AI plays critical role in network security, according to BT boffin

Artificial intelligence (AI) is going to play a critical role in network security in the coming years and is already helping BT defend its infrastructure.

Ben Azvine, the Global Head of Security Research & Innovation at BT, has been at the heart of cutting-edge network security developments at BT for several years and has helped develop a cybersecurity strategy that combines AI-enabled visualization of cybersecurity threats with highly-trained network security personnel. He shared some of his thoughts on the matter with attendees at this week’s Broadband World Forum event.

“We are taking AI and making it help humans to be better… We are more about the Iron Man version of AI than the Terminator version,” he said, sparking ludicrous cinematic pitch ideas in the minds of some of his audience (I mean, Alien vs Predator sort of worked, right?).

Azvine pointed out that with the number of connected devices growing rapidly, old ways of securing assets were no longer relevant: Now, companies (including network operators) need to think about having a cybersecurity strategy comprising three steps – prevention, detection/prediction and response. The response needs to be much quicker than in the past (hours, not days) while the detection/prediction is tough to do without sophisticated analytics and AI algorithms.

What BT is doing is a great example of analytics and AI in action in the communications networking sector, rather than AI as a marketing hype machine — see ‘Why BT’s Security Chief Is Attacking His Own Network’ for more details.

But security is just one of seven key telecom AI use cases, as identified in a recent report, Artificial Intelligence for Telecommunications Applications, from research house Tractica (a sister company to Telecoms.com).

That report identified the seven main use cases as:

1) Network operations monitoring and management

2) Predictive maintenance

3) Fraud mitigation

4) Cybersecurity

5) Customer service and marketing virtual digital assistants (or ‘bots’)

6) Intelligent CRM systems

7) Customer Experience Management.

“The low hanging fruit seems to be chat bots to augment call center workers,” said Heavy Reading Senior Analyst James Crawshaw, who will be one of the expert moderators digging deeper into the use of AI tools by telcos during Light Reading’s upcoming ‘Software-Defined Operations & the Autonomous Network’ event.

“The more challenging stuff is making use of machine learning in network management. That’s still a science project for most operators — Verizon’s Matt Tegerdine was pretty frank about that in his recent interview with Light Reading. (See Verizon: Vendor AI Not Ready for Prime Time).

That analysis from the Verizon executive shows it’s still early days for the application of machine learning in production communications networks. And, as Crawshaw noted, AI is not a magic wand and can’t be applied to anything and everything. “It can be applied to the same things you would apply other branches of mathematics to, such as statistics. But it’s only worth using if it brings some advantage over simpler techniques. You need to have clean data and a clear question you are seeking to answer — you can’t just invoke machine learning to magically making everything good,” adds the analyst, bringing a Harry Potter element to the proceedings.

So what should network operators be ding to take advantage of AI capabilities? BT appears to have set a good example by hiring experts, investing in R&D, applying AI tools in a very focused way (on its cybersecurity processes) and combining the resulting processes with human intelligence and know-how.   “You don’t need to recruit an army of data scientists to take advantage of machine learning,” said Crawshaw. “Nor should you remain totally reliant on third parties. Develop a core team of experts and then get business analysts to leverage their expertise into the wider organisation.”