Skip to navigation

Threat reportVulnerabilityTL-2026-2356

StyleSmuggler — Magento Open Source and Adobe Commerce Unauthenticated RCE 0-Day Under Active Exploitation

criticalACTIVE

A critical unauthenticated remote code execution vulnerability dubbed 'StyleSmuggler' affecting all current versions of Magento Open Source and Adobe Commerce (2.4.6-2.4.9) is under active exploitation since September 4, 2026. Discovered by Sansec, the two-stage attack abuses GraphQL 'styles' property injection to poison log files, then triggers execution when Magento renders Payment Transaction Failed Reminder emails.

Key facts

IdentifiersTL-2026-2356
Also known asStyleSmuggler
SeverityCRITICAL
CVSS9.8 (CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H)
First disclosed
Affected productsAdobe Adobe Commerce; Magento (Open Source) Magento Open Source
Weaknesses (CWE)CWE-94, CWE-74, CWE-502
MITRE ATT&CK techniquesT1005 Data from Local System, T1027 Obfuscated Files or Information, T1036 Masquerading, T1053 Scheduled Task/Job, T1057 Process Discovery, T1059 Command and Scripting Interpreter, T1095 Non-Application Layer Protocol, T1190 Exploit Public-Facing Application, T1219 Remote Access Tools, T1480 Execution Guardrails, T1539 Steal Web Session Cookie, T1552 Unsecured Credentials and 3 more in the attack flow
MotivationUNKNOWN
Target sectorsecommerce, retail
Detection rules9 (1 shown free)
Sources6 cited sources

Detection coverage

As of 2026-09-06, Threadlinqs Intelligence publishes 9 detection rule(s) for TL-2026-2356 across Splunk SPL, Microsoft KQL and Sigma, covering 19 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: StyleSmuggler — Masqueraded Kernel Thread Backdoor (SPL)

Detects the StyleSmuggler Rust backdoor process masquerading as a legitimate Linux kernel worker thread named [kworker/u:8:0]. Genuine kernel kworker threads are always owned by root (UID 0) with zero resident memory. The StyleSmuggler implant runs under the Magento website user account (e.g., www-data, nginx, magento) and has measurable non-zero RSS. This detection also identifies direct crontab spool file modifications (bypassing crontab -e) revealing the persistence mechanism: cron entries executing .gvfsd/gvfsd-user every 5 minutes. Includes detection of the PHP dropper probing multiple execution functions (proc_open, exec, system, passthru, shell_exec, popen) as observed during post-exploitation.

  • Languages: Splunk SPL, Microsoft KQL, Sigma
  • ATT&CK: T1036, T1036.005, T1053, T1053.003, T1564, T1027, T1480
  • Data sources: linux_process_monitoring, linux_auditd, file_integrity_monitoring, process_command_line_parameters
  • Severity: critical
  • Confidence: high

Splunk SPL

index=linux (sourcetype=ps OR sourcetype=auditd:syscall OR sourcetype=linux_secure)
  | regex process="\[?kworker"
  | eval is_bracketed = if(match(process, "^\\[kworker"), 1, 0)
  | eval user = lower(user)
  | where is_bracketed=1 AND user!="root"
  | eval resident_mb = round(if(isnull(rss), 0, rss)/1024, 2)
  | where resident_mb > 0
  | eval threat_name="StyleSmuggler - Masqueraded kworker Backdoor", mitre_techniques="T1036 T1053 T1564 T1027 T1480"
  | table _time, host, user, process, pid, resident_mb, cpu, cmdline
  | sort - _time

| append [search index=linux sourcetype=linux_audit syscall=open file=*/var/spool/cron/crontabs/* OR file=*crontabs* command=*.gvfsd* OR command=*gvfsd-user*
  | eval threat_name="StyleSmuggler - Cron Spool Persistence"
  | table _time, host, user, file, command, pid]
  | sort - _time

Microsoft KQL

Syslog
| where TimeGenerated > ago(24h)
| where SyslogMessage matches regex @"\[kworker"
| where SyslogMessage !contains_cs "root"
| where SyslogMessage matches regex @"rss\s*[:=]\s*[1-9]"
| extend ProcessName = extract(@"(\[kworker[^\]]*\])", 1, SyslogMessage)
| extend ProcessUser = extract(@"user(?:=|\s+)(\S+)", 1, SyslogMessage)
| extend RSSValue = extract(@"rss\s*[:=]\s*(\d+)", 1, SyslogMessage)
| project TimeGenerated, Computer, ProcessName, ProcessUser, RSSValue, SyslogMessage
| sort by TimeGenerated desc

Sigma

title: StyleSmuggler - Masqueraded Kernel Worker Backdoor Process
alias: StyleSmuggler_kworker_Masquerade
id: 4b8c0f92-6d3e-5a92-b4c7-8e9f0a1b2c3d
status: experimental
description: Detects the StyleSmuggler Rust backdoor process named [kworker/u:8:0] running under a non-root website user account with non-zero resident memory, indicating process masquerading
references:
    - https://sansec.io/research/stylesmuggler
    - https://cybernoz.com/unpatched-magento-and-adobe-commerce-zero-day-exploited-to-backdoor-online-stores/
author: Threadlinqs Team
date: 2026-09-06
tags:
    - attack.t1036
    - attack.t1036.005
    - attack.t1053
    - attack.t1053.003
    - attack.t1564
    - attack.t1027
    - attack.t1480
    - attack.persistence
    - attack.defense_evasion
logsource:
    category: process_creation
    product: linux
service: auditd
detection:
    selection_kworker:
        Image|endswith: 'kworker'
        CommandLine|contains: '[kworker'
        User|startswith: 'www-data'
    selection_auditd:
        type: 'PROCTITLE'
        a0|contains: '[kworker'
        auid|gt: 1000
    condition: selection_kworker or selection_auditd
falsepositives:
    - None expected — genuine kernel kworker threads are always root-owned with zero resident memory
level: critical

Known false positives:

  • None expected — genuine kernel kworker threads are always root-owned with zero resident memory

Tuning notes: Correlate with the web server user account (www-data, nginx, magento). Investigate any process with a bracketed kworker name not owned by root. Verify resident memory > 0 via /proc/<pid>/status VmRSS. Check /var/spool/cron/crontabs/ and ~/.local/share/.gvfsd/ for additional persistence artifacts. Hash both /proc/<pid>/exe and the on-disk file as Disrex Group observed differences between in-memory and on-disk binaries.

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)
  • StyleSmuggler — GraphQL Styles Array Injection (KQL) · unlock with TL Creds
  • StyleSmuggler — Masqueraded Kernel Thread Backdoor (KQL) · unlock with TL Creds
  • StyleSmuggler — Multi-Modal C2 and Session Data Theft (KQL) · unlock with TL Creds
  • StyleSmuggler — GraphQL Styles Array Injection (Sigma) · unlock with TL Creds
  • StyleSmuggler — Masqueraded Kernel Worker Backdoor (Sigma) · unlock with TL Creds
  • StyleSmuggler — Multi-Modal C2 and Redis Session Theft (Sigma) · unlock with TL Creds
  • StyleSmuggler — GraphQL Styles Array Injection (SPL) · unlock with TL Creds
  • StyleSmuggler — Multi-Modal C2 and Session Data Theft (SPL) · unlock with TL Creds

Indicators of compromise

As of 2026-09-06, Threadlinqs Intelligence tracks 19 indicators of compromise for TL-2026-2356.

Sample indicators (3 of 19)

  • IP address: 99.84.67[.]186
  • Domain: 247[.]cdnflare[.]xyz
  • File hash: e315687a1dfe61ef4a5a5642214db6d3b2b05d81391285eebc2af664641a26a7

Defanged, TLP:CLEAR. Full set of 19 indicators: unlock with TL Creds.

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

Technical analysis

A Rust-based backdoor disguised as a Linux kernel thread ([kworker/u:8:0]) is deployed for persistence and session data theft via local Redis connections. No CVE has been assigned and no official patch is available from Adobe as of September 6, 2026.

StyleSmuggler is an unpatched (0-day) remote code execution vulnerability in Magento Open Source and Adobe Commerce, under active exploitation since September 4, 2026. The vulnerability requires no authentication and affects all current versions of the platform, including the latest 2.4.9 release. Sansec, a Dutch e-commerce security firm, discovered the campaign and independently reproduced the full unauthenticated attack chain on clean installations of Magento Open Source 2.4.7, 2.4.8, and 2.4.9.

The attack operates in two stages. Stage 1 — code injection: attackers send crafted GraphQL requests manipulating 'styles' array properties to evade input sanitization and inject malicious PHP code into files Magento writes during normal operations. Observed injection points include the var/report/ directory (payment failure reports) and var/log/system.log. The injected payload contains a crafted directive that forces a chain of Magento's own classes — specifically the dependency-injection (DI) compiler code under setup/src/Magento/Setup/Module/Di/Code/ — to execute code intended only for the CLI compiler, ultimately including the attacker-poisoned log file.

Stage 2 — trigger: the attacker deliberately triggers Magento's 'Payment Transaction Failed Reminder' transactional email. When Magento renders that email template, the injected PHP code executes server-side. No user interaction is required — nobody needs to open or even receive the email for the attack to succeed. The attack works even when email delivery fails.

Once the PHP injection executes, a PHP dropper cycles through six different PHP functions (including proc_open) until it finds one capable of spawning a process, then downloads and launches a persistent implant. The implant is a statically linked Rust binary of approximately 1.9 MB, compiled for both x86-64 and ARM64 architectures. It masquerades as a Linux kernel thread named [kworker/u:8:0] — setting its command line to the literal bracketed string so process-table checks against the comm field match nothing. Genuine kernel worker threads are root-owned with zero resident memory, so any bracketed kworker running under a website user account with measurable resident memory is the implant.

The backdoor is persisted via a cron entry written directly into the crontab spool file (/var/spool/cron/crontabs/) to bypass standard system logging, executing every 5 minutes. In one compromised store, the implant had the same cron line repeated 1,728 times and re-added it within one second of removal. The implant's network behavior is notably stealthy: on one confirmed store, it made no outbound internet connections at all, instead opening 28 simultaneous connections to the site's own local Redis instance (port 6379) to read live Magento session data. This design allows the malware to operate almost invisibly to network-based monitoring. On other stores, the backdoor connected to C2 infrastructure via WebSocket over TLS (port 443) and custom NTP-shaped UDP traffic on port 123.

Disrex Group, a Magento hosting firm handling two breached stores, independently analyzed the attack chain. Their packet captures (each >200 MB taken with the implant live) contained zero packets to the C2 addresses Sansec listed, confirming the Redis-only operational mode. They also noted that the binary running in memory differed from the file on disk on one store — indicating defenders must hash both the running process from /proc/<pid>/exe and the on-disk file. Disrex found 26 distinct source IPs across their two compromised stores, primarily a residential proxy pool, indicating broad exploitation infrastructure rather than a single attacker.

No official CVE has been assigned. Adobe's next scheduled security bulletin is September 8, 2026, but it remains unknown whether that release will address this vulnerability. Unofficial mitigations have been published by Disrex Group, ProxiBlue, and Graycore, all of which are characterized as hardening measures rather than definitive fixes. Server-level mitigations including disabling PHP's proc_open function and mounting temporary directories with noexec have been shown effective at preventing the dropper from launching its payload. Disabling the GraphQL endpoint entirely is recommended for stores not relying on headless or PWA storefronts.

MITRE ATT&CK techniques by tactic:

Collection

T1005 Data from Local System

Defense Evasion

T1027 Obfuscated Files or Information; T1036 Masquerading; T1480 Execution Guardrails; T1564 Hide Artifacts

Persistence

T1053 Scheduled Task/Job

Privilege Escalation

T1053 Scheduled Task/Job

Discovery

T1057 Process Discovery

Execution

T1059 Command and Scripting Interpreter

Command and Control

T1095 Non-Application Layer Protocol; T1219 Remote Access Tools; T1571 Non-Standard Port; T1573 Encrypted Channel

Initial Access

T1190 Exploit Public-Facing Application

Credential Access

T1539 Steal Web Session Cookie; T1552 Unsecured Credentials

Remediation

Patches

  • No official patch available as of September 6, 2026
  • Unofficial hardening patches available from Disrex Group, ProxiBlue, and Graycore (composer-patches, GitHub)
  • Adobe's next scheduled security bulletin: September 8, 2026 — unconfirmed if StyleSmuggler is covered

Immediate actions

  • Disable GraphQL endpoint for stores not using headless or PWA storefronts
  • Add proc_open to PHP's disable_functions list
  • Mount /tmp, /var/tmp, and /dev/shm with noexec flag
  • Deploy nginx/Apache rules blocking POST to /graphql with styles[] array parameters
  • Scan for backdoor files in ~/.local/share/.gvfsd/ and /tmp/.kw_* paths
  • Check for [kworker/u:8:0] processes owned by non-root users with non-zero resident memory
  • Search var/report/ for X_TRACE_ markers and var/log/system.log for PHP code or base64 payloads
  • Inspect crontab spool files for suspicious gvfsd entries
  • Monitor for bursts of Payment Transaction Failed Reminder emails
  • Check for unexpected simultaneous Redis connections from non-Redis processes

Workarounds

  • Disable GraphQL endpoint entirely for non-headless storefronts
  • Sansec Shield provides real-time blocking of exploitation attempts (rules deployed Sept 5, 07:15 UTC)
  • eComscan 1.9.7 can detect known implant variants (note: scans document root; implant may reside in home directory above it)
  • Apply Disrex Group's web-server rule blocks for current attack traffic patterns
  • Apply Graycore's module hardening email template block directives, grid URL generation, and Web API error reports

Longer-term hardening

  • Apply Adobe's official patch when released (pending September 8, 2026 security bulletin)
  • Implement GraphQL request validation and schema hardening
  • Deploy WAF rules with behavioral detection for log poisoning patterns
  • Transition to file integrity monitoring on var/report/ and var/log/ paths
  • Implement runtime process monitoring for masqueraded kernel thread names
  • Establish baseline of expected Redis connection counts for e-commerce infrastructure

Affected and fixed versions

  • Adobe — Adobe Commerce
    Vulnerable versions: 2.4.6-p15; 2.4.7; 2.4.7-p2; 2.4.8; 2.4.9
  • Magento (Open Source) — Magento Open Source
    Vulnerable versions: 2.4.7; 2.4.8; 2.4.9

Timeline

  • Adobe's most recent Commerce security bulletin published, pre-dating the StyleSmuggler campaign
  • Store A (Sansec Shield customer running Magento 2.4.8) compromised — hit approximately 23:10 UTC, hours before Shield blocking rules were deployed
  • Sansec identified the ongoing attack campaign at approximately 22:40 UTC
  • First confirmed StyleSmuggler exploitation observed by Sansec at approximately 22:20 UTC
  • Sansec published public disclosure of StyleSmuggler, publishing early 'because stores are being compromised right now'
  • ProxiBlue and Graycore published unofficial hardening patches on GitHub and Packagist
  • Disrex Group published incident response repository with nginx/Apache blocking rules, code-level DI scanner guards, and cleanup guidance; contained both compromised stores approximately 11-14 hours after first contact
  • Sansec independently reproduced the full unauthenticated attack chain on clean installations of Magento Open Source 2.4.7, 2.4.8, and 2.4.9
  • Sansec Shield blocking rules went live at 07:15 UTC, blocking StyleSmuggler exploitation in real time
  • Store B (running Magento 2.4.7-p2) first compromised at approximately 00:55 UTC
  • Widespread industry awareness; Adobe has not issued an advisory, CVE identifier, or patch; no official response from Adobe as of this date
  • Adobe's next scheduled security bulletin — unconfirmed whether this vulnerability will be addressed

Sources cited

Previous and next in vulnerability

Drafted by the Threadlinqs research pipeline from 6 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