Artificial Intelligence and Machine Learning (AI/ ML) – Your Friend or Theirs?

At the recently held Telecom Leadership Forum, 2022, a panel of industry experts discuss Artificial Intelligence (AI) & Cybersecurity- in the network. The intricacies, power and fears of AI/ ML in the security realm.

Artificial Intelligence, also known as AI is an ally, an adversary, a force multiplier, an offensive tactic – everything and anything today. It is being used for contextualised, faster and more powerful attacks. So should Telcos be worried, prepared or still be sitting on the fence when it comes to developing capabilities for security with AI? Some experts from the industry dissected how strong is the offensive and defensive side of AI, what constraints can data availability bring in, why modelling plays a key role in tapping AI’s potential and where AI can bring in real value.

Data –The Real lever of AI’s power

The panel – that was moderated by Anil Chopra, VP-Research and Consulting, CyberMedia Research, started with the question of how worrying the trend of offensive AI is?

“AI-ML is being used for defence as well as by adversaries. Of course, it is a worrying trend because the adversary’s models are strong – there are no boundary conditions for them. However, on the defence side, we are limited with the quantum of information- especially with data sets needed for the training of models. Quality and availability of information on the defender’s end are lesser than the ones present on the offender’s side. That puts us on the back foot. The context and use of AI would depend a lot on the boundary we operate in.” Mathan Babu Kasilingam, CISO, Vodafone Idea Ltd. explained.

Venkat Krishnapur, VP, Engineering, MD, Trellix India added some more layers here. “AI is not new. The ability of computers to match human capability has been in discussion for many years. What has made AI a reality today is the advent of computing power, assisted with the proliferation of Data and the emergence of Cloud. In the past, computer power drove output, today it is the other way around- with data driving compute power. The variety, velocity and veracity of data are tremendous today.

Where-How to use AI?

AI-ML can now be put to a lot of use-cases, as illustrated by Priya Kanduri, CTO, VP- Cyber Security Services, Happiest Minds Technologies. “Advanced areas like a honeypot for luring attackers, using simulated environments to analyse their strategies with sandboxed set-ups – all these can be, and should be, de facto security approaches. Similarly, gathering information and threat prevention can be strengthened as well as transaction-level fraud protection can be fortified with AI. Models can help customers with better protection. We work closely with data scientists to catch stock impersonation, account take-over etc. Wide usage of technology can help with context-aware authentication here.”

Abhish Kulkarni, Practice Manager Advisory and Professional Services (AI, Data Analytics), HPE India gave a perspective of accelerating innovation here. “We have three major areas of focus- hardware, software and services. On the hardware infrastructure, for instance, we work on solutions that can be custom-made and for accelerated training, for quickly churning maximum data to come up with models, and for fast inference. AI is an exploration and not a journey- because you will keep on learning as you go along. We also need to be aware of the anonymisation of data, the bias around data, the knowledge of what data can be shared and what not – which is where software helps. On the service side, we help customers to quickly start this path of ML and AI adoption and fast deployment of models.”

“Malware detection and behaviour analysis are good ways to use AI. Identifying malicious activity and predictive activity work can be done with specific algorithms.” Vishakh Raman, Director, Security Business, Cisco India & SAARC pointed out some specific use-cases which help Telco players when we think of AI for security.

The Flip Side of AI

All these possibilities are flanked with challenges and what-ifs too. Especially around the level of expertise to understand data-sets and to convert them for analytical models. And also on vertical-specific models to enhance usability. Looking from the shoes of a typical attack strategy can also help. Using internal approaches for understanding how the offensive side operates- would elevate security with AI. The industry also needs to stay up to speed with new technology- so that attackers do not have an advantage. Especially when attackers are using reverse engineering, using their models to understand defence-side models, and using technology to confuse with data manipulation. Threat modelling and attack simulation are some of these ways to think like criminals and change data where the exposure and vulnerability get high.

Can CIOs and CISOs leverage AI to fight back? Especially with the large attack surface that a Telco has?

That’s a question that is no more on the back-burner now. And it’s time to dial up attention and action here.