Machine Learning: Our Next Cyber Threat?

Hello! I just wrote an article on Machine Learning and Cybersecurity! This article is also available on Medium. Recommends are appreciated!

The Wannacry virus shut down thousands of computers, including part of the British Healthcare system. Imagine what could happen if hackers used machine learning, too.


Recently, there’s been a lot of cyber attacks. First the Mirai botnet, then the WannaCry ransomware, and then the mysterious Ukraine cyberattack that surfaced in June. We’ve generally accused freelancer hackers or Russian affiliates. But, what if there was a new kind of cyberattack—one that had very little human intervention?

A new way to think

Machine learning is a relatively new phenomenon where one presents an algorithm with a bunch of data, and the computer finds groups or does something else interesting with the data. Machine learning has been used for everything from email filtering to driverless cars.

Machine learning has exploded in popularity and is now starting to make it into the hands of consumers and hobbyists. A quick google search for ‘Machine learning algorithms’ pulls up thousands of results. Now, high-quality machine learning software is available online for free, prompting a college kid to make his Honda driverless.

But what about the bad guys?

But what if cyber-criminals used machine learning to their advantage? Instead of using tried-and-true methods of hacking, such as exploiting a loophole in system security, guessing passwords, or crafting a phishing email, what if cyber-criminals used machine learning to craft more realistic scams?

Phishing has been around since the 1990s, and has more-or-less stayed an effective scamming tool over the years. But, what if a machine learning algorithm found someone’s first name and email on a social media network? It could then craft realistic emails (with near-perfect english) and start a massive spear-phishing phenomenon. What’s worse is that it could measure the rate of people that responded to the email and create even more realistic messages.

Or what if a criminal developed a hacking algorithm, let it run for a bit, and sold it on the dark web? The customer could make thousands of copies of the same machine-learning hacking tool and could distribute them on a botnet. Each computer on the botnet would try hacking a random IP address, and a peer-to-peer botnet could share more and more information and become immensely stronger.

What about other countries?

The notion that most warfare will start as cyber-warfare is commonly believed, especially nowadays. Imagine if a large threat like Russia bought several high-quality, expensive machine-learning algorithms and pitted them against each other. That country could easily create an extremely dangerous algorithm, and they could duplicate it as much as they wanted.

What we need to do to prepare

This is obviously a huge threat, and we need to prepare. The face of cyber-warfare will change dramatically, so we need to make sure we’re on the same page as the bad guys. We need to take advantage of the fact that our government has lots of money and computing power. We need to develop more complex algorithms to find and patch vulnerabilities before anyone else does. We need to be aggressive in cyber-defense, and it should not be the first thing on most companies’ budget-cut list. Cyber warfare will be here to stay, so the least we can do is to be ready for it.

Leave a comment

Your email address will not be published. Required fields are marked *