Skip to navigation

Threat reportPhishingTL-2026-2834

Milk Dragon (NaiLong) AiTM Phishing-as-a-Service Kit Targeting Social Media Shoppers and Bank MFA

highACTIVE

Group-IB details Milk Dragon (NaiLong), an Adversary-in-the-Middle phishing kit sold as PhaaS on Telegram from 300 USDT per month. It spreads through Facebook and TikTok marketplace ads and posts offering heavily discounted goods, using fake WordPress/WooCommerce storefronts and a custom 'BytePress' plugin for C2, real-time keystroke capture, and 3D Secure MFA interception and relay.

Key facts

IdentifiersTL-2026-2834
Also known asMilk Dragon, NaiLong
SeverityHIGH
Affected productsLEGO Consumer brand impersonated by fake storefronts (no vulnerability); Calvin Klein Consumer brand impersonated by fake storefronts (no vulnerability); Aeon Malaysia Retailer impersonated by fake storefronts (no vulnerability); Various financial institutions (36 templates) 3D Secure / OTP pages spoofed for MFA relay (no vulnerability)
MalwareMilk Dragon, NaiLong, telegram
ToolsDocker, Milk Dragon operator panel, WooCommerce, WordPress, socket.io
MITRE ATT&CK techniquesT1056.001 Keylogging, T1071.001 Web Protocols, T1111 Multi-Factor Authentication Interception, T1204.001 Malicious Link, T1557 Adversary-in-the-Middle, T1566.003 Spearphishing via Service, T1583.001 Domains, T1585.001 Social Media Accounts, T1587.001 Malware, T1657 Financial Theft, T1684.001 Impersonation
MotivationFINANCIAL
Target sectorsretail, ecommerce, finance, consumer
Target regionsGlobal
Detection rules9 (1 shown free)
Sources7 cited sources

Detection coverage

As of 2026-10-01, Threadlinqs Intelligence publishes 9 detection rule(s) for TL-2026-2834 across Splunk SPL, Microsoft KQL and Sigma, covering 14 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: Milk Dragon — Social Media Ad Click-Through to Unknown Storefront Checkout (Exploitation)

Detects a client arriving from a Facebook, Instagram or TikTok referrer on a non-allowlisted shop domain and browsing WooCommerce-style product, cart or checkout URLs. Milk Dragon (NaiLong) distributes fake discount storefronts through Facebook and TikTok marketplace ads and posts; this surfaces the victim click-through and checkout interaction (T1566.003 / T1204.001). Referrers are sometimes stripped, so absence of a hit is not proof of safety.

  • Languages: Splunk SPL, Microsoft KQL, Sigma
  • ATT&CK: T1566.003, T1204.001, T1656
  • Data sources: proxy (zscaler:nss:web, bluecoat:proxysg:access:syslog, pan:url, squid:access), CommonSecurityLog (CEF web proxy)
  • Severity: medium
  • Confidence: medium

Splunk SPL

index=proxy sourcetype IN ("zscaler:nss:web","bluecoat:proxysg:access:syslog","pan:url","squid:access") earliest=-24h
```Milk Dragon: social-ad referrer -> unknown shop host with cart/checkout activity```
| eval ref=lower(coalesce(http_referrer,referrer,cs_referer)), u=lower(url), host=lower(coalesce(dest_host,url_domain,dest))
| where match(ref,"^https?://([a-z0-9-]+[.])?(facebook|instagram|tiktok)[.]com/")
| where NOT match(host,"(^|[.])(facebook|fbcdn|instagram|cdninstagram|tiktok|tiktokcdn|whatsapp)[.](com|net)$")
| where match(u,"/(product|shop|cart|checkout)(/|[?]|$)|add-to-cart=|wc-ajax=")
| lookup tl_known_retail_domains host OUTPUT is_known
| where isnull(is_known)
| stats count as hits dc(u) as distinct_urls values(http_method) as methods min(_time) as first_seen max(_time) as last_seen by src, user, host
| where distinct_urls>=2
| convert ctime(first_seen) ctime(last_seen)

Microsoft KQL

let KnownRetail = _GetWatchlist('KnownRetailDomains') | project Host = tolower(tostring(SearchKey));
CommonSecurityLog
| where TimeGenerated > ago(24h)
| where tolower(RequestContext) matches regex @"^https?://([a-z0-9-]+[.])?(facebook|instagram|tiktok)[.]com/"
| extend Host = tolower(DestinationHostName)
| where Host !matches regex @"(^|[.])(facebook|fbcdn|instagram|cdninstagram|tiktok|tiktokcdn|whatsapp)[.](com|net)$"
| where tolower(RequestURL) matches regex @"/(product|shop|cart|checkout)(/|[?]|$)|add-to-cart=|wc-ajax="
| join kind=leftanti KnownRetail on Host
| summarize Hits = count(), DistinctUrls = dcount(RequestURL), Methods = make_set(RequestMethod, 5), FirstSeen = min(TimeGenerated), LastSeen = max(TimeGenerated) by SourceIP, SourceUserName, Host
| where DistinctUrls >= 2
| project FirstSeen, LastSeen, SourceIP, SourceUserName, Host, Hits, DistinctUrls, Methods

Sigma

title: Milk Dragon Social Media Referral to Fake WooCommerce Storefront
id: 7b3e5c1a-92d4-4f68-a1b7-3c0e8d5f2a41
status: experimental
description: Detects proxied requests to shop, cart or checkout URLs whose referrer is Facebook, Instagram or TikTok, matching the Milk Dragon (NaiLong) AiTM PhaaS lure flow. Tune with an allowlist of legitimate retailers.
references:
  - https://www.group-ib.com/blog/milk-dragon-nailong-phishing-kit/
author: Threadlinqs Team
date: 2026-10-01
tags:
  - attack.initial-access
  - attack.t1566.003
  - attack.execution
  - attack.t1204.001
logsource:
  category: proxy
detection:
  selection_referrer:
    cs-referrer|contains:
      - 'facebook.com/'
      - 'instagram.com/'
      - 'tiktok.com/'
  selection_shop:
    c-uri|contains:
      - 'add-to-cart='
      - 'wc-ajax='
      - '/checkout'
      - '/cart'
  filter_social_hosts:
    cs-host|endswith:
      - '.facebook.com'
      - '.fbcdn.net'
      - '.instagram.com'
      - '.cdninstagram.com'
      - '.tiktok.com'
      - '.tiktokcdn.com'
  condition: selection_referrer and selection_shop and not filter_social_hosts
falsepositives:
  - Legitimate retailers reached from social media ads
  - Employees shopping from social media on corporate networks
level: medium

Known false positives:

  • Legitimate retailers reached via Facebook/Instagram/TikTok ads or Shop links
  • Brand-owned storefronts not yet in the known-retail allowlist

Tuning notes: Populate the tl_known_retail_domains lookup / KnownRetailDomains watchlist with retailers your users legitimately buy from. Many proxies strip or omit the Referer header, so also feed the stage-2 detections. Verify that your proxy maps the referrer into RequestContext (CEF) or http_referrer (Splunk CIM) before relying on it.

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)
  • Milk Dragon — Lookalike Brand + Discount-Lure Domain Contacted by Endpoint (Exploitation) · unlock with TL Creds
  • Milk Dragon — [indicator] Plugin Files Written on Web Host (Operator Infrastructure Build) · unlock with TL Creds
  • Milk Dragon — Web Server Process Calling Telegram Bot API (Operator Notification C2) · unlock with TL Creds
  • Milk Dragon — [indicator] Plugin Asset Requested in Web Traffic (Exploitation) · unlock with TL Creds
  • Milk Dragon — [indicator] Plugin Activation or API Base URL Configuration in wp-admin · unlock with TL Creds
  • Milk Dragon — Non-Browser Client Posting to Telegram Bot sendMessage Endpoint (Proxy) · unlock with TL Creds
  • Milk Dragon — Fake Shop Checkout Followed by 3DS/OTP Page on Same Host (MFA Interception) · unlock with TL Creds
  • Milk Dragon — socket.io Keystroke Channel on a Fake WooCommerce Storefront (C2) · unlock with TL Creds

Indicators of compromise

As of 2026-10-01, Threadlinqs Intelligence tracks 14 indicators of compromise for TL-2026-2834.

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

Technical analysis

Milk Dragon (also tracked as NaiLong) is an Adversary-in-the-Middle (AiTM) phishing kit distributed as Phishing-as-a-Service (PhaaS) through Telegram communities, priced from 300 USDT per month with various subscription plans and add-ons. Group-IB reports it has been active since October 2025 and identified 258 phishing pages with victims across 66 countries. Unlike fear-based phishing, the lure is commerce: heavily discounted consumer goods advertised through Facebook and TikTok marketplace ads and organic posts, in some cases from potentially AI-generated profiles with purchased followers to appear legitimate. Impersonated storefronts span 21 consumer brands (cosmetics, fashion, food and beverage, home/baby products, toys; e.g. LEGO, Calvin Klein, Aeon Malaysia) plus regional supermarkets.

The phishing sites are fake WordPress/WooCommerce storefronts. A second, custom plugin called BytePress is installed alongside WooCommerce and provides the command-and-control link: it adds fake credit card and PayPal payment methods and exposes an 'API Base URL' configuration field that ties the fraudulent checkout to the operator's C2 panel. BytePress keeps a persistent socket.io WebSocket connection to the panel, streaming the victim's input character by character without a form submission, so card data is captured in real time even if the checkout is abandoned.

The operator panel is connected directly to live phishing pages. It supports multiple accounts with role-based access control, Telegram bot notifications on victim activity, BIN-based card identification with automatic bank tagging, live session monitoring with keystroke capture, and centralized storage of domain visits, conversion rates, visitor details and card data. The panel is deployed in Docker through an automated installer that uses SSH credentials and auto-provisions the SQL database and API services; hosting is supplied by affiliates.

After card capture, the operator steers the victim to a matching 3D Secure/OTP page. The kit has 36 financial-institution impersonation templates for 2FA/3DS pages, customizable by geolocation. The victim enters the one-time code on the spoofed page and the operator relays it to the legitimate 3DS site to authorize a fraudulent transaction or take over the account. A fake turnstile loading page serves as a distraction during the relay, and the victim is then shown a fake confirmation page. Collected data is retained in the panel for re-targeting.

Group-IB states that full IOCs are restricted to its customers (Threat Intelligence portal), and the report names no individual actors, domains, IPs or hosting providers. No CVEs are involved; the kit abuses legitimate software (WordPress, WooCommerce, socket.io, Docker) and legitimate platforms (Facebook, TikTok, Telegram). Attribution is therefore unknown. The IOC set below is limited to publicly stated artifacts (tooling, platforms, behaviors, impersonated brands).

MITRE ATT&CK techniques by tactic:

Collection

T1056.001 Keylogging

Command and Control

T1071.001 Web Protocols

Credential Access

T1111 Multi-Factor Authentication Interception; T1557 Adversary-in-the-Middle

Execution

T1204.001 Malicious Link

Initial Access

T1566.003 Spearphishing via Service

Resource Development

T1583.001 Domains; T1585.001 Social Media Accounts; T1587.001 Malware

Impact

T1657 Financial Theft

Defense Evasion

T1684.001 Impersonation

Remediation

Immediate actions

  • Monitor for lookalike domains impersonating your brand and initiate takedown before campaigns scale
  • Monitor for unusual checkout patterns on e-commerce properties
  • If a card was entered on a suspect shop, immediately notify the bank or card issuer

Workarounds

  • Verify suspicious shop links with UrlScan.io, VirusTotal or ScamAdviser before purchase
  • Treat any OTP/3DS prompt that appears outside the bank's own app or flow as suspicious

Longer-term hardening

  • Integrate threat intelligence on PhaaS kits and impersonated-brand storefronts into brand-protection workflows
  • Educate customers that steep, limited-time discounts on social media are a red flag

Affected and fixed versions

  • LEGO — Consumer brand impersonated by fake storefronts (no vulnerability)
  • Calvin Klein — Consumer brand impersonated by fake storefronts (no vulnerability)
  • Aeon Malaysia — Retailer impersonated by fake storefronts (no vulnerability)
  • Various financial institutions (36 templates) — 3D Secure / OTP pages spoofed for MFA relay (no vulnerability)

Timeline

  • Kit offered as PhaaS in Telegram communities from 300 USDT per month with subscription plans and add-ons (start of sales not dated precisely; observed from October 2025 activity window)
  • Earliest of the 258 phishing pages Group-IB identified since October 2025 (month precision)
  • Milk Dragon (NaiLong) PhaaS kit becomes active (Group-IB: 'since October 2025'; exact day not stated, month precision)
  • Group-IB states full IOC list is available only to customers via its Threat Intelligence portal; no domains, IPs or hosting providers published
  • Group-IB publishes 'Milk Dragon: Huge Discounts on Social Media? Think Twice Before You Buy'
  • As of Group-IB's report, 258 phishing pages, victims in 66 countries, 21 impersonated consumer brands and 36 financial-institution MFA templates

Sources cited

Previous and next in phishing

Drafted by the Threadlinqs research pipeline from 7 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