The Rise of Machine Learning Makes Our Networks Smarter

By Mark Verbloot, Systems Engineering Director, Asia Pacific, Japan
Share Post

Artificial intelligence (AI) and machine learning are rapidly becoming part of the fabric of our everyday lives. Take my iPhone for instance. I live and work in Sydney, and when I get into my car each morning as soon as my phone pairs to my car, it tells me how long the trip will be to the office. I didn't ask the phone but it has learned that I go from my home to the office on weekday mornings and I drive directly home in the evening. It uses several data sources such as the phone GPS, traffic data and the time of day to make this prediction. It seems amazing that my phone knows where I'm going and how long it will take to there.

It's pretty cool, but it's not always so smart. On some weekends I drive to another house about two hours out of Sydney. I only started doing this about 4 months ago and my phone has learned this is a new regular destination that I go to at a similar time each weekend so it assumes that I'm going to go there every Saturday morning when I get into my car but it is often wrong. It would be smarter if it had access to more data, such as my work and personal calendars. The more data machine-learning algorithms have, the better the results.

What is Machine Learning, Exactly?

AI and machine learning are big buzzwords these days. AI is the broader concept of machines being able to carry out tasks in a way that we would consider smart. Machine learning is an application of AI based on the idea that machines can access massive amounts of data and learn for themselves. In machine learning, the system automatically learns and improves from experience without being explicitly programmed.


AI and machine learning are being democratized, driven by the sheer scale and power of the cloud as well as enhanced algorithms that can tackle more ambitious problems. What was used for highly specialized applications like stock trades, fraud detection, and airport security, is making our homes smarter and boost business intelligence. Machine learning is also making a big impact on networks and cybersecurity.

Making Our Networks Smarter

It's no secret that networks are incredibly complex. Mobile environments have unpredictable connectivity patterns. With BYOD and IoT, there are more devices—and device types than ever—on the network. And people have high expectations for service, even for demanding applications like voice and video. Cybersecurity is also extremely complex, as threats go undetected and investigations take too long, further increasing organizational risk.

Machine learning can make our networks smarter and more automated. We are moving from an era where we rely on really smart network engineers and their toolsets to build, optimize and secure networks to an age where machines can help us with these complex tasks.

A lot of network optimization and troubleshooting is painstaking work. Some organizations have extremely talented network engineers. I have come across many over the years. These are the people we turn to when we have serious issues. Without naming any one individual to let's call one of these engineers Tom. Tom has been called in because no one seems to know what is happening with the network. Tom kicks off some packet captures,  logs into switches, and controllers, and starts entering numerous CLI commands. He examines the packet traces, command outputs and pours over numerous logs. Tom compiles all of this data into a spreadsheet and uses his years of experience and intuition to figure out what is going on and make recommendations on how to remedy the situation. The more customer sites that Tom goes to, the more he learns, and the recommendations are better and better.

What if we could put an extremely talented network engineer like Tom in a box and have him run this kind of analysis on 100's or 1000's of networks continuously?

Using Machine Learning to Improve the Network

With the introduction of Aruba IntroSpect, we can. The power of machine learning, coupled with the cloud, when applied to complex networks, can improve the user experience and stop cyberattacks faster.

IntroSpect leverages machine learning for cybersecurity. It can spot changes in user or device behavior that often indicate that inside attacks have evaded perimeter defenses. Machine learning-based analytics are used to predict the malicious intent of both individual users or devices such as IoT. It does this by first baselining their normal behavior by continuously monitoring their normal traffic patterns. Anything outside of this normal behavior may represent a security threat or even an attack and can be flagged accordingly. That enables security analysts to prioritize security risks based on entity risk scores and put key factors about an attack into context. The end result is that security teams can triage more effectively and respond before damage is done.

With the introduction of machine learning-based solutions, we are entering a new age of possibilities for smarter, automated networks. Through the power of software, we will have brilliant, always-learning network engineers and security analysts continuously watching over our networks and helping us to deliver a better experience to our users.

Learn more about IntroSpect.