The IT Sorcerer’s Apprentice!
Machine learning seemed an odd fit at first. Our company was formed as a simple network discovery tool, as reliable and useful as a carpenter’s hammer. If you don’t know us at Who’s On My WiFi, we started off by offering a platform-agnostic ARP scanning tool to discover connected devices on a network over time. Our company path changed drastically when we started saving all the information we were scanning. We made an important discovery: WWW.WhoIsOnMyWiFi.com
Large amounts of network data is useless without some way to make
sense of all that information!
For example, the first problem people tried to solve using our software was detecting if an unknown device suddenly joined the network.
We initially required that customers tag devices as KNOWN, and then they could be alerted to any UNKNOWN devices. But there is a problem with this, especially on larger networks. Tagging devices is time-consuming and requires constant updating to be useful. Our customers’ IT managers would be tasked with tagging staff and network devices, while reporting on guests that entered their building. It was an up-front workload compounded by the inevitable influx of new devices or [shudders] network equipment overhauls.
The next problem people started solving with our basic network detection was trying to determine the number of people using a public WiFi network over time. Although this sounds simple, to get accurate usage patterns, again, there is an up front cost of going through and tagging all equipment that could possibly be on a public WiFi network to exclude it. Otherwise, always on devices like network equipment or printers incorrectly impacted the results. And what about employees using the public WiFi? Should they be counted as visitors or not?
To painstakingly go through a large public venue, tag all switches, APs, as well as employee equipment and smartphones was too much maintenance for IT administrators to keep up with.
Enter Machine Learning.
Machine learning is great at putting items into categories.
To a computer, the task of painstakingly annotating every active IP address, MAC Address and subnet, and comparing it’s current usage to it’s usage patterns over time or a standard usage template is no more difficult than running any other task. Machine learning that categorizes network devices into network equipment vs employees vs visitors or notices usage anomalies is trivial. Trying to accomplish this same task as a human in IT takes a lot of human hours, effort, and cost.
What we are finding is that in this day and age, data is plentiful. It’s insight from the data that’s scarce. Modern Access Point manufacturers like Meraki, Aerohive, and Aruba offer the option for IT Engineers to export virtually all network data. What you do with that data depends on the money and time in your organization or the needs of your industry or specific problems.
As decision makers recognize the value in analyzing the data, the sorcerer has become ruled by the apprentice as IT professionals are purchasing network hardware based on that hardware’s ability to analyze the network data or it’s ability to export to an analytics package. You can look at the newest interfaces of Meraki, Aerohive, Ubiquiti, etc. to get a sense of this.
Network Analytics and Machine Learning in Action
But if the hardware manufacturers are already starting to integrate some of these tools, then what is the point of knowing about Machine Learning as an IT Professional and how it’s changing things? For one thing, Machine Learning based self-improving defense algorithms and uptime algorithms are learning to protect the network themselves. Hardware maintenance and bandwidth issues are being analyzed and prioritized by a machine before an entry-level tech sleepily clocks-in on a Monday morning. It’s hugely beneficial for us, because the mundane work becomes automated while the nitty-gritty technical implementation and human analysis is left up to the masters.
But keeping the network up and running safely is starting to become an expectation, and there are other uses for analyzing network data. For example, many businesses live and die by the number of customers who visit their location each day. They build their revenue models off of these numbers, yet estimate the average dwell time and return rates of their customers. WiFi provides a wealth of information about visitors to a location and can count guests who don’t even connect to a network. The IT admin can only get this information if ALL network data is exportable.
The future of IT administration is a bright one. Machine learning promises to be a tool that empowers the skilled Network Engineer to not only protect a network, but grow the business. So what’s the takeaway for you as the network administrator?
Pay attention to your network hardware’s ability to export data for 3rd party analysis.
No single hardware system will be able to do the level of analysis that is going to be required to stay competitive in the near future. As machines learn to share data with one another, you don’t want your organization to be at a disadvantage. By safely and securely being able to export your data for third-party analysis, you can improve network security and add value to your organization.
Author: John Kerber Is the Technical Co-Founder of Who’s On My WiFi. www.whoisonmywifi.com. John is a lifelong computer enthusiast with over 20 years experience in IT, Development, Network Security, and Management. John is a geek in the truest sense of the word with an eye for market opportunities. With a degree in computer science and over 5 years experience as the CEO of a startup, he brings a technical mind to business and end user solutions.
***Technical Editor Comment - I have known John for several years and John is one of the Good Guys and always welcomed at LoveMyTool .com! I have used Johns software for years and have recommended to hundreds. It is totally free and you will see all the action on your WiFi ...simple to use and understand the information. Protect your WiFi! I have it on every computer I have!