You've probably heard of the phrase machine learning. It sounds complicated, but the name says it all; machine learning is a way of making computers (machines) learn. Machine learning is a form of AI, Artificial Intelligence. And to get a machine to learn something, you first need software that makes that possible. Of course, that software is created by a human. But the final result is super-smart software that keeps on getting smarter and smarter. In the end, this kind of software works much faster and more accurately than a human.
In fact, the more experience the software has, the better and more accurate it does its job! This makes Machine Learning ideal for outsmarting phishing for example. It will take your cyber security to the next level. We are happy to explain how machine learning works and how it can make your life easier.
To understand how machine learning works, just look at how you learned words as a child. For example, the word "dog". You were shown pictures of a variety of dog breeds. A Labrador, an English Cocker Spaniel, a Great Dane, you name it. After a while you had seen pictures of quite a few dog breeds. And with all of them you learned that they were pictures of a dog. Unconsciously you began to recognize patterns. Four legs, a furry coat, rather large and thin ears in all shapes and sizes, even the different shapes a dog's head can have, after a while you recognized them.
You may not know that this is a Hungarian Vizsla, but you are sure that this is a dog. All thanks to the labels you learned as a child.
At some point you would have seen enough different dogs that you knew even dog breeds that were new to you were dogs. A Hungarian Vizsla, Irish Wolfhound, Finnish Spitz? Never seen one, but definitely a dog! And that's how machine learning learns, too. By supplying lots of data, machine learning recognizes patterns by itself, and that software can determine for itself what something is, even if something is completely new! Only requirement is that when learning, there is more than enough data to learn from and that you know what that data actually is.
As you have read above, it is important to provide as much input or data as possible if you want to make software smart. The more data, the better. That data is used to make the software learn. Of course, you don't do that with just any data.
Do you know what exactly is in which jar? If you don't know and have yet to learn what something is, those tags or labels are indispensable!
That data you use for training machine learning is very well controlled and you know exactly what kind of data it is. Each piece of data has a tag describing exactly what kind of data it is. In the example about how you learned to recognize dogs as a child, the tag is "dog". In Smartlockr, that label is used for personal data, such as a phone number, address, patient number or a BSN. Anything goes. Thanks to the label, the smart software knows what kind of data it is and can learn from it, as long as there are enough examples.
With just a table full of tagged data, you're not there yet. You will need much more data than a simple table from excel. With ten different addresses, the software can never learn to recognize all types of addresses. If the software has to recognize an address without ever having seen it, it won't work properly if you haven't taught it what an address is.
Machine learning needs data. A lot of data.
The more data you use, the better the software learns what an address is. We're talking tens of thousands of sample addresses just to learn the concept of an address. That seems like a lot of work, but once enough data is entered, the software can eventually recognize and process a concept like addresses much better than a human can. But machine learning is more than just throwing a big list of data into the software. You need a programmer to check everything, adjust it and keep it on track.
Of course, it all starts with the programmer who writes the software that must be able to learn everything by itself. But even after that, the programmer's work is far from over. They do a lot more than just inputting data and then relaxing on the couch watching Netflix while the system does its work.
Even with machine learning, the programmer remains essential to making the software truly smart.
From the very first moment, the programmer watches very closely how the system processes the data and where things are or are not going well. This is because Machine Learning also makes mistakes during the learning process. Mistakes are part of the learning process, even in machine learning and just like the mistakes you made as a child.
But you do want to spot the mistakes in time. The programmer can then make the right turns to fix errors. And with each turn, the software gets better and better and so the results become more accurate. Until the software no longer makes mistakes.
The end result of this whole process is a system that understands what the information you provide means and what it can do with it. Most importantly, machine learning also processes new, unfamiliar data and assigns the right label to it. "Yes, I can do that too!", we hear you thinking. That's right. But machine learning can do two things much, much better than humans.
First of all, machine learning works super-fast. The data you can process on your own in a week is a chuckle to systems based on machine learning. These types of systems do the work that would take you a week to do in mere seconds, at most. So machine learning is very well suited to handle the ever-increasing stream of data that is created daily all over the world.
A mistake is easy to make; even the ancient Greeks suffered from it. Fortunately, today you can rely on the benefits of machine learning.
Second, machine learning makes no (or hardly any) mistakes. No matter how precise someone is and how carefully they try to work, mistakes come with the territory. Just think how quickly you are distracted at work. Or those times you hit the wrong button. Machine learning doesn't suffer from that. You can virtually eliminate human error with machine learning!
And, also convenient; machine learning never gets tired! While you're catching up with your colleagues on Monday morning about the weekend, the smart software just keeps working all weekend. Error-free and super fast.
At Smartlockr, we make grateful use of machine learning. This allows our system to recognize (personal) data even better and knows immediately what to do with it. You can send an email with peace of mind; machine learning keeps an eye on it and tells you when something needs to be secured or encrypted or if you still have unauthorized recipients in the cc. So your daily mail traffic becomes even more efficient!
Also useful is that we use machine learning to recognize phishing. And that's kind of necessary. Because billions (!) of phishing emails go around the world every day. Some are easy to recognize, but techniques like spear phishing make them better and better. And because it is becoming increasingly difficult for the user to recognize phishing easily and quickly, it is important to use the best high-tech tool for this. You guessed it; machine learning! Machine learning recognizes even the very latest phishing techniques and instantly alerts you to potential dangers. That way you can't even accidentally click on a wrong link.
It's an open door if you've already read this entire blog, but machine learning makes your working life a whole lot nicer. You don't have to worry about secure mail in any form. Smartlockr’s system encrypts your emails and attachments when necessary and is always alert for phishing. This way, you can be sure that you are not the cause of a data breach!
Want to know more about machine learning and what it can do for your organization? Download the free Whitepaper now.