Page 179 - Cyber Defense eMagazine April 2023

of network breaches. Some of the most devastating attacks endured by this country’s corporations have
started with a single compromised password.

How Does ChatGPT Do Damage?

ChatGPT utilizes AI-based “transformer” algorithms—based on the same principles that connect human
neural cells—to scan the internet for germane material based on keywords, prompts, or other search
parameters, generating what’s referred to as “natural language text.” This same capability can be
leveraged by malicious hackers to create phishing emails and malware designed to infiltrate a corporate
network with far more speed, efficiency, and stealth than ever before.

Not only does this type of technology create natural language, it can also aid programmers in creating
code—and we all know not all code is leveraged for good. Sophisticated AI allows hackers to create
blocks of nefarious code almost instantaneously; or to automate portions of a cyberattack launch, such
as the initial infection code.

In the case of phishing and imposter attacks, the ability for AI-based bots to scrub the internet for accurate
details on a victim’s identity makes it nearly impossible for those users to recognize the attack. Gone are
the days when phony emails were glaringly marked by spelling mistakes and generic requests for
information. Today’s attacks will include the name of the user’s actual bank or healthcare provider, it will
refer to their home city or local pharmacy (“geo-phishing”), or will supposedly have come from a vendor
with whom the user regularly transacts.

Data is gathered with lightning speed through AI bots, pulled from existing references that populate the
internet. The email incorporating these details then typically supplies a link leading the recipient to a
brand imposter site where they will volunteer their user names and passwords.

The New Wave of AI-generated Security Threats

Just as CAD-based design programs now allow anyone with basic computer knowledge to function as a
designer, advanced AI content generators deliver programming capabilities and automated coding short-
cuts to even the least accomplished of wanna-be hackers. It is exponentially expanding the base of
criminals who can crank-out malware.

The rise of AI technologies like ChatGPT, Replica, and YouChat stands to produce a frightening
escalation of malicious activity targeting business networks, since it hands-over powerful tools to the ill-
intentioned. The increase in cyberattacks will likely include a wave of activity that targets email data,
since these systems are typically a point of entry for malware, viruses, and other threats.

Malicious software only needs to compromise a single account to infiltrate an entire network. Many
viruses will sit dormant on the system for a length of time, to avoid detection that can be traced back to
the date of infection, making the breach harder to mitigate. The viral code is later activated and works its
way through the network, scraping data for sale, collecting private credentials, or conducting denial-of-
service activities.

179