A new cybersecurity system traps malicious software inside isolated digital containers, preventing it from spreading across a computer network even if it evades initial detection. Current antivirus software often fails against sophisticated phishing emails and zero-day attacks—new exploits that have never been seen before. Many solutions also slow down computers or generate false alarms that overwhelm security teams. This project combines two defences: a patented machine-learning model that analyses email content for suspicious patterns before delivery, and micro-virtualization that seals each application in its own container. Inside those containers, artificial intelligence monitors system calls in real time. If malicious code does execute, the container stops it from moving laterally to other parts of the system. If successful, this approach could replace resource-intensive antivirus software, cutting licensing costs while maintaining performance. It would protect endpoints—the laptops, phones, and servers that connect to corporate networks—against ransomware and unknown threats without requiring constant updates. For organisations handling sensitive data, such as hospitals or financial firms, this means fewer system-wide infections and reduced downtime. The system is designed to be scalable, so it could work across large networks without slowing them down.
View original technical description
Our project presents an innovative Endpoint Detection and Response (EDR) solution that enhances protective measures against spam, phishing, and various other threats by employing a dual-layered strategy. At its foundation, the system utilises a patented machine learning technology derived from UK Patent Application No. 2410759.1 to identify phishing emails at both the mail server and endpoint levels. This approach markedly improves accuracy by reducing the incidence of false positives and false negatives that are prevalent among current industry leaders. This sophisticated machine learning model conducts a thorough analysis of email content, proficiently detecting suspicious patterns and malicious intent prior to their delivery to end users. In addition to this, the solution utilises micro-virtualization by encapsulating each application within isolated containers. Within these environments, artificial intelligence models consistently monitor and filter system calls to detect and mitigate threats in real time. In the circumstance that a fraudulent email containing malicious content evades the preliminary detection mechanisms, the containerised environment guarantees that any executable code is effectively isolated, thereby inhibiting lateral movement and preventing a system-wide infection. This comprehensive containment strategy not only diminishes the risk of ransomware and zero-day exploits but also obviates the need for conventional, resource-intensive antivirus software, thereby decreasing licensing expenses and maintaining optimal system performance. By amalgamating these two defensive mechanisms, our solution attains a proactive and holistic security posture that is both scalable and efficient. The integration of advanced machine learning techniques for early threat detection, coupled with the strategic isolation of potential threats, guarantees the security of endpoints even in the face of sophisticated attacks. This approach establishes a new standard in cybersecurity by providing robust protection while maintaining minimal performance overhead.
Plain English summaries and category classifications on this site are generated by AI and may not perfectly reflect the original research.
Is something wrong? Let us know