How AI Platforms Like ChatGPT will Heighten Imposter Attacks—and How to Fight Them

• By Rom Hendler
• Jun 20, 2023

Everyone in recent months has been busy either debating the merits of ChatGPT and similar natural language generators—or have been busy using them. The amount of material these AI-driven platforms will produce is mind-boggling, since there seems to be no shortage of businesses and content writers who feel they can benefit from the striking shortcuts these platforms provide. ChatGPT was in fact able to reach 100 million users in its first two months of public existence.

However, legitimate businesses aren’t the only ones who can benefit tremendously from these AI-powered capabilities to generate meaningful passages. Many pundits warn that this ability to create article-length narratives in mere seconds will make it frighteningly simple for criminals to create more persuasive phishing and imposter attacks, and at a far greater volume. This onslaught of new threats will hugely accelerate vulnerability, where even the savviest of network users could be tricked into turning over log-in credentials, healthcare information, or financial data.

It’s not surprising that AI would allow the ranks of cyber criminals to grow. Technology has often opened fields of expertise up to amateurs, making it easier for laymen with minimal skills to master tasks that formerly required much more training and effort. Consider that automated software allows anyone with a CAD program to draft impressive 3-dimensional designs, and WordPress and Wix allow users with even the most basic of abilities to create professional websites. We can view ChatGPT in the same light, as a tool for hackers. It not only allows anyone with an Internet connection to compose believable and supposedly informed text, but it also empowers hackers who start out with even the most rudimentary of skills to swiftly generate scripts and launch language for imposter cyber-attacks.

These imposter events come in various forms. In the corporate community, Business Email Compromise (BEC) occurs when nefarious actors breach the account of a high-level executive, often a CEO. The hacker will send emails from the CEO’s account directing other senior executives to do things like make large wire transfers or reveal sensitive log-in information. These “socially engineered” BEC attacks have increased by 65% since 2019, according to reports from software research site Gitnux, and are expected to spike dramatically along with the new sophistication of language generators.

Brand imposter attacks are when hackers create a credible mock-up of a site that the victim frequents, such as a financial institution, cloud provider, transport company, or healthcare organization. The criminals will send well-composed and convincing emails requesting that the victim click a link to their site due to some matter that needs attention. The user is then brought to the clever look-alike site, and prompted to enter their user names, passwords, banking details, address, or identifying healthcare information.

Here are some ways that ill-intentioned hackers can now produce code more quickly, launch attacks more precisely, and compose phishing content more eloquently than ever:

ChatGPT allows overseas hackers to write grammatically correct, accurately composed language. In the earlier days of phishing, hackers in unregulated foreign countries were often foiled by spelling mistakes, awkward phrasing, and unprofessional grammar that tipped-off readers. Natural language generators will produce well-composed email copy that is completely indistinguishable from ordinary native speech, since the text is not composed by an outsider. It’s composed by an AI algorithm, pulling from existing native sources.

ChatGPT makes it easier for cyber criminals to write effective malware. Not only do AI-based language generators instantly create prose, they can also quickly write code, aiding programmers in developing applications. Researchers have already reported evidence on the dark web of malicious actors abusing ChatGPT to speed the creation of new malware or fine-tune existing malicious programs. As usual, cyber criminals are intelligent and resourceful—they have already found ways to circumvent ChatGPT’s inherent safeguards.

How to Protect Against Heightened Attacks

All this makes it more critical than ever for businesses to use AI-driven email protection. The only way organizations can guard against the power and speed of advanced AI is to leverage the same technologies in their cyber security solutions. The challenge is that even many top-tier software packages don’t utilize best-in-class AI, because they were designed before these sophisticated tools had even been developed.

Many existing security solutions rely on traditional SEG (security email gateway) methods as their legacy technique. This involves the blacklisting of known malicious IP addresses. Yet contextual attacks like the BEC scenarios above simply can’t be detected by these SEG-based solutions. Cyber security solutions must employ powerful AI to interpret the text of ill-intended emails, identifying keywords like “wire transfer” and “credit card” or even recognizing attachments with sensitive images such as healthcare ID cards. Without these intelligent AI-based tools, which include optical character recognition, companies are vulnerable to a ramp-up in breaches now that criminals have access to tools like ChatGPT.

Organizations should consider solutions from new, next generation cybersecurity providers, especially those who specialize in email security, including solutions for anti-malware, anti-virus, and data loss protection. Outbound email protection like best-in-class encryption is also advisable, since hackers can’t exploit emails that they can’t decode. Businesses should also demand email security protection that is easy to use, in order to foster greater adoption across the organization. Technology that doesn’t get used is pointless.

In the end, the only genuine strategy for combatting the increased level of AI-based attacks from these platforms is to use the same AI tools against them. Don’t let your organization be swept up in the watershed of ChatGPT-assisted schemes.