Penetration Testing in the Age of Artificial Intelligence

The world as we know it is rapidly being impacted by A.I.-driven technology. It was only a decade ago when smartphones came to prominence, and now, the A.I. landscape is slowly hogging all the spotlight.

Markets rise and fall based on predictive algorithms, smart homes perform menial tasks based on people’s behavior and self-driving cars drive more accurately by the day, among other uses.

These innovations have been made possible by increased internet speeds, stronger computing hardware, and the rise of technologies like Edge and Cloud Computing.

However, all of the benefits come with equivalent hazards. Now that our personal data is situated in the cloud, it’s more vulnerable than ever to theft.

That’s why it’s not surprising that a heavy emphasis on cybersecurity and secure server practices have been strictly adhered to in recent years.

But, before we dive right in to how A.I. fits in the whole cybersecurity puzzle, we first need to discuss the individual concepts in their current state.

White Hat, Black Hat

Cybersecurity, as a whole, encompasses a wide range of aspects from the hardware-level all the way to the social-level. It serves as a direct response to the malicious practice commonly known as hacking.

While hacking itself is an even more general term, ranging from phishing scams to malware attacks, hackers themselves aren’t all bad.

“White hat” hackers, for instance, use the same toolkit and adhere to the same practices as their more hostile counterparts, but their intent leans towards the improvement of security rather than breaking it.

Penetration testing or (Pen test), otherwise known as Ethical Hacking or White Hat Hacking, is conducted by white hat hackers to combat the threat of malicious hackers, commonly known as “black hats.”

Pen testing is an authorized simulated cyberattack on a computer system with the main goal of detecting vulnerabilities and weaknesses of a system.

The whole process is an end-to-end test that starts from gathering the necessary information all the way to reporting all of the detected weak spots.

Contrary to popular belief, penetration testing doesn’t just involve hardware and software components, it also employs social engineering tactics to weed out weak employees.

White hat hackers who conduct social engineering penetration testing do this by deceiving employees into giving out sensitive data or perform actions that will create security weaknesses that allow the hackers slip through. 

Automate Everything

Unfortunately for white hat hackers, their black hat counterparts are up-to-date on the latest cutting-edge technologies themselves.

A.I. is being used as drones for large-scale bot networks (or botnets) to enact massive Distributed Denial-of-Service attacks, among many other illicit activities.

In order to keep up with the pervasive threat, white hat hackers must be willing to keep up and adapt to the ever-changing landscape.

But, as with other industries that are automatable, there is a looming question of whether or not Artificial Intelligence would eventually replace the human aspect of penetration testing.

To answer that, we first need to examine the current state of A.I and how it could supplant the need for manual intervention.

Even though science fiction novels and movies have led us to believe that A.I. would be so advanced in the present day that they could easily pass as humans, unfortunately that’s not the case.

Chatbots have made great strides over the last decade, but they have a long way to go before they effectively mimic the way humans speak.

It’s an underrated aspect but one that could make or break the social engineering part of spoofing potential attack vectors.

On the plus side, A.I. can sift through a hundred thousand lines of data in a matter of seconds. Even the search parameters doesn’t have to be extensive since they can adjust on-the-fly.

Throw in a good Optical Character Recognition (OCR) plugin and they could easily have the ability to read text in pictures and handwritten notes.

Unlike humans that get tired, A.I. can run 24/7 non-stop. Plus, they could easily multiply and are highly extensible. You wouldn’t need to pay them for their services as well.

All of these are highly relevant to the information gathering process when performing penetration testing.

However, given the unpredictability of human emotions and the probability of human error, tactics would need to be adjusted on-the-spot given whatever context presents itself.

This is something that A.I. could eventually learn to analyze, but currently their models need to be trained further to cater to the inherent randomness.

Given a proper setup, A.I. can seamlessly interface to different systems and it can follow its protocol exactly as it was designed – leaving none to minimum margins of error.

A.I. decision making might be rigid (to an extent), but it is objectively infallible, especially compared to humans. Not to mention, it can generate extremely detailed reports in a blink of an eye.

Humans are prone to error, lapses in judgment and, depending on the sensitivity of the information to be handled, hard to trust than technology that can be programmed or designed to “learn” new information.

These glaring weaknesses can be easily bypassed through the use of A.I. So, why hasn’t A.I. fully taken over this whole process yet?

Man and Machine Working Together

Even though A.I. has come a long way in the past two decades, it still has a long way to go before it can fully take over the different types of penetration testing processes.

Despite the growing stack of benefits, the biggest argument against handing complete control of penetration testing to A.I. is its reliability.

While the A.I. can follow a set of instructions, it can easily be exploited by hackers that are prepared to take on the automated defense system.

Here’s an example of how hackers can carry out their attack on AI-based cybersecurity systems. 

Machine learning – an application of A.I. – learns and gets “smarter” by observing patterns found in data, and making assumptions about its meaning – whether on a large neural network or on individual computers. 

So if a certain action within computer processors occurs at the same time that particular processes are running, plus, the action gets repeated on the specific computer or neural network, the system will learn that the action means that a cyber-attack is happening. 

This also prompts the system that the necessary actions need to be taken to address the attack.

The tricky part though is that A.I. – savvy malware, for instance, can insert false data for the security to read – the goal of which is to disrupt the patterns that machine learning algorithms utilize to make their decisions.

This means that fake data could be injected into a database to make it seem like a process that’s copying sensitive information is part of the regular IT system routine, and therefore, can just be ignored.   

Thus, A.I.-centric approaches might be the future decades down the line, but for the time being, human-led pen testing still remains as the go-to for many prominent companies.

But, that doesn’t discount what A.I. can do for pen testing today. While it may not be a viable alternative to give it autonomy, pen testers could still leverage A.I. as a tool to aid their practices.

As mentioned earlier, A.I.-supported information gathering can help ease the burden of having to sift through piles of information. That would leave human pen testers more time to focus on other aspects.

White hat bots can be employed to combat malicious bots, and automated sniffers can be used to detect fraudulent sites before they can do any significant damage.

Reports can be automatically generated and steps can be easily documented with the help of automated tools.

A.I. – powered tools are also used to “look at” rendered web pages to determine the ones that most likely have actionable leads. 

The current penetration testers’ method are to do this task manually – which can take up a lot of time since they have to check each screenshot one at a time. 

With the latest AI technology, however, and deep neural networks, performing this task – visually inspecting web-pages – can now be done through an automated process.  

There has never been an easier time to get into penetration testing. If you’re interested to start a career, you can read up on tech articles to help you get started on your journey.

Don’t feel too pressured that you need to catch up quickly with the latest trends. There’s a lot of ground to cover and you would need a lot of time to practice.

What’s next?

While it might be tempting to go for a DIY approach when it comes to protecting your site against cyber attacks — especially since there are a lot of available tools out there in the market — you might do yourself more harm than good.

For the most part, it is a good practice to work with reliable cyber security companies that do penetration testing since they have specialists who work on cybersecurity day in and day out.

With the help of experts running pen tests on your network, your network security is bound to get rid of its security gaps.

(3)

Share

Article Link: http://digitalforensicsmagazine.com/blogs/?p=3039