtual and real

Reputation is the perception of the user by other people.In real life this happens everywhere.Whether you suspect it or not, when you first meet a person you have certain expectations or premonitions about him. Then, as you talk or interact, you update this view. All these factors go into reputation.

Reputation is not only about people, but also about companies, government organizations and any structures with which we interact.I don't use social networks often and I only knowmost popular: Facebook and Instagram. In them, people communicate using comments and photographs. The concept of reputation is not widely used on these popular platforms because most of the interactions there are interpersonal. On Facebook, we do not open a bank account, make transactions, or manage finances. In a broad sense, we don't need a lot of trust to communicate on these platforms because if our interlocutor turns out to be a bad person, we don't have much to lose.

The reputational system is mainly used on commercial sites such as Amazon and eBay in the USA or Ozon and Yandex.Market in Russia.This requires trust between the seller andbuyer. If you pay money and the goods are not sent, you lose something. Therefore, reputation and trust are important for economic reasons. Typically these platforms use a centralized reputation system, which you're probably familiar with. For example, Amazon has a five-star rating system where you can write to the seller after receiving the item and give it a rating. This is one of the most common reputation systems. It is centralized because it takes control of the algorithm. If you give three stars, and then someone else gives four or five, the system will automatically average these values ​​and show the result to other users. This type of assessment is very effective for commercial sites.

There are also decentralized or distributed reputation systems.They are used where there is no centralauthority, for example, in peer-to-peer file sharing systems. There is no authority here that will tell you about the reputation of another user. But everyone has their own idea about this or that person. I may have interacted with a user who didn't send me the file or had a very slow connection. And another could have a very pleasant experience of interacting with him. So we have different ideas about this person and his reputation is not the arithmetic average of our opinions.

I'm trying to figure out which algorithms are used in popular networks.For example, Stack Overflow or Quora (popularquestion and answer systems - "High-Tech"). These are where you ask a question and then people answer it. If an answer becomes more popular, the author increases his reputation. This is a general scheme, but each platform uses different algorithms. Typically, their scheme is based on feedback between user interactions: if the interaction was successful, the reputation increases, otherwise it decreases.

We recently developed an algorithm that is completely independent of this type of feedback.The only thing we use is time andinteraction frequency. How often we interact really affects reputation. It does not matter whether it is good communication or not, the main thing is its number and frequency. We compared our development with the standard Stack Overflow algorithm on a single data set. It turned out that their results are very similar. This means that without knowledge of the perception of the user by other participants, one can quite accurately determine his reputation.

AI, fake accounts and agitation

Recently, social networks have become important because of the huge number of registered users.Now political parties and commercialorganizations have a new channel of influence on all these people. Previously, if a company wanted to promote a product, it would use TV commercials or banners on the side of the road. But since social networks became popular, everyone has realized that advertising on them is much more effective. Marketing within a social network is cheaper, and information spreads much faster. Now it’s enough to create a small piece of news instead of buying an expensive TV spot.

The main role in the popularization of social platforms was played by the younger generation.When I was younger, I hated reading newspapers andother press. On social networks everything is different. If someone says that something interesting is happening, the rumor will spread very quickly. That’s why social networks have replaced traditional newspapers and magazines.

Now AI can create a fake account and conduct its own activity.Since social platforms have become importantinformation medium, they are used to influence people for political purposes. With their help, parties try to lure people to their side. To do this, they broadcast a selective part of the information that is beneficial to them. This is implemented using AI and learning algorithms. The important thing here is to choose the right group of people. If you want to be voted for, you need to choose the right person. They should not be your ardent opponent or supporter. This may have the opposite effect. It is necessary to choose people “on the border” who can change their minds. And, of course, such people are easy to identify based on their interests and interactions on social networks.

Surveillance in social networks, Korea and the vulnerability of personal data

On large media and social platforms it is impossible to avoid information leaks.Because, first of all, you put it there.Second, even if big companies like Facebook say they won't share your information, their long and confusing user agreement that no one actually reads will allow that to happen.

The main problem in the scandal with Cambridge Analytica is that the analytic company gained access to friends and all friends' data.They developed an application that determinedpolitical views of users. Facebook was supposed to provide information only about people who used it. Although even if the leak did not take place, there are many algorithms that would make it possible to draw conclusions about the entire network based on data about part of it.

The transfer of information to third parties is not something new or rare.Facebook is a company, not an NPO, and it needsget a profit. To do this, you have to share some of the information with other companies. How exactly she does this, how much and what information she gives out, no one knows. So the user must be aware of this danger from the very beginning. Everything related to social platforms is new. Therefore, the government does not know how to react to such scandals. Appropriate policies have yet to be developed. Right now, the only one who can protect your data is you. I don't want to say that you shouldn't use social networks. They are very convenient and allow you to make new friends. But it will not be possible to completely prohibit the transfer of information to third parties - this will happen constantly. Therefore, we need the right approach to this type of situation - not at the individual level, but at the government level.

In Korea, there is a very popular KakaoTalk messenger.Many thought that the state was looking throughcorrespondence history in it to find out how people feel about the current government. At that time, some of my friends refused to use certain words. They distorted names and words to avoid observation. I think in many countries the government actually does this. Of course, without admitting it publicly, since it is illegal. However, can the government effectively use this information? If you accuse someone based on a photo from a private chat, you must admit that you were illegally monitoring that person. We never know what is really going on inside government structures. At the individual level there is no way to change the situation.

Having enough information you can find out everything about a person, if you need it.The same case with Cambridge Analytica:They just developed an app - a small psychological test that makes inferences about personality based on a few questions. This allowed them to determine the political background of respondents and use this in the election race. Knowing where you live and what university you went to can determine your entire cultural background. Even with a small amount of information - who you talk to, who your friends are - you can infer so many things. This is why user information on social media has become critical.