Skip to navigation

Threat reportSupply ChainTL-2026-2823

Bitget $387.5M Cryptocurrency Theft via Third-Party Security Product Zero-Day (Suspected DPRK / TraderTraitor)

criticalACTIVE

Attackers stole about $387.5 million from Bitget hot and warm wallets on 24-25 September 2026 after exploiting a zero-day in an unnamed third-party security product (Product A) and compromising a second one (Product B), then using a custom withdrawal tool.

Key facts

IdentifiersTL-2026-2823
Also known asBitget hack, Bitget wallet infrastructure breach
SeverityCRITICAL
First disclosed
Affected productsUnnamed third-party vendor Security product A (service on node; zero-day); Unnamed third-party vendor Security management platform B (appliance with web shell); Bitget Centralized exchange hot and warm wallet infrastructure
Weaknesses (CWE)CWE-78
AttributionTraderTraitor (medium confidence)
Nation-state nexusNorth Korea
MalwareAvalanche, Web shell on security appliance B
ToolsCustom Bitget withdrawal tool (deleted, recovered by SlowMist)
MITRE ATT&CK techniquesT1059 Command and Scripting Interpreter, T1071 Application Layer Protocol, T1078 Valid Accounts, T1190 Exploit Public-Facing Application, T1195 Supply Chain Compromise, T1505.003 Server Software Component: Web Shell, T1552 Unsecured Credentials, T1565.002 Data Manipulation: Transmitted Data Manipulation, T1570 Lateral Tool Transfer, T1657 Financial Theft
MotivationFINANCIAL
Target sectorsfinance, cryptocurrency
Target regionsGlobal, Asia-Pacific
Detection rules9 (1 shown free)
Sources8 cited sources

Detection coverage

As of 2026-10-01, Threadlinqs Intelligence publishes 9 detection rule(s) for TL-2026-2823 across Splunk SPL, Microsoft KQL and Sigma, covering 5 indicator(s) of compromise. The whole corpus is readable without an account. Every threat comes with one complete detection in Splunk SPL, Microsoft KQL and Sigma, free for everyone; a free account adds daily TL Creds to unlock the rest. Paid plans add a larger daily allowance, correlation and the MCP server. Threadlinqs MCP server · View plans.

Free detection: Bitget TTP — Security Product Service Process Spawning Hidden Shell/Script (Exploitation)

Detects a service process on a security-appliance node (java, node, python, php-fpm, httpd, nginx and similar) spawning a shell or interpreter with download, reverse-shell, base64 or script-execution arguments. Mandiant/SlowMist report the Bitget intrusion began with a zero-day in a service on a Product A node that ran hidden scripts under the service process (first seen 2026-08-31, spreading to more nodes 2026-09-23 to 09-25). Scope with the security_product_hosts lookup.

  • Languages: Splunk SPL, Microsoft KQL, Sigma
  • ATT&CK: T1190, T1195, T1059
  • Data sources: sysmon_for_linux, linux_auditd, microsoft_defender_for_endpoint_linux
  • Severity: critical
  • Confidence: medium

Splunk SPL

index=linux sourcetype="sysmon:linux" EventCode=1 earliest=-24h
```Service process spawning shell/interpreter on security-product nodes (T1190/T1059)```
| where match(ParentImage,"/(java|node|python3?|php-fpm|httpd|apache2|nginx|uwsgi|dotnet)$") AND match(Image,"/(sh|bash|dash|zsh|python3?|perl|ruby|nc|ncat)$")
| where match(CommandLine,"(?i)(base64|/dev/tcp|mkfifo|curl |wget |nohup |chmod [+]x|[.]sh)")
```security_product_hosts lookup: host, product_role (Product A / Product B nodes)```
| lookup security_product_hosts host OUTPUT product_role
| where isnotnull(product_role)
| stats count min(_time) as first_seen max(_time) as last_seen values(CommandLine) as commands values(User) as users by host product_role ParentImage Image
| convert ctime(first_seen) ctime(last_seen)

Microsoft KQL

let service_parents = dynamic(['java','node','python','python3','php-fpm','httpd','apache2','nginx','uwsgi','dotnet']);
// Populate with the Product A / Product B node names
let mgmt_hosts = dynamic([]);
DeviceProcessEvents
| where Timestamp > ago(24h)
| where InitiatingProcessFileName in~ (service_parents)
| where FileName in~ ('sh','bash','dash','zsh','python','python3','perl','ruby','nc','ncat')
| where ProcessCommandLine matches regex @'(?i)(base64|/dev/tcp|mkfifo|curl |wget |nohup |chmod [+]x|[.]sh)'
| where array_length(mgmt_hosts) == 0 or DeviceName in~ (mgmt_hosts)
| project Timestamp, DeviceName, AccountName, InitiatingProcessFileName, InitiatingProcessCommandLine, FileName, ProcessCommandLine
| order by Timestamp desc

Sigma

title: Security Product Service Process Spawning Hidden Shell or Script
id: 7c1e5b3a-2d4f-4e8a-9b61-0a3f5c7d9e11
status: experimental
description: Detects a web/application service process on a security appliance node spawning a shell or interpreter with download, reverse-shell or script-execution arguments, as reported in the Bitget zero-day intrusion (hidden scripts under a third-party security product service).
references:
  - https://www.bleepingcomputer.com/news/security/bitget-hacked-via-zero-day-in-third-party-security-products/
  - https://thehackernews.com/2026/10/bitget-confirms-third-party-zero-day.html
author: Threadlinqs Team
date: 2026/10/01
tags:
  - attack.initial-access
  - attack.t1190
  - attack.t1195
  - attack.execution
  - attack.t1059
logsource:
  category: process_creation
  product: linux
detection:
  selection_parent:
    ParentImage|endswith:
      - '/java'
      - '/node'
      - '/python3'
      - '/python'
      - '/php-fpm'
      - '/httpd'
      - '/apache2'
      - '/nginx'
      - '/uwsgi'
      - '/dotnet'
  selection_child:
    Image|endswith:
      - '/sh'
      - '/bash'
      - '/dash'
      - '/zsh'
      - '/python3'
      - '/python'
      - '/perl'
      - '/ruby'
      - '/nc'
      - '/ncat'
  selection_cmd:
    CommandLine|contains:
      - 'base64'
      - '/dev/tcp'
      - 'mkfifo'
      - 'curl '
      - 'wget '
      - 'nohup '
      - 'chmod +x'
      - '.sh'
  condition: all of selection_*
falsepositives:
  - Security product health-check or update scripts launched by the service
  - Administrators running maintenance through the appliance management service
level: high

Known false positives:

  • Vendor update or health-check scripts spawned by the appliance service
  • Admin maintenance performed via the management service

Tuning notes: Populate security_product_hosts / mgmt_hosts with Product A and B nodes. Baseline legitimate vendor scripts for 14 days and exclude by exact CommandLine. Product vendors are unnamed in public reporting, so scope by host role rather than product name.

8 more detection rules (Splunk SPL, Microsoft KQL, Sigma) · unlock with TL Creds. Sign up free to unlock the detection rules or query them with the Threadlinqs MCP server (Purple)
  • Bitget TTP — Command Injection into Security Management Platform Task Parameters (Exploitation) · unlock with TL Creds
  • Bitget TTP — Log Tampering and Audit Evasion on Security and Wallet Hosts (Post-Exploitation) · unlock with TL Creds
  • Bitget TTP — Lateral Tool Transfer from Security Appliance to Wallet Job Servers (Lateral Movement) · unlock with TL Creds
  • Bitget TTP — Hidden Script Harvesting Service Environment Credentials and Connecting to Database (Exploitation) · unlock with TL Creds
  • Bitget TTP — Interactive Logon to Security Product Hosts from Unapproved Source (Post-Exploitation) · unlock with TL Creds
  • Bitget TTP — Withdrawals Executed Without Valid Risk-Control Approval or Burst Across Chains (Impact) · unlock with TL Creds
  • Bitget TTP — Service Process Writing Web Shell, Relay or Malicious Program Files (Post-Exploitation) · unlock with TL Creds
  • Bitget TTP — Security Appliance Service C2 Egress and Community-Observed DPRK Infrastructure (C2) · unlock with TL Creds

Indicators of compromise

As of 2026-10-01, Threadlinqs Intelligence tracks 5 indicators of compromise for TL-2026-2823.

5 indicators of compromise · unlock with TL Creds. Compare plans

Technical analysis

Bitget attributes the intrusion to North Korean actors based on IP behavior and on-chain analysis; Mandiant and SlowMist are investigating.

Bitget disclosed unauthorized transfers from its hot and warm wallets on 24 September 2026 (18:31 UTC; 02:31 on 25 September UTC+8). Initial estimates of $351.6M were revised to $387.5M after further ZEC and TRX losses were found. Bitget says private keys and cold wallets were not compromised and the exchange's User Protection Fund (reported above $464M) covers losses.

Per the interim SlowMist and Mandiant findings reported on 30 September 2026, the earliest malicious activity was on 31 August 2026 against 'Product A', an unnamed third-party security product, through a zero-day in a service running on a node. The attacker ran hidden scripts under that service process, read an environment variable holding the database password, and connected directly to the database. Similar hidden-script activity appeared on further nodes between 23 and 25 September.

On 24 September (about 16:07 UTC) the attacker moved into 'Product B', a security management platform, reportedly using an internal employee identity. SlowMist reports three consecutive system command injection attempts into task parameters, use of a web execution endpoint, relay communication files written to the server, and batch assembly and upload of malicious programs (first activity about 17:49 UTC). Mandiant found a web shell on security appliance B that established a command-and-control connection and allowed lateral movement to Bitget's production wallet job server, where malicious packages were deployed.

SlowMist recovered a deleted, highly customized tool built around the wallet system's withdrawal logic. It forged risk-control parameters, built withdrawal requests and invoked withdrawals, spoofing transaction data to trigger the exchange's normal authorization signing process. Theft ran in waves across 11 chains (Ethereum, XRP Ledger, Zcash, TRON, Arbitrum, Optimism, Base, BNB Smart Chain, Avalanche, Algorand, Celestia), with the largest wave of about $185M reported in roughly one minute around 19:16 UTC. The on-chain window closed around 21:23 UTC. Failed BTC withdrawal attempts and attempted log alterations were also observed.

Attribution to North Korean actors rests on IP behavior patterns and on-chain analysis. Elliptic and TRM Labs report wallet overlaps with earlier DPRK-attributed hacks, and Elliptic ties the laundering pattern to TraderTraitor. Funds were swapped within hours and routed through cross-chain venues including THORChain and Chainflip; Bitget asked THORChain to refuse service to attacker addresses, and about $632,700 was frozen by Circle, Tether and NEAR Intents. No government attribution has been issued. No CVE, vendor/product names, hashes, IPs, domains or wallet addresses had been published at the time of research.

MITRE ATT&CK techniques by tactic:

Execution

T1059 Command and Scripting Interpreter

Command and Control

T1071 Application Layer Protocol

Initial Access

T1078 Valid Accounts; T1190 Exploit Public-Facing Application; T1195 Supply Chain Compromise

Persistence

T1505.003 Server Software Component: Web Shell

Credential Access

T1552 Unsecured Credentials

Impact

T1565.002 Data Manipulation: Transmitted Data Manipulation; T1657 Financial Theft

Lateral Movement

T1570 Lateral Tool Transfer

Remediation

Patches

  • Vendor patches for the unnamed third-party products were pending at time of reporting; apply them when released

Immediate actions

  • Hunt for hidden scripts spawned under security-appliance service processes and for reads of service environment variables that hold database credentials
  • Review security-appliance management platforms for command injection into task parameters, unexpected web-execution endpoint use and relay files written to disk
  • Audit use of employee identities against security management platforms and rotate those credentials
  • Treat withdrawal/signing pipelines as untrusted if the upstream risk-control or job server was reachable from a compromised appliance

Workarounds

  • Disable the affected third-party functionality until fixes ship (the step Bitget took)

Longer-term hardening

  • Segment security appliances and their management planes from wallet job servers and signing infrastructure
  • Add independent, out-of-band validation of withdrawal requests and risk-control parameters
  • Keep cold wallets and private keys isolated from application-layer backends

Affected and fixed versions

  • Unnamed third-party vendor — Security product A (service on node; zero-day)
    Vulnerable versions: unknown
  • Unnamed third-party vendor — Security management platform B (appliance with web shell)
    Vulnerable versions: unknown
  • Bitget — Centralized exchange hot and warm wallet infrastructure
    Vulnerable versions: not applicable

Timeline

  • Earliest malicious activity: zero-day in a service on a node of third-party security Product A exploited; hidden script reads the database password from an environment variable
  • Hidden-script activity begins appearing on additional nodes (observed through 25 September)
  • On-chain theft window closes around 21:23 UTC; failed BTC withdrawal attempts and log alteration attempts observed; Bitget suspends withdrawals and discloses the breach
  • 18:31 UTC (02:31 on 25 Sep UTC+8): first on-chain outflows from hot and warm wallets via the custom withdrawal tool; largest wave about $185M in roughly one minute near 19:16 UTC
  • Around 16:07 UTC the attacker moves into security management platform Product B using an internal employee identity and attempts three command injections into task parameters; malicious programs start about 17:49 UTC
  • Bitget attributes the attack to suspected North Korean actors based on IP behavior and on-chain analysis; initial loss estimate $351.6M; Elliptic and TRM Labs report wallet overlaps with prior DPRK hacks
  • Losses revised to $387.5M after additional ZEC and TRX theft is found; Bitget asks THORChain to refuse service to attacker addresses
  • SlowMist and Mandiant publish interim findings: two third-party security products compromised, web shell with C2 on appliance B, lateral movement to the production wallet job server
  • Bitget confirms the third-party zero-day vector, notifies the vendor and disables affected functionality pending fixes; about $632,700 frozen by Circle, Tether and NEAR Intents

Sources cited

More from TraderTraitor

Previous and next in supply chain

Further reading

Drafted by the Threadlinqs research pipeline from 8 sources; published ; last updated —.

Threadlinqs Intelligence — Real-Time Threat Detection Platform

// threat_feed
$ sort --newest
Showing all threats

Live intelligence console

Threat level

Latest Threats