Research: 
Connectivity and coordinated behavior in complex and encrypted systems, using machine learning and network theory. Developing models to advance security, resilience, and predictive analytics across domains like cyber-security, finance, and biology.

Interview with Dr Somin:

Dr. Somin's research focuses on identifying overt and covert connections between people, even when the information is partial, fragmented, or encrypted. She develops tools based on machine learning and temporal network analysis that are used to detect coordination and assist in the fields of cyber, finance, and preventing the spread of fake news.

The events of recent months have made it clear how important it is to be able to identify terrorist activity early, in all its branches. However, identifying terrorist cells is a complex operation: branches for recruiting activists, for example, will not operate on the same platforms as branches for raising funds or military branches. Furthermore, in the name of caution, it is very possible that the activists have received instructions not to talk to each other at all on social networks or other means of communication. So how can we detect an organization preparing for hostile activity in advance, before the attack is carried out? "This approach is called proactive cyber, and its goal is to detect hostile activity at an early stage "The planning, long before the attack is carried out," says Dr. Shahar Somin. "This is one of the issues I dealt with during my postdoctoral research: the question of how to uncover the hidden coordination mechanisms that lie behind connections between people, even though they are not directly visible in the data."
Dr. Somin joined the Faculty of Engineering in April 2025. She is 38 years old, married, and a mother of three children. She holds a bachelor's degree in mathematics and computer science and a master's degree in computer science, with a focus on machine learning. Upon completing her master's degree, she went into industry and worked for five years at a startup called Andor, where she worked as a data scientist and head of the machine learning department. At the same time, she began her doctoral studies in industrial engineering and management at Tel Aviv University. "My goal was to apply my knowledge of machine learning to the field of relationships between people," she explains. "Classical machine learning often focuses on the personal characteristics of each individual - such as name, height, marital status, area of residence, salary - and tries to predict their behavior based on this data alone. But people's behavior is influenced no less by their relationships with others. For example, when we tried to predict the spread of the coronavirus, the most important question was not whether I was married or how many children I had – but who I came into contact with. Likewise, if all my friends switched to a certain bank – the chances increase that I will also switch to that bank. And if all my friends joined ISIS – the chances increase that I also have such an orientation. There is a network of influences here in which the structure of connections is just as important as personal characteristics, and this was the focus of my doctoral research.
After graduating, she continued her postdoctoral work at the MIT Media Lab, where she took the research one step further, and tried to uncover hidden mechanisms between people, even when they are not directly reflected in the data. "The world is spread across a very wide variety of platforms, and there is hardly any body that has access to all of these platforms to draw conclusions from all of the wealth of information, not to mention the fact that there are factors that would intentionally hide these connections. My goal was to try to locate connections between different people who operate across different platforms, and who seemingly have no connection in the data, and to find indications that they are acting in coordination for a common goal – for example, that they are part of the same organization or are engaged in planning hostile activity."