DEV Community

Cover image for Lazarus Group's 19-Day A/B Test: How North Korean APT Pivoted from Airdrops to Fake CVEs to Dream Jobs

Lazarus Group's 19-Day A/B Test: How North Korean APT Pivoted from Airdrops to Fake CVEs to Dream Jobs

KL3FT3Z on April 08, 2026

description: "Technical analysis of three consecutive Lazarus Group campaigns targeting the same GitHub users with different social engineering vec...
Collapse
 
gnomeman4201 profile image
GnomeMan4201

This looks like outer-loop automation ,lure generation, segmentation, and A/B testing paired with a human-governed inner loop for high-risk actions. The 19-day pivot suggests the Observe/Orient cycle is tightening, but burn rate and attribution pressure still limit full end-to-end autonomy. In other words, sensing and adaptation are scaling faster than execution can safely keep up.

Collapse
 
gnomeman4201 profile image
GnomeMan4201 • Edited

feels less like persistence and more like steering.
If telemetry or observability can be influenced, you can feed the team a stable but false picture of the system and the developer’s OODA loop becomes the attack surface.
The AI part is what makes it nastier: bad instrumentation doesn’t just hide the attack, it can train the model toward the wrong fix.

Collapse
 
toxy4ny profile image
KL3FT3Z

You're absolutely right! I love watching their evolution and getting the opportunity to use the techniques for practice phishing tests within RedTeam.

Thread Thread
Collapse
 
copyleftdev profile image
Don Johnson
Collapse
 
toxy4ny profile image
KL3FT3Z

its cool! Thanx! Very good job!