AI Is Slowly Outperforming Human-written Phishing Emails, and It Is a Cause of Concern! Every day, millions of phishing emails are sent worldwide by threat actors to lure users into divulging their Personally Identifiable Information or other critical information, such as credit card credentials or bank account details. Here's how a team of researchers leveraged AI to write phishing emails, which seemed more realistic than human-written ones

By Remesh Ramachandran

Opinions expressed by Entrepreneur contributors are their own.

You're reading Entrepreneur India, an international franchise of Entrepreneur Media.

Remesh R

Spear phishing is a social engineering technique targeted towards a targeted individual to divulge confidential information. But creating highly targeted mass spear-phishing emails could take a lot of effort and time. In a recent test conducted by a team of researchers, it was found that they could use Natural Language Processing (NLP) to devise targeted phishing emails. At the end of the research, the team revealed that AI/ML could be used to develop spear-phishing campaigns at a devastating scale.

AI Can Write Better Phishing Emails Than Humans!

In the recently held Black Hat Defcon security conference in Las Vegas, a team of researchers hailing from the Singapore Government Technology Agency presented the results of their AI/ML generated phishing email test. Here's how the event unfolded:

  • The team stated that they created phishing emails using the Generative Pre-trained Transformer 3 (GPT 3) autoregressive language model in combination with other AI-as-a-service platforms.
  • They tested it by sending targeted phishing emails to more than 200 of their colleagues. Along with that, they also sent phishing emails created manually without the help of GPT-3.
  • But unlike actual phishing emails, these emails didn't contain any malicious links or attachments but simply links that sent the total click-through count back to the team.
  • The team was surprised to see that the success rate of GPT-3 generated emails was much higher than the manually crafted emails.

Why It Is A Cause For Concern

According to Eugene Lim, a cybersecurity specialist from the team that ran this test, it takes a lot of time and effort to train and develop a good AI model, and it can cost a lot of money. But if one succeeds in developing one and use it to create an AI-as-a-service model, it can save them a significant amount of time and future efforts to use that service to send phishing emails.

And if a threat actor gets their hands on this methodology, it could turn out to be disastrous as they would be able to send bulk emails targeted towards thousands of unsuspecting individuals.

How OpenAI Is Doing its Part in Prevent Malicious Use Of Their AI

OpenAI, the creators of GPT-2, GPT-3 and DALL-E, have stated that every request for GPT 3 access is monitored along with its production use before the product goes live.

OpenAI also stated that they are constantly working to enhance the accuracy of their system so that it doesn't reach the hands of malicious actors. Even though getting access to Open AI API is tough as one has to go through various checks and measures, there are other AI-as-a-service players in the market where one can get easy free trials.

Final Words

The team behind this phishing experiment utilized the facility of AI-as-a-service products and the GPT 3 deep learning model to learn the traits and behavior of their colleagues and generated targeted spear-phishing emails on a massive scale at once. This activity has demonstrated how AI tools can prove to be a double-edged sword; on the one hand, they can help spam detection tools identify and segregate phishing emails from regular ones, while on the other hand, if misused, they can enable threat actors to launch phishing attacks on a massive scale.

Remesh Ramachandran

CISO | Security Researcher | Ethical hacker

Remesh Ramachandran is an ethical hacker. He has solved several sophisticated cybercrime and real-world hacking cases, and has worked for the government and various other national and international agencies. Remesh is currently working as a CISO (Chief Information Security Officer) for an organisation.

Business Ideas

70 Small Business Ideas to Start in 2025

We put together a list of the best, most profitable small business ideas for entrepreneurs to pursue in 2025.

Technology

TCS Q1 Profit up 6%, Revenue Down 3.1% on Delayed Client Decision Making

Operating margins for the June quarter narrowed to 24.5 per cent from 24.7 per cent in the corresponding quarter of the previous year but expanded 30 basis points from 24.2 per cent in preceding three months.

Leadership

5 Lessons I Wish I Didn't Learn the Hard Way During My 20 Years in Business

Starting a business has been one of the most exciting and fulfilling chapters of my life — but it's also been unpredictable, humbling and often much harder than it looks from the outside. Here are a few pieces of hard-earned wisdom I wish I'd known from day one.

News and Trends

India Accelerator Launches Two Flagship Programs to Build Future Startup Leaders

These flagship programs are tailored to cultivate entrepreneurial and venture capital talent—nurturing the builders, operators, and investors who will shape the next wave of Indian innovation.

Growth Strategies

Rare Earth Crisis: Vedanta Is On a Mine To Magnet Strategy

Once the exploration efforts of rare earth bear viable fruit, the authorities can opt for a price discovery model. This will mitigate the key challenge of limited geological exploration data and turn this sector into an investor-friendly one, says Priya Agarwal Hebbar, chairperson, Hindustan Zinc

Devices

Boost Office Output with This 27″ 4K Monitor Deal

Upgrade your team's workspace and boost employee performance with $60 off a 27" 4K monitor.