Skip to navigation

Threat reportVulnerabilityTL-2026-2172

Suspected Chinese-Speaking Threat Actor Exploits ownCloud and WordPress Flaws to Steal Philippine Nuclear and Naval Data

criticalACTIVE

A suspected Chinese-speaking operator chained CVE-2023-49105 (ownCloud pre-signed URL authentication bypass) and CVE-2024-28000 (LiteSpeed Cache WordPress privilege escalation) to breach a Philippine nuclear research body and a naval marine-engineering contractor, stealing roughly 9 GB of reactor, personnel, and encrypted-credential data plus 195 MB from the contractor's WordPress site.

Key facts

IdentifiersTL-2026-2172, CVE-2023-49105, CVE-2024-28000
SeverityCRITICAL
CVSS9.8 (CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H)
EPSSCVE-2024-28000: 68.2% (99.2th percentile); CVE-2023-49105: 11.0% (95.5th percentile)
Affected productsownCloud ownCloud Server; LiteSpeed Technologies LiteSpeed Cache (WordPress plugin)
Weaknesses (CWE)CWE-287, CWE-266, CWE-665
Nation-state nexusChina
MalwareMeterpreter (Android), Mettle
ToolsMettle, Sliver - S0633, rockyou.txt
MITRE ATT&CK techniquesT1020 Automated Exfiltration, T1059.006 Command and Scripting Interpreter: Python, T1068 Exploitation for Privilege Escalation, T1071 Application Layer Protocol: Web Protocols, T1074.001 Data Staged: Local Data Staging, T1074.002 Data Staged: Remote Data Staging, T1083 File and Directory Discovery, T1105 Ingress Tool Transfer, T1110.001 Brute Force: Password Guessing, T1136.001 Create Account: Local Account, T1190 Exploit Public-Facing Application, T1204 User Execution: Malicious File and 9 more in the attack flow
MotivationESPIONAGE
Target sectorsgovernment administration, nuclear, defense, maritime, research
Target regionsSoutheast Asia, Asia-Pacific
Detection rules9 (1 shown free)
Sources8 cited sources

Detection coverage

As of 2026-09-02, Threadlinqs Intelligence publishes 9 detection rule(s) for TL-2026-2172 across Splunk SPL, Microsoft KQL and Sigma, covering 40 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: ownCloud CVE-2023-49105 Pre-Signed URL Forgery & WebDAV PROPFIND Enumeration (SPL)

Detects exploitation of CVE-2023-49105, the ownCloud pre-signed URL/WebDAV signing bypass, by identifying forged OC-Signature requests against /remote.php/dav/files/<account>/<path>, bursts of PROPFIND (Depth: 1) enumeration across multiple paths for a single account, and the throttled 1.5-6 second request cadence the recovered oc_hm_dl.py / oc_vps_download.py scripts use to evade rate-limiting.

  • Languages: Splunk SPL, Microsoft KQL, Sigma
  • ATT&CK: T1190, T1083, T1213
  • Data sources: owncloud_web_access_logs, reverse_proxy_waf_logs, webdav_access_logs
  • Severity: high
  • Confidence: high

Splunk SPL

index=web (sourcetype=owncloud_access OR sourcetype=access_combined OR sourcetype="nginx:plus:kv") earliest=-24h
| rex field=_raw "\"(?<http_method>GET|PROPFIND)\s+(?<uri_path>/remote\.php/dav/files/[^\s\?]+)"
| where isnotnull(uri_path)
| eval has_oc_sig=if(match(_raw, "OC-Signature="), 1, 0)
| where has_oc_sig=1
| bin _time span=5m
| stats count as request_count, dc(uri_path) as distinct_paths, values(http_method) as methods, min(_time) as first_seen, max(_time) as last_seen by src_ip, _time
| where request_count >= 5 AND (like(methods, "%PROPFIND%") OR distinct_paths >= 3)
| eval avg_interval_sec=round((last_seen-first_seen)/request_count, 1)
| where avg_interval_sec >= 1 AND avg_interval_sec <= 8
| sort - request_count

Microsoft KQL

CommonSecurityLog
| where TimeGenerated > ago(24h)
| where RequestURL has "/remote.php/dav/files/"
| where RequestURL has "OC-Signature" or AdditionalExtensions has "OC-Signature"
| extend HttpMethod = tostring(column_ifexists("RequestMethod", "")), SourceIP = tostring(SourceIP)
| summarize RequestCount = count(), DistinctPaths = dcount(RequestURL), Methods = make_set(HttpMethod), FirstSeen = min(TimeGenerated), LastSeen = max(TimeGenerated) by SourceIP
| extend AvgIntervalSec = datetime_diff('second', LastSeen, FirstSeen) / RequestCount
| where RequestCount >= 5 and (Methods has "PROPFIND" or DistinctPaths >= 3)
| where AvgIntervalSec between (1 .. 8)
| project SourceIP, RequestCount, DistinctPaths, Methods, AvgIntervalSec, FirstSeen, LastSeen

Sigma

title: ownCloud CVE-2023-49105 Pre-Signed URL Forgery / PROPFIND Enumeration
id: 8a1f2c4e-6b3d-4a9f-9e12-1d5c7f0a3b41
status: experimental
description: Detects forged OC-Signature WebDAV requests and PROPFIND (Depth 1) enumeration bursts against /remote.php/dav/files/ consistent with CVE-2023-49105 exploitation using an empty PBKDF2-HMAC-SHA512 signing secret.
references:
  - https://hunt.io/blog/chinese-speaking-operator-philippine-nuclear-naval-contractor
  - https://nvd.nist.gov/vuln/detail/cve-2023-49105
author: Threadlinqs Team
date: 2026/08/27
tags:
  - attack.initial-access
  - attack.discovery
  - attack.collection
  - attack.t1190
  - attack.t1083
  - attack.t1213
  - cve.2023.49105
logsource:
  category: webserver
detection:
  selection_path:
    cs-uri-stem|contains: '/remote.php/dav/files/'
  selection_sig:
    cs-uri-query|contains: 'OC-Signature'
  selection_propfind:
    cs-method: 'PROPFIND'
  condition: selection_path and selection_sig and selection_propfind
falsepositives:
  - Legitimate ownCloud desktop or mobile sync clients performing routine PROPFIND sync operations
  - Backup or migration tooling issuing signed WebDAV requests at a similar cadence
level: high

Known false positives:

  • Legitimate ownCloud desktop/mobile sync clients performing bulk PROPFIND during initial sync
  • Backup or migration tooling issuing signed WebDAV requests at a similar cadence

Tuning notes: Baseline normal ownCloud desktop-client PROPFIND sync volume per account before enabling the >=5 requests/5m threshold; exclude known sync-client user agents and allowlisted service accounts, and confirm the instance has no signing key configured (the root cause of CVE-2023-49105) as corroborating context.

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)
  • ownCloud CVE-2023-49105 Pre-Signed URL Forgery & WebDAV PROPFIND Enumeration (KQL) · unlock with TL Creds
  • LiteSpeed Cache CVE-2024-28000 Privilege Escalation & XML-RPC Brute Force (KQL) · unlock with TL Creds
  • Post-Exploitation ELF Loader & Mettle/Sliver C2 Staging Infrastructure Callback (KQL) · unlock with TL Creds
  • ownCloud CVE-2023-49105 Pre-Signed URL Forgery & WebDAV PROPFIND Enumeration (Sigma) · unlock with TL Creds
  • LiteSpeed Cache CVE-2024-28000 Privilege Escalation & XML-RPC Brute Force (Sigma) · unlock with TL Creds
  • Post-Exploitation ELF Loader & Mettle/Sliver C2 Staging Infrastructure Callback (Sigma) · unlock with TL Creds
  • LiteSpeed Cache CVE-2024-28000 Privilege Escalation & XML-RPC Brute Force (SPL) · unlock with TL Creds
  • Post-Exploitation ELF Loader & Mettle/Sliver C2 Staging Infrastructure Callback (SPL) · unlock with TL Creds

Indicators of compromise

As of 2026-09-02, Threadlinqs Intelligence tracks 40 indicators of compromise for TL-2026-2172.

Sample indicators (3 of 40)

  • IP address: 31.58.209[.]241
  • Domain: fine-work-team[.]com
  • File hash: 7447d0d0c34779d4c519823b39bf6ddc16d2b34a226b82ee69da6f5b4a77ad82

Defanged, TLP:CLEAR. Full set of 40 indicators: unlock with TL Creds.

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

Technical analysis

Hunt.io discovered the campaign after finding an openly accessible attacker-controlled server staging Python exploit scripts, an ELF loader, and a Mettle payload alongside the stolen files.

Hunt.io identified an open directory on 31.58.209[.]241:8000 on 2026-08-13 that exposed the tooling and take of an intrusion set targeting Philippine government-linked organizations. The initial foothold against the nuclear research body's ownCloud deployment exploited CVE-2023-49105, a critical authentication bypass (CVSS 9.8) in the pre-signed URL/WebDAV signing mechanism: when an ownCloud instance has no signing key configured, the server's signing routine silently falls back to an empty secret, letting anyone who knows a valid username forge a PBKDF2-HMAC-SHA512-signed WebDAV request (10,000 iterations, empty b"" key) and retrieve or manipulate that user's files without ever supplying credentials.

Five near-identical Python scripts recovered from the staging server implement this technique. Four target a single ownCloud account each, issuing GET requests against /remote.php/dav/files/<account>/<path> with a forged OC-Signature and randomized 3-6 second delays between requests to evade rate-limiting. The fifth, oc_vps_download.py, is a more advanced enumerator that issues WebDAV PROPFIND requests with Depth: 1 to recursively map account file trees, tightens its delay to 1.5-3.5 seconds, and logs every retrieval to /root/oc_download.log. Using this access the operator staged roughly 9 GB of data referenced in a CSV manifest named after the victim's parent ministry; Hunt.io itself recovered 176 files (~372 MB) across five categories spanning reactor-operations records, radiation-safety documentation, strategic/IT planning material, and 130 MB of personnel/PII, plus a KEYS subfolder containing KeePass databases, AxCrypt-encrypted files, and BitLocker recovery keys. A separate 192 MB SQL dump of a ZKTeco BioTime attendance/personnel database, also recovered from the top-level of the same open directory, referenced additional Philippine government science and research organizations.

Against the naval marine-engineering and shipbuilding contractor's WordPress site, the operator exploited CVE-2024-28000, an unauthenticated privilege-escalation flaw (CVSS 9.8) in the LiteSpeed Cache plugin. The flaw lets an attacker who obtains or brute-forces a debug-log hash spoof the plugin's role-simulation feature to impersonate an administrator, then create a new administrator account via the /wp-json/wp/v2/users REST API endpoint. Compiled Go exploit binaries (wp28000, wp28000_cp) and their source (main.go) were found alongside brute_xmlrpc.py, which brute-forced WordPress credentials via XML-RPC wp.getUsersBlogs calls using the rockyou.txt wordlist, and brute_result.txt logging successful hits. Three archives totaling 195 MB -- the full site tree, a database dump, and the media library -- were staged from this intrusion.

The same 31.58.209[.]241 host also served a stage-1 ELF loader (multi_backupd, SHA-256 7447d0d0c34779d4c519823b39bf6ddc16d2b34a226b82ee69da6f5b4a77ad82) and, from port 8090, a Mettle (portable Meterpreter) stage-two payload (stage2_payload.bin, SHA-256 10df3451915ea35bcb17efe121415f24182680e2d07fc09df07ee695072104c1). A Sliver C2 framework installation was also present on the server, though Hunt.io did not directly attribute its use to a specific intrusion step. Additional service fingerprints on the host included a cloned ownCloud login page on port 80, OpenSSH 9.6p1 Ubuntu on port 22, a Python 3.12.3 BaseHTTP responder on port 8080, and a raw TCP listener on port 54329.

Attribution rests on circumstantial but consistent evidence: stolen files were sorted into folders labelled in simplified Chinese (财务/Finance, 辐射安全/Radiation Safety, 核材料账目/Nuclear Material Accounts, IT规划/IT Planning, 系统文档/System Documentation), and script docstrings describe "low-speed download" operations against nuclear-material and radiation-safety documents in Chinese. No named group was attributed. Hunt.io disclosed its findings to CERT-PH under TLP:AMBER, which notified the affected organizations before the embargo lifted on 2026-08-25; Hunt.io published its report on 2026-08-26, assessing with medium confidence that the activity reflects targeted rather than opportunistic collection, given the reactor/naval-specific targeting despite the underlying vulnerabilities being broadly exploitable.

MITRE ATT&CK techniques by tactic:

Exfiltration

T1020 Automated Exfiltration

Execution

T1059.006 Command and Scripting Interpreter: Python; T1204 User Execution: Malicious File

Privilege Escalation

T1068 Exploitation for Privilege Escalation

Command and Control

T1071 Application Layer Protocol: Web Protocols; T1105 Ingress Tool Transfer

Collection

T1074.001 Data Staged: Local Data Staging; T1074.002 Data Staged: Remote Data Staging; T1213 Data from Information Repositories; T1560 Archive Collected Data

Discovery

T1083 File and Directory Discovery

Credential Access

T1110.001 Brute Force: Password Guessing; T1552.001 Unsecured Credentials: Credentials In Files

Persistence

T1136.001 Create Account: Local Account

Initial Access

T1190 Exploit Public-Facing Application

Defense Evasion

T1218 System Binary Proxy Execution: Mshta

Resource Development

T1583.003 Acquire Infrastructure: Virtual Private Server; T1587.001 Develop Capabilities: Malware; T1587.004 Develop Capabilities: Exploits; T1588.002 Obtain Capabilities: Tool; T1608.002 Stage Capabilities: Upload Tool

Remediation

Patches

  • ownCloud Server 10.13.1 (CVE-2023-49105) / 10.13.3 vendor-recommended bundle
  • LiteSpeed Cache plugin 6.4

Immediate actions

  • Patch ownCloud Server to 10.13.1 (fixes CVE-2023-49105) or the vendor-recommended 10.13.3 (also fixes CVE-2023-49103/CVE-2023-49104)
  • Update the LiteSpeed Cache WordPress plugin to version 6.4 or later
  • Rotate all credentials, signing keys, and secrets stored in or accessible via unpatched ownCloud or WordPress instances
  • Audit WordPress administrator accounts for entries created without a known owner via the REST API
  • Block and hunt for connections to/from 31.58.209.241

Workarounds

  • Explicitly configure an ownCloud signing key on every instance so pre-signed URL validation cannot fall back to an empty secret
  • Disable or restrict the LiteSpeed Cache role-simulation/debug-log feature until the plugin is patched

Longer-term hardening

  • Deploy WebDAV/REST API anomaly detection for PROPFIND enumeration bursts and mismatched OC-Signature/OC-Credential headers
  • Disable XML-RPC on WordPress installations that do not require it, or restrict access to trusted IPs
  • Enforce MFA and request rate-limiting on all internet-facing ownCloud and WordPress logins
  • Isolate nuclear/critical-infrastructure research data behind access controls independent of the file-sharing platform's own authentication

Affected and fixed versions

  • ownCloud — ownCloud Server
    Vulnerable versions: 10.6.0-10.13.0
    Fixed in: 10.13.1 (CVE-2023-49105 fix); 10.13.3 (vendor-recommended, also fixes CVE-2023-49103/CVE-2023-49104)
  • LiteSpeed Technologies — LiteSpeed Cache (WordPress plugin)
    Vulnerable versions: 1.9-6.3.0.1
    Fixed in: 6.4

Timeline

Last updated · 3 updates · revalidated 3× · latest source

  • ownCloud discloses CVE-2023-49105, a critical pre-signed URL authentication bypass in ownCloud Server versions before 10.13.1, alongside CVE-2023-49103 and CVE-2023-49104.
  • LiteSpeed Cache plugin version 6.4 is released, patching the unauthenticated privilege-escalation flaw later assigned CVE-2024-28000.
  • Patchstack and LiteSpeed disclose CVE-2024-28000, an unauthenticated privilege-escalation flaw in the LiteSpeed Cache WordPress plugin (versions 1.9-6.3.0.1), patched in version 6.4.
  • Latest records recovered from the ZKTeco BioTime attendance-system SQL dump on the actor's staging server are dated December 2024, indicating the actor's data holdings were current through at least that period.
  • A Japanese-language security blog flags a NoChain malware service-worker script later found co-located on the same staging infrastructure, though not confidently linked to the nuclear/naval intrusion operator.
  • Hunt.io identifies an open directory on locked-ioc-13:8000 hosting five custom Python exploit scripts, the multi_backupd ELF loader, a Mettle stage-two payload, and staged data stolen from a Philippine nuclear research body and a naval marine-engineering contractor.
  • Hunt.io verifies active EtherHiding-style compromise via analysis of the Ethereum smart contract (0x58460d0b3d4d6b03761c89120393c0c676676496) used to deliver ClickFix loader content.
  • Hunt.io's coordinated disclosure to CERT-PH under TLP:AMBER concludes; CERT-PH had notified the affected Philippine organizations ahead of public release.
  • Hunt.io publishes its full technical report on the campaign at 16:55 UTC.
  • CISA adds CVE-2023-49105 (ownCloud) to its Known Exploited Vulnerabilities catalog, citing confirmed active exploitation tied to this campaign against the Philippine nuclear research agency.
  • Cyber Security News republishes and summarizes the Hunt.io findings, bringing the campaign to broader public attention.
  • Security Affairs and other outlets publish summary coverage of the Hunt.io report, drawing renewed attention to the campaign.
  • CISA BOD 26-04 remediation deadline for U.S. federal agencies to patch CVE-2023-49105.

Update history

Sources cited

Previous and next in vulnerability

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