A New Approach to Machine Learning Based Cybersecurity and Fraud Detection


Crossword Cybersecurity Plc (AIM:CCS, “Crossword”, the “Company” or the “Group”), the technology commercialisation company focused solely on cyber security and risk, has today announced the launch of Nixer CyberML, a new family of machine-learning based security and anti-fraud software products, that help organisations easily and quickly build these capabilities into applications. Nixer CyberML is a new tool for businesses that want to solve advanced security and cybercrime problems, such as detecting and dealing with compromised accounts, fraud and in-application denial of service attacks.

Many of today’s security and fraud problems occur within applications and are difficult, if not impossible, to detect externally to the applications. For example, if a fraudster has obtained a user’s login details via a credential attack, their access to the site while logging in can appear normal – but once inside the site, can start to behave maliciously. Nixer CyberML allows development teams to rapidly add machine learning based detection to online applications (online banking, ecommerce systems, ticket sites, critical business apps, etc.) that can learn to accurately distinguish between good and bad user behaviour.

This initial release designed for developers, includes the Nixer CyberML architecture, code libraries for Spring framework based applications, and a local Nixer CyberML Engine designed to help with credential protection functionality. The Nixer CyberML Engine, stores and processes anonymous application event data, and contains the machine learning algorithms which determine whether events are normal or potentially malicious. Future versions of Nixer CyberML will give developers access to cloud-based machine learning algorithms.

The benefits of using Nixer CyberML include:

• Detecting, mitigating and preventing fraud and other user-based attacks that Web Application Firewalls (WAFs) and Distributed Denial of Service (DDoS) tools cannot mitigate, by using advanced machine-learning integrated into business applications

• Quick and reliable deployment by using existing code and architecture, and, where needed, working with a specialist partner familiar with machine-learning, cyber security and cyber-crime prevention to provide support with data science and algorithm creation.

Jake Holloway, Crossword’s Chief Product Officer, said, “Machine Learning based approaches to cybersecurity are not uncommon in very advanced applications, but most development teams are maxed out and do not have the right data science skills and tools to implement effective solutions quickly. This is why Nixer CyberML has been designed – it allows organisations to build in security and anti-fraud capabilities easily and quickly, without needing to employ machine learning specialists.”

Tom Ilube, Crossword CEO also said, “The first product to come out of the Nixer CyberML family of products is part of Crossword’s CyberAI initiative. It will be followed up with further services to help Enterprises, SaaS Vendors and developers accelerate projects that add intelligence to their applications. As the UK’s leading cyber security commercialisation company we work with universities regularly, as we have with Imperial College London on Nixer CyberML’s machine learning algorithm design and selection.”

Jan Broniowski, Nixer CyberML Architect said, “By putting security closer to the application layer, we create rich, valuable context for algorithms and analytics. We offer Nixer CyberML as Java libraries through GitHub – a design decision that provides control, explainability and configurability to developers. Nixer CyberML helps to merge the gap between security analytics and software development.”
To find out more about Nixer CyberML, visit www.nixer.io or download the code libraries from Github.