IT security, with a focus on data protection and cryptography, used to be my daily job when I worked for Oxford Computer Group in the ’90s, and, after that, when I started Project Botticelli. I moved to data mining in the mid-2000’s. I remember presenting a few sessions at Microsoft TechEd and other conferences in which I suggested that applied machine learning (then called data mining) could help security. I theorised that a system could monitor itself by continually updating a machine learned model to detect outliers in its own behaviour. I suggested it could take action and autonomously prevent some attacks if the probability of a live event being an outlier was sufficiently high. Nowadays, this would be recognised as security AI! Needless to say, the repercussions are both thrilling and frightening.

Perhaps thankfully, none of that happened in those days—at least, not amongst my usual clients. I think it is time to revisit those ideas—carefully. We have more understanding of how to handle false positives, which are still a major issue, and, above all, we have better data. We also have more of it, which is not necessarily a good thing, and more work needs to be done to filter and transform it. We even have more anomaly detection algorithms than in the past, although algorithm is never the magic bullet: clear goals and good data matter far more.

I am happy I will be working again with my former colleagues from Oxford Computer Training, who specialise in Identity Management and Security, especially on the Microsoft Azure platform. Together, we are offering you a free video in The Future Series and a follow-up webinar. I look forward to working with Hugh Simpson-Wells, Founder and CEO of Oxford Computer Training, who also happens to be my former, nice boss :), and whose suggestions have helped me over the years.

In this new video, Hugh and I discuss the applicability of machine learning for making us more secure. I respect his thoughts and instincts about security, very much. Watch it here:

• The Future Series: Machine Learning for Security Applications

Once you have watched the video, please join us for a free webinar on 2 October 2019 in which Hugh and I will discuss the subject in more detail, and we will also give you a chance to ask questions live. Hopefully, it will make the subject relevant to your job and to your interests.

• [Updated 04 Oct 2019] Webinar recording is here: Machine Learning for IT Security: From ML to Security AI

See you in October!

Rafal