Back to Intelligence

Operation QUICSILVER: China-Nexus VHD-Delivered Go Backdoor and FTP Dead Drop RATs — Defender's Detection and Response Guide

SA
Security Arsenal Team
August 30, 2026
10 min read

The latest Security Affairs malicious software newsletter (Round 112) aggregates three research threads that deserve immediate attention from SOC and IR teams: Operation QUICSILVER, a China-nexus espionage campaign targeting Myanmar diplomatic personnel using virtual hard disk (VHD) attachments to deliver a Go-based backdoor; a novel technique abusing FTP banners as dead drop resolvers to stage and deliver previously unseen remote access trojans; and the August 2026 state-of-play on AI-enabled malware, spanning brand abuse to agentic, self-directed malicious tooling.

None of this is theoretical. QUICSILVER is an active, targeted intrusion set against diplomatic targets — and the VHD-delivery tradecraft it uses has already been adopted by commodity actors precisely because it defeats Mark-of-the-Web-based defenses. The FTP banner dead drop technique is a C2-resilience pattern that makes infrastructure takedown dramatically harder. If your detection stack assumes attachments are files with MOTW and that C2 resolution happens over DNS or HTTPS, you have coverage gaps today.

This post breaks down the attack chains, then gives you deployable Sigma, KQL, and Velociraptor detections plus hardening guidance.

Technical Analysis

Operation QUICSILVER: VHD-Delivered Go Backdoor

Targeting: Myanmar diplomatic entities — a classic espionage target set consistent with China-nexus strategic collection requirements around regional foreign policy and negotiations.

Attack chain (defender's view):

  1. Delivery: Spear-phishing email carrying a .vhd / .vhdx (virtual hard disk) attachment. VHD/ISO container files are favored because files extracted or executed from a mounted container do not inherit the Mark-of-the-Web (MOTW) — the Zone.Identifier alternate data stream does not propagate. This silently neutralizes MOTW-dependent controls: Office Protected View, SmartScreen prompts, and many email-to-endpoint trust signals.
  2. Execution: The victim (or a lure document) mounts the VHD — Windows mounts it natively on double-click via explorer.exe / dllhost.exe. Inside the container: a lure document alongside a Go-compiled backdoor, often sideloaded or launched via shortcut (lnk) or script.
  3. Payload: A Go backdoor providing unauthorized remote access — command execution, file staging, and exfiltration. Go payloads are attractive to operators: large static binaries that evade signature engines tuned for PE metadata, cross-platform reuse, and dense, analyst-hostile disassembly.

Exploitation status: Confirmed active, targeted in-the-wild campaign. This is espionage tradecraft, not a CVE — there is no patch; the fix is architectural (attachment policy, container blocking, execution control) plus behavioral detection.

FTP Banners as Dead Drop Resolvers

This technique repurposes a legacy protocol as resilient C2 infrastructure. The operator stores encoded configuration — real C2 addresses, keys, or next-stage URLs — inside the FTP welcome banner of an attacker-controlled or abused public FTP server. The implant connects to the FTP server (TCP/21), reads the banner, decodes the resolver data, and only then beacons to the actual C2.

Why defenders should care:

  • Takedown resistance: Burn the resolved C2 and the operator just updates the banner. The resolver layer is cheap and disposable.
  • Detection blind spot: FTP is legacy enough that many egress policies either ignore it or blanket-allow it for legacy business processes. Implant-initiated port 21 traffic from a workstation is almost never legitimate in 2026.
  • Staging channel: The same FTP session can pull second-stage payloads — the research documents novel RAT families delivered through this pipeline.

AI-Enabled Malware: August 2026 Snapshot

The newsletter's third thread tracks AI-enabled malicious software maturing from brand abuse (AI-generated phishing pages and impersonation content indistinguishable from legitimate brand assets) toward agentic malware — tooling that autonomously performs reconnaissance, adapts its execution path to the host, and makes lateral-movement decisions without operator tasking. The defensive implication: static indicators decay faster than ever, and behavior-based detection (execution chains, egress anomalies, identity abuse) is now the only durable layer.

Detection & Response

Sigma Rules

YAML
---
title: Virtual Disk Image File Created in User-Writable Location
description: Detects creation of VHD/VHDX/ISO/IMG container files in user directories, consistent with phishing-delivered container payloads used in Operation QUICSILVER. These containers bypass Mark-of-the-Web propagation and are rarely created by end users.
references:
  - https://securityaffairs.com/198133/malware/security-affairs-malware-newsletter-round-112.html
  - https://attack.mitre.org/techniques/T1566/001/
author: Security Arsenal
date: 2026/08/10
tags:
  - attack.initial_access
  - attack.t1566.001
logsource:
  category: file_event
  product: windows
detection:
  selection_ext:
    TargetFilename|endswith:
      - '.vhd'
      - '.vhdx'
      - '.iso'
      - '.img'
  selection_path:
    TargetFilename|contains:
      - '\Users\'
      - '\AppData\Local\Temp\'
      - '\Downloads\'
      - '\Desktop\'
  filter_archive_tool:
    Image|endswith:
      - '\7z.exe'
      - '\winrar.exe'
      - '\ explorer.exe'
  condition: selection_ext and selection_path and not filter_archive_tool
falsepositives:
  - IT administrators deploying virtual machines
  - Users legitimately receiving disk images from software vendors
level: high
---
title: Mount-DiskImage or Diskpart Mounting of Virtual Disk via Command Line
description: Detects interactive or scripted mounting of virtual disk images through PowerShell Mount-DiskImage or diskpart, a step in VHD-delivered malware execution chains. While double-click mounting via Explorer is more common, adversaries and scripts frequently mount images programmatically.
references:
  - https://securityaffairs.com/198133/malware/security-affairs-malware-newsletter-round-112.html
  - https://attack.mitre.org/techniques/T1059/001/
author: Security Arsenal
date: 2026/08/10
tags:
  - attack.execution
  - attack.t1059.001
logsource:
  category: process_creation
  product: windows
detection:
  selection_ps:
    CommandLine|contains:
      - 'Mount-DiskImage'
      - 'Mount-VHD'
      - 'Add-PartitionAccessPath'
  selection_diskpart:
    Image|endswith: '\diskpart.exe'
    CommandLine|contains:
      - 'vdisk'
      - 'attach'
  condition: selection_ps or selection_diskpart
falsepositives:
  - Legitimate virtualization administration by IT staff
  - Backup and recovery tooling
level: medium
---
title: Outbound FTP Connection from Non-FTP Client Process
description: Detects workstation processes initiating outbound connections to TCP/21 that are not recognized FTP clients. Consistent with implants using FTP banners as dead drop resolvers or FTP as a staging/exfiltration channel. Legitimate interactive FTP from end-user workstations is exceptionally rare in modern environments.
references:
  - https://securityaffairs.com/198133/malware/security-affairs-malware-newsletter-round-112.html
  - https://attack.mitre.org/techniques/T1071/002/
  - https://attack.mitre.org/techniques/T1102/
author: Security Arsenal
date: 2026/08/10
tags:
  - attack.command_and_control
  - attack.t1071.002
  - attack.t1102
logsource:
  category: network_connection
  product: windows
detection:
  selection:
    DestinationPort: 21
    Initiated: 'true'
  filter_known_clients:
    Image|endswith:
      - '\ftp.exe'
      - '\FileZilla.exe'
      - '\WinSCP.exe'
      - '\psftp.exe'
      - '\curl.exe'
  condition: selection and not filter_known_clients
falsepositives:
  - Legacy line-of-business applications with embedded FTP transfers (tune per-environment by Image hash)
level: high

KQL — Microsoft Sentinel / Defender Hunt

KQL — Microsoft Sentinel / Defender
// Hunt: VHD-delivered malware + FTP dead drop resolver behavior (Operation QUICSILVER / FTP banner RATs)
// Part 1: Virtual disk container files landing in user-writable paths
let ContainerFiles =
DeviceFileEvents
| where TimeGenerated > ago(14d)
| where FileName has_any (".vhd", ".vhdx", ".iso", ".img")
| where FolderPath has_any ("\\Users\\", "\\Downloads\\", "\\Desktop\\", "\\AppData\\Local\\Temp\\")
| project ContainerTime=TimeGenerated, DeviceName, FileName, FolderPath, SHA256, InitiatingProcessFileName, InitiatingProcessCommandLine;
// Part 2: Outbound FTP (TCP/21) from non-standard processes
let FtpEgress =
DeviceNetworkEvents
| where TimeGenerated > ago(14d)
| where RemotePort == 21
| where InitiatingProcessFileName !in~ ("ftp.exe", "FileZilla.exe", "WinSCP.exe", "curl.exe", "psftp.exe")
| where RemoteIPType == "Public"
| project FtpTime=TimeGenerated, DeviceName, InitiatingProcessFileName, InitiatingProcessCommandLine, RemoteIP, RemoteUrl;
ContainerFiles
| join kind=fullouter FtpEgress on DeviceName
| order by DeviceName asc
KQL — Microsoft Sentinel / Defender
// Correlation: host wrote a disk container AND spawned processes shortly after (mount + execute chain)
DeviceFileEvents
| where TimeGenerated > ago(14d)
| where FileName endswith ".vhd" or FileName endswith ".vhdx"
| extend DeviceName, FileWriteTime = TimeGenerated
| join kind=inner (
    DeviceProcessEvents
    | where TimeGenerated > ago(14d)
    | where ProcessCommandLine has_any ("Mount-DiskImage", "diskpart", "vdisk")
       or FileName in~ ("rundll32.exe", "regsvr32.exe", "wscript.exe", "mshta.exe")
) on DeviceName
| where TimeGenerated between (FileWriteTime .. FileWriteTime + 15m)
| project DeviceName, FileWriteTime, FileName, FolderPath, ProcessExecTime=TimeGenerated, FileName1=FileName, ProcessCommandLine, AccountName

Velociraptor VQL Hunt

VQL — Velociraptor
-- Hunt: VHD container artifacts and suspicious FTP (TCP/21) egress on Windows endpoints
-- Combines filesystem artifact sweep with live network state for dead drop resolver detection

SELECT * FROM foreach(
  row={
    SELECT FullPath, Size, Mtime
    FROM glob(globs=['C:/Users/*/Downloads/*.vhd',
                     'C:/Users/*/Downloads/*.vhdx',
                     'C:/Users/*/Downloads/*.iso',
                     'C:/Users/*/Desktop/*.vhd',
                     'C:/Users/*/Desktop/*.vhdx',
                     'C:/Users/*/AppData/Local/Temp/*.vhd*'])
    WHERE Mtime > now() - 1209600
  },
  query={
    SELECT FullPath AS SuspiciousContainer, Mtime AS ContainerWritten
    FROM scope()
})

SELECT Pid, Name, Path AS ProcessPath, CommandLine
FROM pslist()
WHERE CommandLine =~ 'Mount-DiskImage|diskpart|attach vdisk'

SELECT Pid, Name, Status, Laddr, Raddr
FROM netstat()
WHERE Raddr =~ ':21$'
  AND Status =~ 'ESTABLISHED|SYN'

Remediation & Verification Script

PowerShell
# Security Arsenal — QUICSILVER / VHD-Malware Endpoint Verification Script
# Run elevated on endpoints or deploy via your RMM/Intune. Read-only audit; safe for production.

# 1. Enumerate currently mounted virtual disk images
Write-Host "[*] Mounted virtual disk images:" -ForegroundColor Cyan
Get-DiskImage | Where-Object { $_.Attached -eq $true } |
    Select-Object ImagePath, Attached, DevicePath | Format-Table -AutoSize

# 2. Sweep user profiles for recently written container files (last 14 days)
$cutoff = (Get-Date).AddDays(-14)
$containers = Get-ChildItem -Path "C:\Users" -Recurse -Include *.vhd,*.vhdx,*.iso,*.img -ErrorAction SilentlyContinue |
    Where-Object { $_.LastWriteTime -gt $cutoff }
Write-Host "[*] Container files written in last 14 days: $($containers.Count)" -ForegroundColor Cyan
$containers | Select-Object FullName, Length, LastWriteTime | Format-Table -AutoSize

# 3. Audit outbound TCP/21 listeners of history via established connections (point-in-time)
Write-Host "[*] Active outbound FTP (TCP/21) connections:" -ForegroundColor Cyan
Get-NetTCPConnection -RemotePort 21 -ErrorAction SilentlyContinue |
    Where-Object { $_.State -eq 'Established' } |
    ForEach-Object {
        $proc = Get-Process -Id $_.OwningProcess -ErrorAction SilentlyContinue
        [PSCustomObject]@{ RemoteAddress=$_.RemoteAddress; Process=$proc.ProcessName; Path=$proc.Path }
    } | Format-Table -AutoSize

# 4. Verify ASR rules relevant to container/child-process abuse (Audit or Block)
Write-Host "[*] Attack Surface Reduction rule states:" -ForegroundColor Cyan
$asr = Get-MpPreference
$asr.AttackSurfaceReductionRules_Ids | ForEach-Object {
    $i = $asr.AttackSurfaceReductionRules_Ids.IndexOf($_)
    [PSCustomObject]@{ RuleId=$_; State=$asr.AttackSurfaceReductionRules_Actions[$i] }
} | Format-Table -AutoSize
# Recommended: enable "Block executable files from running unless they meet prevalence/age/trusted list"
# (rule 01443614-cd74-433a-b99e-2ecdc07bfc25) in Block mode after a 2-week audit period.

# 5. Verify Defender SmartScreen and MOTW handling are not weakened
$motw = Get-ItemProperty "HKLM:\SOFTWARE\Microsoft\Windows\CurrentVersion\Policies\Attachments" -ErrorAction SilentlyContinue
Write-Host "[*] SaveZoneInformation policy (1 = preserve MOTW): $($motw.SaveZoneInformation)" -ForegroundColor Cyan

Remediation

There is no CVE to patch here — this is adversary tradecraft, so remediation is control-based:

Email and content control

  • Block or detonate .vhd, .vhdx, .iso, and .img attachments at the email gateway. There is no legitimate business case for unsolicited disk images arriving by email; quarantine-and-review is the minimum acceptable posture.
  • Rewrite or strip container attachments from external senders by default; allow-list verified workflows only.

Endpoint execution control

  • Deploy the Sigma rules above and validate against your EDR telemetry. The FTP egress rule in particular should have near-zero baseline noise on workstations — if it fires constantly, you have a shadow-IT problem worth finding anyway.
  • Enable Defender ASR rule 01443614-cd74-433a-b99e-2ecdc07bfc25 (block untrusted/unsigned executables) after an audit-phase, and consider AppLocker/WDAC policies restricting execution from mounted volumes and user-writable paths.
  • Where feasible, disable auto-mount of disk images for standard users via policy, or gate it behind administrator approval.

Network egress

  • Block outbound TCP/21 from workstations at the perimeter. FTP belongs on tightly scoped server VLANs with explicit allow-lists, if anywhere. Log and alert on any residual attempts.
  • Treat any process-initiated FTP session as a hunt trigger: pull the binary hash, check prevalence in your EDR console and against threat intel, and inspect the session content where TLS inspection doesn't apply (FTP is plaintext — capture it).

Threat hunting and IR readiness

  • Run the KQL and VQL hunts across the last 14–30 days. Diplomatic and government-adjacent organizations — and their suppliers — should extend the lookback to 90 days given espionage dwell times.
  • If you find a mounted-image-plus-LOLBin execution chain, treat it as an intrusion, not a malware event: isolate the host, capture memory, preserve the VHD container as evidence (it contains the full delivery package), and hunt laterally for the same container hash across the fleet.

Strategic

  • The AI-enabled malware trend means indicator-driven defense has a shrinking half-life. Rebalance detection engineering effort toward behavioral chains (container mount → script interpreter → unsigned child process → anomalous egress), identity telemetry, and egress anomaly detection rather than hash/signature feeds.

Category Note for Analysts

If your organization handles diplomatic, NGO, or regional-government data touching Southeast Asia, raise QUICSILVER-style TTPs to your threat model explicitly: VHD delivery, Go backdoors, and resolver-layer C2 are the pattern to brief to your SOC this week.

Related Resources

Security Arsenal Incident Response Services AlertMonitor Platform Book a SOC Assessment incident-response Intel Hub

Is your security operations ready?

Get a free SOC assessment or see how AlertMonitor cuts through alert noise with automated triage.