Intercast February 2025 – Cybersecurity & Machine Learning

Welcome to the February 2026 edition of Intercast’s monthly newsletter for cybersecurity professionals. As always, we’ll bring you the latest news and views to bring you up to speed.

In This Issue:

  • Cybersecurity & Machine Learning
  • Machine Learning Limited By ‘Averaging’ Flaw
  • Ransomware Negotiator Admits Ransomware Scams
  • Army Makes AI and Machine Learning A Dedicated Career Path
  • Nearly Half Of All Firms Outsource Security
  • Machine Learning Predicts When Machines Will Break
  • Best of the Rest

Client Insight

Each month we ask our clients what’s on their minds to help us get a broader perspective on the industry. This month we’ve had a lot of conversations about the overlap between cybersecurity and machine learning.

As you may have read in our recent 2025 Impact Report or spotted in our newsletters, we’ve expanded our focus to cover machine learning and AI roles alongside our existing cybersecurity specialism. That’s prompted some good-hearted debate about the similarities and differences between the two fields.

For example, one analogy had cybersecurity pros being a defensive line preventing damage, with machine learning pros the more creative and productive offense squad. (We presume that means those whose expertise covers both areas are truly “special teams”!)

Others argued that the skills are transferable: both require creative thinking and an understanding of the benefits and limitations of automating core tasks. It seems the debate will continue…

Machine Learning Limited By ‘Averaging’ Flaw

MIT researchers say applying existing machine learning models to new types of data may be less effective than first thought. That’s because poor results are hidden by averaging.

They cited several cases where a model trained on a large amount of data was put into a new setting and compared with other models. In this scenario, testing often involves identifying the model that performs the best as an overall average.

However, the researchers say that when you break the dataset down, that same model is often the worst for any individual piece of data. In one case, the best average model for diagnosing chest x-rays was also the worst for 75 percent of the individual x-rays.

One possible explanation is that the models are so well-trained on the types of data they’ve already encountered that it drags up their average and hides their consistently poor performance on new types of data. Another is that adding the new data exposes the model using “spurious correlations” that weren’t problematic with previous datasets. Either way, the takeaway is that models need broader testing when applied to new datasets.

 

Ransomware Negotiator Admits Ransomware Scams

In a case of gamekeeper turned poacher, a ransomware negotiator has admitted planting ransomware. He and a security incident response manager both pled guilty to an extortion conspiracy charge, admitting to working with an unnamed third party.

Former incident response supervisor Ryan Goldberg and ransomware negotiator Kevin Martin used a “ransomware as a service” package called ALPHV BlackCat against multiple victims. One victim paid them the equivalent of $1.2 million in Bitcoin, of which 20% went in commission to the ransomware creators.

The Justice Department said the men “used their sophisticated cybersecurity training and experience to commit ransomware attacks — the very type of crime that they should have been working to stop.”

Army Makes AI and Machine Learning A Dedicated Career Path

The US army is launching a dedicated “career field” in artificial intelligence and machine learning. It will mean officers can specialize in the subject, getting hands-on experience alongside their regular military training and progression.

It will be a new “area of concentration” option which teaches officers specific skills that can be used in a military context. A similar option for “Army Cyber” already exists.

The Army recently hired four senior execs from AI and machine learning companies to lead a new detachment that “will work on targeted projects to help guide rapid and scalable tech solutions to complex problems.”

Nearly Half Of All Firms Outsource Security

Security threats are the top concern of most businesses, outranking even economic downturns or supply chain issues. That’s according to a survey of senior executives which also found 43 percent of businesses are now outsourcing some of their cybersecurity services.

Rimini Street surveyed senior executives in multiple countries and found 54 percent said cybersecurity threats were their biggest concern. The category was at or near the top of the list across every country and industry.

As well as the 43 percent who said they already outsource cybersecurity, another 46 percent said they are “considering in the near future.” That’s partly caused by difficulties finding and retaining internal staff teams. Some respondents cited a vicious circle where understaffed teams are forced to spend more time on system maintenance, leaving them with less capacity for incident response and boosting defenses.

Machine Learning Predicts When Machines Will Break

Researchers at Stellenbosch University are using machine learning to better detect when industrial equipment is degrading. It could mean a much better balance of cost and risk.

Many manufacturers currently use one of two maintenance models. A fixed maintenance model works on a scheduled timetable, which risks unnecessary expenditure on machines that are working well. Meanwhile, the “break-fix” model of waiting until a machine fails before repairing it can mean increased costs and delays getting production back up and running.

The researchers believe their “predictive maintenance model” offers a middle ground. It uses Internet of Things sensors to spot patterns that are reliable indications of impending breakages. For example, some machines will start vibrating differently when motors or pumps begin to suffer from wear and tear. (Or as The Conversation wonderfully described it, “Machines whisper before they scream.”)

Best of the Rest

Here’s our round up of what else you need to know:

Back to Blog