After demonstrating unprecedented success in Computer Vision first and more recently in Natural Language Processing (NLP), Artificial Intelligence (AI) is quickly gaining traction in the Cyber Security arena as well. The aim is to provide rapid detection of novel attacks. Such attacks are more and more directed at the user via the web (and not at the infrastructure anymore). Being devoid of stringent controls that regulate their security, the web (including social media, messaging applications, and all online communication channels), in fact, represents the preferred means for cybercriminals to carry out attacks. According to the IBM Cost of a Data Breach Report 202310, the global average cost of a data breach in 2023 was $4.45 million, 15% more than in 2020. The global indicator ‘Estimated Cost of Cybercrime’ in the Cyber Security market is forecast to continuously increase to reach 13.82 trillion U.S. dollars in 202811.  In response, 51% of organizations plan to increase Cyber Security spending this year.  

Phishing and stolen or compromised credentials were the two most prevalent attack vectors in 2023 report, with more than 60% of data breaches globally are caused by attacks that use the web at large as a vector. Moreover, 74% of attacks12 also exploit the human element (via Phishing and Social Engineering of various forms). 

In fact, although attack techniques and protection technologies continue to evolve, the human remains the weak point common to all organizations, who to date is responsible for 95% of successful cyber-attacks. We therefore face a context where in almost all cases a cyber-attack is successful due to a lack of awareness, negligence, distraction, or due to human error. With billions of online people to protect, we need to provide technical solutions that cut short the time of detection of new attacks. In fact, attacks that use the web as a vector and are directed at humans today represent a constant threat on any device, in any network. Existing protection systems are often ineffective, as they are based on slow, poorly scalable detection techniques that do not consider all communication channels into account with which the user interacts.  

With nearly 11 months (328 days) to identify and contain data breaches resulting from stolen or compromised credentials, we need solutions that prevent attacks rather than mitigate them afterwards. 

The AI4CTI project aims to explore, design, develop and implement new Artificial Intelligence technologies applied to the field of Cyber Security — and web security specifically — to provide rapid protection against cyber-attacks in their very initial phase, to effectively and timely prevent them. We provide the methodologies, the tools, and the information necessary to conduct security activities and investigations. In detail, the objective of AI4CTI is to study, develop and implement new approaches to keep end-users protected when they access the web or use apps on any device they use, be it a mobile phone, laptop or remote desktop. The key idea is to provide timely detection of new threats as soon as these become active on the internet, leveraging continuous data collection and CTI feeds, and providing AI models that can detect a new threat even before it becomes alive and widespread.