What's Out There About You on the Dark Web? Scan for Leaks Now
Dark Web Profile: CyberXero
12 Mins Read
Summarize with:

Dark Web Profile: CyberXero

CyberXero is a Russian-speaking, financially motivated Initial Access Broker (IAB) active on Dark Web forums, including Dread and HackForums. The actor runs two parallel operations: an opportunistic global campaign that mass-exploits WordPress, Magento and other e-commerce platforms for access and payment data, and a directed campaign that conducts curated reconnaissance against Ukrainian energy and utilities organizations.

What sets CyberXero apart is its AI layer, with up to 51 specialized Claude Code agents on its primary workstation and a PentAGI deployment wired into a Cobalt Strike Team Server through the Anthropic API. The SOCRadar Threat Research Unit has tracked the actor’s infrastructure since July 2026, and CyberXero remained active at the time of publication.

This profile is based on the findings of our prior research into CyberXero.

Who Is CyberXero?

CyberXero came to light after SOCRadar identified an exposed open directory on an attacker-controlled HTTP server. The server held more than 90,000 files and 3,000 subdirectories from an active Linux workstation, including engagement folders named after target domains, exploitation toolkits, AI agent configurations, session logs and victim database exports.

CyberXero threat actor card
CyberXero threat actor card

The actor’s own Dread post describes a progression from having no prior knowledge to dumping emails and password hashes from vulnerable databases, running into tokenization while trying to obtain credit card data, and moving toward digital skimming. That account matches the order of the compromises documented in its infrastructure, although the actor’s five to nine years of web hosting administration contradicts the claim of having “started from scratch”.

CyberXero’s activity centers on three revenue streams:

  • Access brokering: Compromised WordPress administration panels, webshells and network footholds, consistent with an IAB model built around selling access
  • Payment and personal data theft: Database dumps, payment tables and a card skimmer endpoint tested against a production payment flow
  • Cryptocurrency payments: A self-hosted payment gateway that accepts Bitcoin, Litecoin, Ethereum, USDT (ERC-20 and TRC-20) and Monero

From an Open Directory to the CyberXero Alias

Attribution to the CyberXero alias rests on three independent lines of evidence. The open directory contained the operator’s Claude Code account data, the session logs showed the actor citing a Dread post written by the CyberXero account as proof of “owning” the forum, and external verification confirmed that the post was published by CyberXero and describes the same TTP progression observed in the infrastructure.

Five independent identity anchors were collected across Anthropic, Apple, Google, Twitter/X and OpenAI accounts, and a record on a Russian freelance marketplace that requires identity verification adds further weight. Two additional Telegram accounts receive notifications from the actor’s operational bot and are assessed as either a second alias or a trusted collaborator; their exact relationship to the primary operator could not be resolved with the available data.

The actor works in the Moscow time zone (UTC+3), supported by a provisioning timezone directive and 1,269 timestamped activity events peaking at 22:00 MSK, with virtually no activity between 02:00 and 08:00. Russian is the primary operational language, confirmed through script comments, a custom Claude Code instruction file and a mid-session language switch after a model refusal.

What Are the Targets of CyberXero?

SOCRadar identified more than 40 affected organizations across Ukraine, Poland, China, Pakistan and the Netherlands, excluding the 4,500+ WordPress sites hit through mass exploitation.

Geographic distribution of confirmed CyberXero victims
Geographic distribution of confirmed CyberXero victims

Ukraine carries the highest concentration, with 14 or more entities identified at different stages of the compromise chain, mostly in the energy and utilities sector, plus four organizations in telecommunications, e-commerce and CRM compromised through the opportunistic pipeline. Seven Ukrainian energy and utilities entities, including the national transmission system operator and the country’s largest private energy holding, were subjected to curated reconnaissance; one file alone lists 95 enumerated subdomains of the two largest, covering mail, VPN, network dispatch and data platforms.

Poland follows with four to five affected entities, two of them confirmed e-commerce compromises, and 284 scanned Polish domains with a heavy concentration of government institutions. The actor also compromised a Chinese corporate network cluster spanning at least three internal segments, attempted WAF bypass and credential spraying against two Pakistani national security government entities without confirmed success, and accessed a Dutch company through an exposed Magento configuration.

Sector distribution of confirmed CyberXero victims
Sector distribution of confirmed CyberXero victims

Energy and utilities account for 28% of confirmed victims, followed by e-commerce and payment (24%), government (20%), ISP and telecom (16%), corporate networks (8%) and CRM and SaaS (4%).

The scale of the confirmed impact stands out in two figures:

  • More than 628,000 records: Personal data of Ukrainian individuals in the actor’s possession, including home addresses and login IP addresses of residents of Kharkiv, a city on an active war front
  • 429 administration panels: WordPress panels confirmed accessible in a single automated run that scanned 4,708 targets and deployed shells on 32 hosts in 61 seconds

The opportunistic pipeline fits a conventional IAB model, while the focus on Ukrainian critical infrastructure has no obvious direct monetization path. The most plausible explanation is access brokering to a buyer with a specific interest in that infrastructure, although no evidence of a sale, buyer communication or tasking was found.

What Are the Techniques of CyberXero?

CyberXero runs an opportunistic pipeline and a targeted pipeline from the same workstations, sharing part of the same arsenal during the same period. Its infrastructure spans eight correlated nodes across three ASNs, linked through unique provisioning tokens, a shared internal hostname, command histories and staging logs.

Initial Access Through Mass WordPress Exploitation

The actor’s in-house tool, wp2shell, is a five-module Python package that abuses a desynchronization in the WordPress REST API batch endpoint (/wp-json/batch/v1), where the validation and execution handlers apply different authorization contexts. It adaptively selects union-based, error-based or time-based blind SQL injection, writes a new administrator directly to the database, logs in and deploys a WSO-family webshell registered as an active plugin.

Rogue administrators follow the pattern wp2_ plus eight random hexadecimal characters. A companion script, sc.py, works through the target queue and pushes each compromise to a rotating Telegram bot with the panel URL, credentials and shell status. Across the actor’s files, more than 56,000 URLs were scanned, 500 hosts were exploitable and 72 rogue administrator accounts were created during the primary observation period.

Targeted Intrusions Against Ukrainian Organizations

The targeted pipeline relies on deeper, manual chains. Against a Ukrainian classifieds and e-commerce portal, the actor combined blind SQL injection to recover an administrator email, a time-constrained password reset token extraction and a PHP deserialization chain built with phpggc, reaching remote code execution and dumping MySQL, MongoDB and Redis databases with 24,132 active sessions and 864 API tokens.

Against Kharkiv’s district heating service, no exploitation was needed: a hardcoded administrator credential let a multithreaded scraper pull 564,073 subscriber records and 213,340 login log entries. Other confirmed exfiltration hit a regional ISP (63,621 customer records and 75,811 cleartext access credentials) and a gas provider (27 employee accounts with active session tokens).

AI Augmentation and Safety Evasion

CyberXero uses AI in two independent layers. On its primary workstation, a swarm of 51 Claude Code agents covers offensive, reconnaissance and post-exploitation roles, from c2-operator and exploit-chainer to scada-attacker and data-exfiltrator. On the Team Server, PentAGI connects to Cobalt Strike through the Anthropic API, with the payload-generation role assigned the highest token limit and reasoning level.

A dedicated scope-guard agent wraps every task in a fictional “authorized pentest on own infrastructure” framing to lower the chance of model refusals, and the same ownership claim appears verbatim across four unrelated victims. When a model still refused, the actor reopened the task in a fresh session with the framing preloaded. The logs record repeated refusals nonetheless, including four declined requests to scan Dread, where the model asked for token-based ownership verification and rejected the actor’s attempt to paste the token back into the chat.

Rapid Weaponization of Vulnerabilities

CyberXero turns disclosed vulnerabilities into working exploits quickly. CVE-2026-4815 in Support Board 3.8.7 was published in June 2026, and a functional exploit with proxy rotation and automatic timing calibration appears in files dated July 2026, a cycle of less than 30 days. The arsenal also includes CVE-2026-63030 and CVE-2026-60137 in WordPress and CVE-2015-1397 in Magento.

Command and Control With Cracked Cobalt Strike

The actor runs Cobalt Strike 4.9.1 from a pirated build distributed through a Telegram channel, with a “Major Cobalt Strike” TLS certificate shared by about 184 other hosts that should not be treated as CyberXero infrastructure. The C2 receives HTTPS beacons on port 1002, serves HTTP staging on port 8044 and delivers PowerShell stagers on port 8033.

Lateral Movement in a Chinese Network Cluster

A separate campaign gained access to a Chinese corporate network by abusing a misconfigured Redis server through a rogue master-replication technique that planted the actor’s SSH key. From there, beacons enumerated the network, brute-forced a ZTE gateway with default credentials, moved laterally over WMI through a passwordless account, checked hosts for EternalBlue patch status and created the UpdSvc Windows service for persistence. Successful EternalBlue exploitation was not confirmed in the available logs.

Monetization and Competitor Suppression

A cryptocurrency payment gateway built on FastAPI and Celery, with local Litecoin and Monero full nodes, receives payments for sold access and data. The actor also built a card skimmer endpoint against a production payment flow, which had reached functional testing with about 50 masked card records. A script named abuse_monitor.py files abuse reports against competitors to URLhaus, VirusTotal and hosting providers under a fake fraud-analyst persona, a TTP undocumented in comparable IAB profiles.

CyberXero’s opportunistic and directed pipelines, orchestrated through the AI layer
CyberXero’s opportunistic and directed pipelines, orchestrated through the AI layer

How Can Organizations Defend Against CyberXero?

The controls below map to CyberXero’s documented TTPs, ordered along its intrusion chain from initial access to post-compromise activity.

Hunt for Rogue WordPress Administrators

Review WordPress user tables for any account whose username matches wp2_ followed by eight hexadecimal characters, and check the plugins directory for paths matching wp2shell_ with the same suffix. Either artifact indicates that wp2shell touched the host, even if the webshell is no longer active, so treat the site as compromised and rotate all administrator credentials.

Patch Exposed Web Applications Within Days

With a weaponization cycle under 30 days, patch windows measured in months leave WordPress, Magento, PrestaShop and Support Board deployments exposed. Prioritize CVE-2026-63030, CVE-2026-60137 and CVE-2026-4815, and confirm that configuration files such as Magento’s app/etc/env.php are not publicly reachable.

Remove Hardcoded and Default Credentials

Several CyberXero intrusions needed no exploit at all. Audit administration panels for shared or hardcoded credentials, replace default passwords on network devices such as gateways and routers, and eliminate local accounts with blank passwords that enable WMI-based lateral movement.

Lock Down Internet-Facing Redis

Do not expose Redis to the internet, require authentication, and disable or rename the SLAVEOF/REPLICAOF command where replication is not needed. Monitor authorized_keys files on Redis hosts for unexpected changes.

Block and Hunt Known C2 Infrastructure

Block the IP addresses listed in the IOC section and search historical logs for outbound connections to 42.193.227.214 on ports 1002, 8044 or 8033. Hunt for the UpdSvc service and PowerShell download cradles on Windows endpoints.

Treat AI Agent Session Logs as Sensitive Data

The CyberXero investigation shows that AI agent session logs record every planned task, target and workaround, making them the equivalent of a keylogger of the operator’s activity. Organizations running AI agents with terminal, network or data access should apply the same retention, encryption and access controls to these logs as to credential stores, and audit them for repeated attempts at the same refused task.

Integrate Dark Web and Threat Intelligence Monitoring

CyberXero advertises on Dread and HackForums, and IAB listings often surface before access is used. Monitoring Dark Web forums and Telegram channels for references to your domains, brands and infrastructure gives an early signal that access may be for sale. SOCRadar Threat Intelligence’s Advanced Dark Web Monitoring (ADWM) module provides this coverage.

What Are the MITRE ATT&CK TTPs of CyberXero?

TacticTechniqueID
ReconnaissanceActive Scanning: Vulnerability ScanningT1595.002
ReconnaissanceGather Victim Network Information: DNST1590.002
ReconnaissanceGather Victim Org InformationT1591
Resource DevelopmentAcquire Infrastructure: Virtual Private ServerT1583.003
Resource DevelopmentObtain Capabilities: ToolT1588.002
Resource DevelopmentObtain Capabilities: ExploitsT1588.005
Initial AccessExploit Public-Facing ApplicationT1190
Initial AccessValid Accounts: Local AccountsT1078.003
Initial AccessBrute Force: Password SprayingT1110.003
ExecutionCommand and Scripting Interpreter: PowerShellT1059.001
ExecutionCommand and Scripting Interpreter: Unix ShellT1059.004
ExecutionWindows Management InstrumentationT1047
ExecutionSoftware Deployment ToolsT1072
PersistenceServer Software Component: Web ShellT1505.003
PersistenceCreate or Modify System Process: Windows ServiceT1543.003
PersistenceAccount Manipulation: SSH Authorized KeysT1098.004
PersistenceCreate Account: Local AccountT1136.001
StealthObfuscated Files or InformationT1027
StealthMasquerading: Match Legitimate Resource Name or LocationT1036.005
Credential AccessOS Credential Dumping: LSASS MemoryT1003.001
Credential AccessOS Credential Dumping: Security Account ManagerT1003.002
Credential AccessInput Capture: Web Portal CaptureT1056.003
Credential AccessSteal Web Session CookieT1539
DiscoveryNetwork Service DiscoveryT1046
DiscoverySystem Information DiscoveryT1082
DiscoverySystem Network Configuration DiscoveryT1016
Lateral MovementUse Alternate Authentication Material: Pass the HashT1550.002
Lateral MovementRemote Services: SMB/Windows Admin SharesT1021.002
CollectionData from Local SystemT1005
CollectionData from Information RepositoriesT1213
CollectionData from Cloud StorageT1530
Command and ControlApplication Layer Protocol: Web ProtocolsT1071.001
Command and ControlWeb Service: Bidirectional CommunicationT1102.002
Command and ControlProxy: Multi-hop ProxyT1090.003
ExfiltrationExfiltration Over C2 ChannelT1041
ExfiltrationExfiltration Over Web Service: Exfiltration to Cloud StorageT1567.002
ImpactData Manipulation: Stored Data ManipulationT1565.001

What Are the IOCs of CyberXero?

TypeIndicatorDescription
IP46.21.250.135Primary workstation, open directory seed node
IP45.88.106.2Provisioning server (port 1500)
IP212.193.31.189Cryptocurrency payment gateway
IP42.193.227.214Cobalt Strike Team Server and PentAGI
IP91.208.184.148Secondary workstation
IP45.88.106.78Mass scanning server
IP49.234.12.182Staging and payload delivery server (Chinese cluster)
IP42.193.100.94Redis exploitation node and PentAGI agents
IP:Port42.193.227.214:1002Cobalt Strike HTTPS beacon
IP:Port42.193.227.214:8044Cobalt Strike HTTP staging
IP:Port42.193.227.214:8033Cobalt Strike PowerShell stager delivery
SHA25692a789444708fa1cb4cc5a89e0aa6cc7a279b62a7b8a4d2857255a18197d0090stager32.bin, Cobalt Strike x86 stager (4.9.1 pirated build)
MD53870a559b5b6ea6076f5f90029c180b7stager32.bin, Cobalt Strike x86 stager
SHA256a5ae0aab352871bc0b038b3aa43b03eb223425628c884af7b9fc592cd34eb86cObfuscated WSO webshell
SHA25671e21094c1ac1cf0275c91ed377965e248bca1c12711f6dd20e2682681fc11921.bin, Chinese cluster staging payload
SHA25659c535f47ab4d35f6d9fc8b8aec4a72438ea0b89e5a4dcea74b05d2d32cfa483config.bin, Chinese cluster staging payload

Behavioral indicators, weaponized CVEs and YARA rules are covered in the full threat research report. For additional IOCs and detection rules, you can visit the CyberXero threat actor profile on the SOCRadar platform.

Threat Actor

SOCRadar Dark Web Team

Intelligence in your inbox every Thursday.

Join 30,000+ security professionals reading SOCRadar research.

Subscribe