Background: Increasing sophistication of attacks by generative AI

Traditional phishing attacks have mainly focused on mass transmission "for unspecified numbers" methods. The unnatural nature of the body and subject line made it easy for recipients to be alert and easy for security software to detect. However, with the advent of generative AI, attackers can take advantage of:

• Natural sentence generation: Natural grammar and tone that are difficult to distinguish from human-written emails
• Enhanced targeted phishing: Generate personalized emails that mimic internal information or employee names
• Diversify attack methods: can be carried out on all channels, including attachments, link guidance, and via chat and messaging apps.

This can result in significantly higher email open and link click-through rates, which can cause widespread damage. Leakage, especially when sensitive or credential information is compromised, increases the risk of secondary damage and financial loss.

 

Attack Techniques and Risks

The main techniques of phishing attacks using generative AI include:

• AI-generated emails
  With high-precision text generation, the sender's name and text are cleverly disguised. By distracting users, they encourage them to open attachments or click on links.
• Spear Phishing
  Targeting specific employees and departments, using personal information and work-related information to send highly credible emails. There are also an increasing number of cases targeting administrative privileges and financial information.
• Information leakage risk
  Stealing credentials and sensitive data through email links and attachments can impact your business, damage your credibility, and pose legal risks.

These attacks are difficult to prevent with traditional rule-based or signature email security alone. In particular, natural text using generative AI is easy to bypass traditional detection techniques.

 

Measures and implementation procedures

To combat generative AI phishing, it is important to use a combination of human and technical countermeasures.

1. Email Flow Visualization
   Understand the reception and communication routes, and detect abnormal communications on internal mail servers and gateways at an early stage. It is possible to contain them in the early stages of an attack.
2. Filtering and detection settings
   • Enhanced sender authentication such as SPF/DKIM/DMARC
   • Introduce AI-based detection to detect generative AI emails and suspicious text patterns
3. Employee Training
   Conduct regular pseudo-attack drills and security training. We will disseminate the characteristics and precautions of attack emails to prevent damage before they occur.
4. Implement automated rules
   Automate suspicious email quarantine and warnings to reduce the risk of users inadvertently opening them.
5. Periodic Evaluation and Improvement
   Attack methods evolve daily, so we regularly review email filtering settings and employee training to stay updated with the latest threats.

 

What it should be

Ideally, a combination of human and technical measures is established to prevent damage in advance against generative AI phishing attacks.
Employees can identify attack emails, and the system can automatically quarantine and warn suspicious emails. This double defense minimizes the success rate of attacks.

 

Conclusion

Phishing attacks using generative AI are becoming more sophisticated and accurate than before.

• Mail Flow Visualization

• AI-based detection

• Employee Training

• Automated Isolation and Alerting

• Regular Evaluation and Improvement

Defense-in-depth combined with these is the key to minimizing risk for your organization. It is necessary to make efficient use of limited resources and prevent attacks before they occur.