We develop multi-level systems for managing critical infrastructure protection, as well as systems for securing key state services against kinetic and cyber threats. Our algorithms develop solutions successfully used in the United States to improve the security of airports, marine terminals, transmission infrastructure, and to increase the protection of endangered animal species in national parks in many places around the world. In addition, we create dedicated algorithmic solutions for detecting and preventing crime. We also develop artificial intelligence-based solutions that increase the effectiveness of unmanned platforms in security applications.
Protection of critical infrastructure
In the last two decades, thanks to the increase in computer computing power, it has become possible to directly model both many potential threats and possible defense measures, especially in the context of critical infrastructure protection.
Particularly popular here is an approach based on the Stackelberg game model, in which a critical infrastructure defender distributes limited security resources to protect a group of objects against various types of attack. Thanks to the use of innovative optimization methods, it is possible to design defense strategies that significantly increase the chance of preventing a potential attack, which has been repeatedly demonstrated in many practical applications.
In particular, Stackelberg’s games have been used to optimize the protection of selected facilities in the USA (Los Angeles LAX airport, ports in Boston and New York), national parks and nature reserves in the Americas, Africa and Asia to counteract opportunistic crimes, as well as in cybersecurity. Based on these experiences, our team creates advanced systems to protect critical infrastructure in Poland.
Data analysis for security
Data analysis plays a key role in increasing defense capabilities, detecting and preventing crimes. While there are many commercially available data analysis software packages for security applications, they do not cover the latest scientific developments, including those in artificial intelligence. This means that the time and work of analysts are not optimally used, which is particularly important in the context of growing volumes of available data. Therefore, there is an urgent need to build and implement dedicated solutions based on artificial intelligence for the security sector, which use the latest scientific achievements to analyze large amounts of data and enable increasing the capabilities and efficiency of experts’ work.
Our team specializes in developing efficient algorithms for security applications. We specialize in particular in solutions based on graph algorithms and machine learning algorithms. Our interests and experience include a wide range of applications: from modeling and detecting financial fraud to the use of artificial intelligence for analysis, forecasting and crime prevention. For example, by drawing on known forensics methods that allow for the identification of key variables, we can effectively train relatively accurate machine learning models to predict crimes in a given area. Then, using explainable machine learning techniques (explainable AI), we are able to determine the contribution of individual variables to the obtained result to obtain a much deeper understanding of the causes leading to the occurrence of various types of crimes in a given area.
We have initiated an important research direction: adversarial social network analysis. It will help you understand how individuals and groups can avoid the techniques, algorithms and other tools used in this analysis. Within this paradigm, we directly model the strategic behavior of both members of social networks and the party whose goal is to analyze such networks. This research has important practical implications. On the one hand, they can help protect privacy, for example by effectively hiding undisclosed social relationships from connection prediction algorithms. On the other hand, research can help understand and counter the stealth strategies used by criminal and terrorist organizations.
Unmanned platforms (drones)
Our team also develops advanced solutions for various types of unmanned vehicles based on machine learning and other AI technologies. Recently, we have been particularly interested in multimodal methods of human-operator communication with an unmanned vehicle or a swarm of such vehicles. For example, this research includes the creation of artificial intelligence-based mission planning software that can understand and implement operational objectives based on a simple verbal description.
AI SHIELD project
Stefan Dziembowski