[Report] Operation ShadowCommit - A Covert Operation on GitHub - CODE BLUE 2026 #codeblue_jp #codeblue2026

[Report] Operation ShadowCommit - A Covert Operation on GitHub - CODE BLUE 2026 #codeblue_jp #codeblue2026

A report on the CODE BLUE 2026 session "Operation ShadowCommit - A Covert Operation on GitHub."
2026.11.18

This page has been translated by machine translation. View original

Hello, this is Usuda.

Everyone, are you taking security measures? (greeting)

This is a report on the following session held at CODE BLUE 2026.

It was a session that reported on research including the activities of actual attackers.

Operation ShadowCommit - A Covert Operation on GitHub

This talk explains in detail the results of an analysis of an operation carried out over two years by an attack group linked to North Korea. It goes beyond analysis of the malware itself, examining the infrastructure they use, their attack techniques, their targets, the damage caused, and even the purpose and true nature behind it all. Even though their activities were exposed by their use of public services, they continue to carry out these attacks without significant hindrance. We invite you to join this talk, which unravels the profile of these attack actors, something never before revealed.

Speakers: Jun Hyeong Lee, Daehoo Lee

Report

  • Works in the field of DFIR
  • The only DFIR-specialized company in South Korea
  • ShadowCommit, an attacker campaign believed to have ties to North Korea
  • Focuses on the actors behind it rather than the malware itself
  • Cryptocurrency theft incident
    • An investigation request from a victim whose cryptocurrency was stolen
    • Initially thought it was a common case
    • But tracing back revealed a big overall picture
    • The attacker targeted cryptocurrency influencers and others
    • Stole user information
    • Sends messages to a list
    • When information is uploaded to GitHub and matches, activity begins
    • Ultimately steals the cryptocurrency
    • Everything is done on GitHub
    • How do they find targets?
      • At first, influencers and the like
      • They have a member roster
      • A good list for attackers
      • The attacker's folder contained phishing email templates, decoy files, and actual account materials
      • There is a member roster that was obtained
      • Sends phishing emails to members
      • To members, it looks like an email about a course they are taking
    • Online communities are targeted
      • Cafe magic llama
      • Cafe ID
      • Pi coin community
      • Places where cryptocurrency holders gather are being targeted
    • Phishing emails
      • Contents vary
        • Ones impersonating a friend
        • Claiming to have sent the wrong seed and following up with a message telling them not to open it
        • Password-protected files
        • The password is always written in the email body
      • LNK malware
        • RATs and so on
        • PowerShell is executed, establishing persistence, etc.
        • Files are on GitHub
        • Accesses private areas with a PAT
        • A new file means the number of infected PCs has increased
        • The attacker looks at the contents to determine whether it is a user's PC or an analyst's environment
        • One PC serves multiple roles
      • If it is a target, a RAT is planted
        • The method is unique
        • Rewrites a file and commits it
        • The contents change and become a RAT
        • XenoRAT is used
        • It looks broken but gets deployed
      • The attacker infiltrates using the RAT
        • A keylogger is planted
        • Records are saved as CHANGELOG.txt and uploaded
        • Sent from most victims
      • Finally, the cryptocurrency is siphoned off
        • Steals ID images from the victim's PC
        • Seeds are saved as photos or in KakaoTalk
        • Those are also stolen from the PC
    • From this analysis, the attacker's GitHub was analyzed
  • ShadowCommit campaign
    • Active for over two years since 2024
    • 3,380 victims observed
    • Including about 6 government-related cases
    • The damage is more widespread
    • Some victims have large amounts of keylogs, and some had RATs planted multiple times
    • Why did they use GitHub?
      • It is seen as legitimate traffic
    • Servers can be created easily
      • If private, it cannot be seen from outside
      • Hard for analysts to find
    • Can be centrally controlled via API
      • Automation is also easy
      • The attacker made and used a tool that could create and delete repositories with one click
    • The malware to be distributed can also be swapped easily
      • Swapping a Beacon for a RAT allows distribution without changing the address
      • In an investigation, you must not check only the current files
    • How victims are identified
      • Normally IP addresses are important
      • But all are private addresses
      • Here, to identify victims, scripts are distributed from GitHub or the RAT is made to connect to the C2
      • Most are managed by the victim's name
      • The repo name is the victim's name
      • That made it possible to determine the number of victims
    • Who are the targets?
      • There was phishing impersonating law firms, among others
      • Impersonating accounting firms, they requested signatures, ID documents, bank accounts, etc.
      • In both cases, files were sent claiming they were password-protected for personal information protection
      • There were also emails that looked like responses to the victim's own requests, such as those from child consultation centers or futures trading
      • There was even one about a New Year's gift
      • They are tailored to the victim
    • How does the attacker achieve their goals?
      • Automation
        • LNK generation
        • Mass sending of phishing emails is also automated
          • Uses Zoho, rotating 12 accounts
          • The body is HTML
          • The subject includes the target's email address
        • RAT infection and connection
          • Swaps the Beacon
          • Estimating RAT usage, once used, it was operated for 2-3 hours
      • Browser extension
        • Can log in as the target user with one click
        • Managed the value of targets
      • Using AI to bypass KYC
        • Exchanges normally request face photos or videos
        • AI-generated images and videos of non-existent victims were used
    • An example where a YouTuber had Pi stolen
    • From the attacker's tool usage history, logs disguised with AI could be confirmed
  • About the attacker
    • Believed to have ties to North Korea
    • Fluent in Korean and knows South Korea well
    • Emails sent
      • There are set times for sending
      • Passwords are attached
      • The target's name is used to make it look related
      • KakaoTalk is used
      • They understand Korean people well
    • Work like business hours, such as the timing of repository creation
      • Active at the same time every day
      • Not one person but a team
      • 7 repositories running simultaneously
      • 7 Telegram accounts
      • Team of 5-6 people
      • rama.god
      • Members request malware creation from the leader
      • A clear division of labor
    • Malware upload
      • Could also be confirmed on the actual attacker's screen
      • Repo management
      • Many repos under one account
      • One victim per repo
      • Everything finishes within minutes inside GitHub
    • They are still evolving their methods
      • Because LNK has become more suspected than before
      • Sometimes switching to EXE
      • Sometimes sent as internal confidential documents
      • Described as security documents
      • Sent encrypted as a service review request
      • An OTP is requested and split files are fetched in the background
      • Distributed from Dropbox
  • Summary
    • GitHub was used for malware distribution and for retrieving stolen content
    • Fluent in Korean and active at set times
    • We should be wary of activity that blends into normal services

Impressions

The details of the attacker activity behind this were very interesting.

I want to stay careful.

Share this article