Paranoia or a real threat? Find out how the new BLOCK tool will help you recognize bots and social media manipulation.
Unfair actions on the Internet cause huge damage, estimated in 2019 at $78 billion. However, data scientists are actively working to solve this problem. Alexander Nwala, an assistant professor of data science at William & Mary, in collaboration with colleagues at Indiana University, recently published a paper in EPJ Data Science. It presents BLOCK — a universal language model for analyzing behavior in social networks.
BLOC, which stands for "Behavioral Languages for Online Characterization", uses two types of "alphabets" to characterize user actions: one for actions, the other for content. These "alphabets" can be used to create lowercase codes, which are then converted to vectors. These vectors are analyzed using machine learning algorithms, which allows you to identify potentially malicious actions.
The project not only makes it easier to detect automated bots, but also helps identify similarities between human-managed accounts. "If two accounts are doing pretty much the same thing, you can investigate their behavior using BLOC to see if they are controlled by the same person," Nwala said.
The University of William and Mary is already developing a BLOCK-based tool for researchers, journalists and the general public that will analyze suspicious activity on social networks.
Research in this area is complicated by the limitations that social networks impose on the use of software interfaces. Nvala emphasizes that such restrictions harm not only researchers, but also society as a whole, as such research helps to identify manipulations in social networks and form effective policies in this area.
Unfair actions on the Internet cause huge damage, estimated in 2019 at $78 billion. However, data scientists are actively working to solve this problem. Alexander Nwala, an assistant professor of data science at William & Mary, in collaboration with colleagues at Indiana University, recently published a paper in EPJ Data Science. It presents BLOCK — a universal language model for analyzing behavior in social networks.
BLOC, which stands for "Behavioral Languages for Online Characterization", uses two types of "alphabets" to characterize user actions: one for actions, the other for content. These "alphabets" can be used to create lowercase codes, which are then converted to vectors. These vectors are analyzed using machine learning algorithms, which allows you to identify potentially malicious actions.
The project not only makes it easier to detect automated bots, but also helps identify similarities between human-managed accounts. "If two accounts are doing pretty much the same thing, you can investigate their behavior using BLOC to see if they are controlled by the same person," Nwala said.
The University of William and Mary is already developing a BLOCK-based tool for researchers, journalists and the general public that will analyze suspicious activity on social networks.
Research in this area is complicated by the limitations that social networks impose on the use of software interfaces. Nvala emphasizes that such restrictions harm not only researchers, but also society as a whole, as such research helps to identify manipulations in social networks and form effective policies in this area.
