Ransomware family
Hunters International Ransomware Decryption & Data Recovery
- Inactive
- High
- No public decryptor
Hunters International was a RaaS operation built on Hive's source code from October 2023, marked early on by the .locked extension and a Contact Us.txt note. From v6 it encrypted silently without renaming files or dropping notes. The crew shut down in July 2025 and rebranded as the data-theft-only brand World Leaks.
- First seen
- 2023-10
- File extensions
- .locked
- Ransom notes
- Contact Us.txt
- Affected platforms
- Windows / Linux / VMware ESXi
Family profile
- File extensions
- .locked
- Ransom notes
- Contact Us.txt
- Contact patterns
- Per-victim Tor (.onion) negotiation portal (linked from early ransom notes)
- Hunters International DLS leak site on Tor (.onion), where victim names and sample data were staged
- A clearnet mirror of the leak site also existed while the brand was running
- From v6 no note is dropped; operators contact victims directly by phone, email or social media using OSINT
- Aliases / versions
- Water Ouroboros(Trend Micro 命名)、Hunters International v1–v6、Hive 代码承接团伙、World Leaks(2025 年更名后的纯窃密品牌)
- First seen
- 2023-10
- Status
- Inactive
- Operational status
- Rebranded
- Threat level
- High
- Affected platforms
- Windows
- Linux
- VMware ESXi
- Tags
- Defunct
- Rebranded
- Ransomware-as-a-Service
- Double extortion
- Targets virtualization
- Exploits vulnerabilities
- Rebranded as
- World Leaks
There is no trustworthy public decryptor. No More Ransom lists KISA's Hive (v1–v4) Decryptor, but Hunters International inherited the later Rust-based Hive branch, so that tool does not apply, and no public decryptor for the Hunters International encryptor itself has ever been released.
The "free decryptors" announcement of 3 July 2025 needs context. When the crew announced it was closing the project, it said it would supply decryption software and recovery guidance free of charge to affected organisations through its own site. That is not a tool validated by a vendor or law enforcement: it requires victims to contact the criminals and retrieve a binary from their onion site, with no way to verify that it works, that it matches the build in question, or that it carries no secondary payload. No public report has confirmed a victim organisation actually decrypting data this way, and there is no way to confirm the criminals' own infrastructure is still reachable now that the brand is closed. We neither recommend nor pursue that route on a client's behalf.
Any recovery conclusion for Hunters International therefore has to come from analysing real samples: identify the encryptor branch (Windows, Linux or ESXi, pre- or post-v6) and which regions of each file were overwritten, then decide between structural repair, backup rollback and low-level carving.
Latest activity
Public tracking shows World Leaks' last victim posting on 21 July 2026, with its onion site persistently unreachable since. The original Hunters International leak site and onion addresses also remain offline.
SourcesWorld Leaks, the Hunters International successor, leaked Tata Electronics data - allegedly Apple-related component schematics, PCB designs and SDK files. No Hunters International encryption activity has resumed.
Sources
Overview
Hunters International first appeared in the wild in mid-October 2023: Group-IB dates the first disclosure on its leak site to 13 October 2023, and public trackers place the start of the brand in the same part of that month. The crew claimed it had bought Hive's infrastructure and ransom binary and had "fixed the bugs in Hive that prevented file decryption in some cases". Group-IB assesses with moderate confidence that the group is possibly a Hive rebrand, citing heavy code overlap and underground forum posts that refer to it in Russian simply as "хайв" (Hive). Trend Micro tracks the actor as Water Ouroboros. The encryptor is written in Rust and ships for x86, x64 and ARM.
It ran as classic ransomware-as-a-service: the core team supplied the encryptor, the negotiation portal, the leak site and an exfiltration utility known as Storage Software, while affiliates handled intrusion and deployment. Microsoft-tracked Storm-0501 was one such affiliate, having deployed Hive, BlackCat, Hunters International, LockBit and Embargo in turn.
Over roughly two years the brand publicly claimed close to 300 attacks. Public trackers record 307 victim organisations across 53 countries, about 150 of them in the United States, 16 in the United Kingdom and 15 in Canada. Trend Micro's breakdown puts construction, IT, manufacturing and healthcare at the top, while public trackers rank professional services, manufacturing and technology first; either way the victims skew towards small and mid-sized organisations. Higher-profile cases include Fred Hutch Cancer Center (hundreds of thousands of patient records, with extortion emails sent directly to patients), Integris Health, US Navy shipbuilder Austal USA, Japanese optics maker Hoya, AutoCanada and Tata Technologies.
The lifecycle is closed. On 17 November 2024 the group first signalled it was winding down, citing falling profitability and rising government scrutiny. On 1 January 2025 it launched World Leaks, a pure data-extortion brand with no encryption at all. Group-IB documented the transition in April 2025, and on 3 July 2025 the operators formally announced the closure of the Hunters International project. World Leaks kept posting victims into mid-2026 - public trackers list roughly 170 organisations under that brand, with its last public posting on 21 July 2026. For enterprises today the family matters in two ways: legacy .locked data that still needs a recovery assessment, and the fact that its intrusion playbook - poisoned installers of sysadmin tools, and going straight for ESXi - is still in wide use by other crews.
How to identify it
The trap with this family is that it changed its entire signature in August 2024.
v1-v5 (October 2023 - August 2024)
- Extension: .locked appended to the full original filename (vendor reports usually render it .LOCKED), so report.xlsx becomes report.xlsx.locked. The base name is neither rewritten nor randomised.
- Ransom note: Contact Us.txt dropped in affected directories, in English, claiming both encryption and data theft, and supplying a victim ID plus a Tor negotiation portal.
- Leak site: the organisation's name appears on the Hunters International DLS with a sample data pack, confirming exfiltration.
v6 onwards (from 14 August 2024) The encryptor stopped appending extensions and stopped dropping ransom notes entirely - a deliberate stealth choice intended to keep ordinary staff unaware of the breach and to keep the communication channel in the attackers' hands. Identification then rests on:
- Files that still open by type but contain garbage, with sizes essentially unchanged and icons and extensions intact.
- A region of the file header left intact: per Group-IB the encryptor writes a fixed marker at offsets 0x45-0x58 to flag files it has already processed, and ciphertext starts at offset 0x59, so magic numbers and Office or PDF headers generally survive - "the file type is recognised but the file will not open" is the classic symptom.
- No note at all; instead the operators phone, email or message executives directly using OSINT-gathered contact details.
- Host-side traces: shadow copies wiped, database and backup agent services stopped, and a long list of business processes force-killed.
ESXi cases. Virtual machines are shut down in bulk via vim-cmd, and vmx, vmdk, vmss and nvram files under /vmfs/volumes are rewritten. Encrypted files carry a marker at offset 52 consisting of four random bytes plus a 17-byte constant, with 640 bytes of RSA-encrypted metadata appended at the end, and the free space of the datastore is filled with random data.
Bottom line. The .locked extension alone is not enough to attribute the family - several unrelated strains have used it - so the note structure, file trailer markers and sample characteristics must be assessed together. Equally, the absence of a changed extension and of any note does not rule this family out.
Infection vectors
The defining access pattern for Hunters International and its affiliates was poisoning the IT administrators themselves rather than ordinary staff - going straight for the people holding privileged credentials:
- Search ads and SEO poisoning delivering trojanised sysadmin tools. In August 2024 the group distributed a digitally signed fake installer, ipscan-3.9.1-setup.exe, from sites impersonating Angry IP Scanner (and earlier Advanced IP Scanner), dropping the SharpRhino C# backdoor used for privilege escalation and PowerShell execution. In 2025 affiliates used malvertising to push a trojanised installer for RVTools, a VMware management utility, steering administrators to a fake download page that delivered the SMOKEDHAM PowerShell backdoor.
- Abuse of legitimate monitoring software. Once inside, Kickidler employee-monitoring software was deployed to capture an administrator's keystrokes and browsing, harvesting cloud backup and hypervisor credentials without tripping detection.
- Exploitation of internet-facing applications. Trend Micro observed exploitation of known Oracle WebLogic flaws (CVE-2019-2725, CVE-2017-10271, CVE-2019-2729), while affiliate Storm-0501 leaned on Zoho ManageEngine (CVE-2022-47966), Citrix NetScaler (CVE-2023-4966) and ColdFusion 2016, alongside heavy use of credentials bought from access brokers.
- Lateral movement and deployment. Kitty for reverse SSH tunnelling, WinSCP for exfiltration and payload delivery, Total Commander to browse and archive network shares, and VMware PowerCLI to drive ESXi hosts and run the encryptor across them; AdFind for domain reconnaissance, ProcessHacker and PCHunter to disable endpoint protection.
- Exfiltration. A bespoke Storage Software utility handled bulk data theft; it was later upgraded into the dedicated exfiltration tool used by World Leaks.
The practical lesson is blunt: the first domino is usually an engineer downloading a plausible-looking tool from the top of a search results page. Controlling where admin tooling comes from, allow-listing applications on administrator workstations and restricting network reachability of ESXi management interfaces pays far better than adding another endpoint agent afterwards.
Encryption behavior
Windows / generic encryptor. Written in Rust, file content is encrypted with AES using a randomly generated 128-bit key per file, driving the CPU's AES instruction extensions (AESKEYGENASSIST, AESIMC, AESENC, AESENCLAST) for speed. Each symmetric key is wrapped with an embedded RSA public key and stored alongside the file; the private key never touches the victim network, so there is no mathematical way back without it.
File layout. A region of the header is not overwritten: per Group-IB the encryptor checks and writes a fixed marker at offsets 0x45-0x58 to tell whether a file has already been processed, and ciphertext begins at offset 0x59. That detail matters forensically: magic numbers survive, so file type detection still works even though the content is unusable, and the surviving header retains some value for structural repair.
ESXi encryptor. Virtual machines are shut down in bulk via vim-cmd, then /vmfs/volumes is walked and vmx, vmdk, vmss and nvram files encrypted. Synacktiv's analysis of one incident shows this variant uses AES-256 CTR and encrypts only part of each file - roughly the first 10% of the data after the marker (capped at about 100 MiB) plus a further limited amount - with 640 bytes of RSA-encrypted metadata appended and a marker at offset 52 made of four random bytes plus a 17-byte constant. It then writes a temporary file that fills the datastore's remaining free space with random data, a deliberate anti-forensics step that severely limits what can later be carved from unallocated space. That description comes from a single vendor's single case, so whether it holds for every ESXi build has to be measured per incident.
Recovery inhibition. The encryptor carries a default list of dozens of processes and services to terminate (Group-IB's published list runs to roughly fifty entries: database engines, backup agents, Office, browsers and more), tunable via -k/--kill and -s/--skip. Several commands clear shadow copies and backup catalogs, including vssadmin.exe delete shadows /all /quiet, and the build can wipe free disk space. Other switches - -w/--wait to delay execution, -l/--low-key for a quiet run that skips process termination and disk partition mounting, -R/--no-remote to disable remote host enumeration, -E/--no-erase to skip wiping, -a/--admin to impersonate a domain administrator - mean the amount of collateral damage varies widely between incidents and must be measured case by case rather than assumed.
Double extortion. The operators openly prioritised data theft over encryption: files were exfiltrated with Storage Software before the encryptor ran, non-payment led to staged publication on the leak site, and in some cases extortion emails were sent directly to the victim's patients or customers. That preference eventually became World Leaks, which dropped encryption altogether.
Assess before you act
Recoverability assessment
There is no shortcut with Hunters International, and every recovery conclusion has to be reached sample by sample. We do not pay ransoms and do not negotiate; our work is technical recovery, breach impact assessment and attribution.
1) Public decryptor: none exists. The Hive (v1-v4) Decryptor on No More Ransom targets the early Hive branch and does not transfer to the Rust codebase Hunters International inherited. The "free decryptors" the crew offered in July 2025 require contacting the criminals, cannot be validated, and no public report confirms a victim recovering data through them, so it is not a dependable path. Real work therefore starts at option two.
2) Repair space created by the file layout and encrypted range (depends on the encryption pattern). The Windows encryptor leaves a header region intact and overwrites from offset 0x59, while the ESXi variant, per Synacktiv, encrypts only a leading portion of each file. The extent and depth of overwriting must therefore be measured for each critical file type. Where large files - database MDF/LDF, DBF, ibd, virtual disks such as vmdk and vhdx, mail stores - have only their leading regions damaged while the data pages survive, page-level extraction and logical rebuilds are worth attempting. Where a file was overwritten end to end, this path simply does not exist. Sample and measure first; never conclude from the family name.
3) Backups, snapshots and shadow copies. Shadow copies and backup catalogs are usually gone, but each source still deserves a check: offline and offsite backups, storage array or NAS volume snapshots, hypervisor snapshots, untouched copies on the backup server, and cloud version history. This is typically the highest-yield path for this family. Note that in ESXi cases the operators frequently logged into the backup platform with stolen credentials and deleted jobs first, so the backup system's own audit log must be reviewed.
4) Unencrypted copies and log replay. File server recycle bins, endpoint caches, reporting and BI staging databases, ERP archive exports, database transaction logs and application audit logs can all support reconstruction or point-in-time replay of critical records.
5) Low-level carving. Some builds write a new encrypted file and delete the original, leaving source data in unallocated clusters. Be realistic, though: this encryptor can wipe free space, and the ESXi variant fills the datastore's remaining space with random data, so the path is often actively destroyed before responders arrive - success rates are noticeably lower than with other families. It still requires stopping all writes immediately.
6) Breach impact assessment, which is mandatory here. Because the crew prioritised theft and has a record of contacting victims' patients and customers directly, even a clean file-level recovery leaves work to do: establish the scope and timeline of exfiltration, drive internal notification and contractual or regulatory obligations from it, and rotate the affected accounts, keys and credentials.
What we commit to is a verifiable assessment and a clearly bounded recovery scope. Any claim of complete data restoration is unfounded: the achievable ratio depends on the sample, the damage done on site and the state of the backups, and can only be expressed as a measured range.
Our response plan
Hit by Hunters International ransomware? What to do
Containment and forensic preservation
Isolate affected hosts, ESXi servers and storage paths, and disable suspicious administrator accounts and VPN sessions - but do not power off or reboot. SharpRhino and SMOKEDHAM implants, along with memory-resident processes and connections, are lost the moment you do, and attribution and partial recovery go with them. Image or snapshot the domain controller, backup server and hypervisor management hosts first. Export logs from the firewall, VPN gateway, Active Directory, ESXi and the backup platform - login and job-deletion records on the backup platform matter especially. Preserve three to five encrypted samples plus the original Contact Us.txt where one exists; for post-v6 incidents preserve the attackers' calls or emails instead.
Family identification and encryptor branch analysis
Confirm Hunters International from file trailer markers, the encryption start offset, note structure and sample characteristics, and separate the Windows, Linux and ESXi encryptors as well as pre- and post-v6 behaviour. Three things matter most: how far the overwrite from offset 0x59 actually extends, whether switches such as -l/--low-key or -E/--no-erase were used (which determines how much process termination, shadow copy deletion and free-space wiping actually happened), and whether free-space wiping ran. Because .locked is shared with unrelated strains and post-v6 builds leave neither extension nor note, this step cannot be skipped.
Parallel recoverability and breach impact assessment
Run two tracks. On recovery: sample-measure the overwritten range across critical databases, virtual disks and documents, inventory offline backups, storage and hypervisor snapshots and unencrypted copies, and run small-scale repair trials. On exposure: reconstruct what was stolen and when, from exfiltration tooling artefacts, egress traffic and leak site postings. Deliver a written assessment stating which systems go the structural repair route, which rely on backup rollback and which are unrecoverable, with expected recovery ranges, timelines and a business priority order - then execute only after sign-off.
Recovery execution and business verification
All work happens on images or copies, with originals kept read-only. Restore in business priority order: domain controllers and identity first, then core databases such as ERP, MES and finance, then file and mail systems. In ESXi cases, repair virtual disk structures and mount them to extract inner data rather than creating new VMs on the original LUN or re-initialising the datastore. After each batch, run integrity checks and business-side verification - reconciliation, report comparison, application start-up tests - and record everything in a traceable recovery manifest.
Attribution, hardening and compliance closure
Reconstruct the full kill chain: whether an administrator workstation installed a trojanised tool from a search ad (an Angry IP Scanner or RVTools lookalike), whether an internet-facing application was unpatched, where admin credentials were captured by keylogging, and when SSH was enabled on ESXi. Remove backdoors, monitoring software such as Kickidler, rogue accounts and scheduled tasks. Reset credentials domain-wide and enforce MFA. Apply application allow-listing on admin endpoints and govern where sysadmin tooling is downloaded from. Segment the ESXi management network, disable SSH by default and enable lockdown mode. Rebuild backups to a 3-2-1 design with immutable copies. Finally, complete notification and credential rotation driven by the exposure assessment, and close with an incident report and handover checklist.
Risk warning
What not to do
- Do not reboot or power off affected hosts and ESXi servers - losing memory-resident key material, implants and connections destroys forensic evidence and any partial recovery chance at the same time.
- Do not follow a note or an attacker phone call to their onion site to collect a so-called free decryptor; the source cannot be validated, it may carry a secondary payload, and no public report confirms the route ever returned anyone's data.
- Do not run downloaded "Hunters International decryptors" against original disks. No trustworthy public decryptor exists for this family, such tools are usually re-extortion or destructive, and any trial belongs on isolated copies.
- Do not delete encrypted samples, Contact Us.txt or records of the attackers' calls and emails, and do not rush into antivirus clean-up - they are the only basis for build identification and for scoping the data exposure.
- Do not format, reinstall or rebuild RAID sets and storage pools, and never re-initialise an ESXi datastore. The encryptor already wipes free space, so any additional write finishes off what remained of low-level carving.
- Do not reconnect backup tapes, external drives or the backup server to a network that has not been cleaned - the operators often already hold backup platform credentials and target jobs and copies first.
Emergency response
Data already encrypted? Stop and let an engineer look first
We do not pay ransoms and we do not negotiate with attackers. Engineers run a free assessment first, then propose a recovery plan and a firm quote.
Related scenarios
ESXi / Hyper-V Virtualization Encrypted by Ransomware
Hypervisor-level encryption causes the widest blast radius of any ransomware event: dozens of production VMs go dark within an hour or two. This page covers what Linux ESXi encryptors actually do — shut down guests, encrypt vmdk, delete snapshots — the recovery value of flat disk files, and how Hyper-V and Proxmox cases differ.
Domain Controller Compromise and Estate-Wide Encryption
A compromised domain controller hands the attacker a legitimate administrator identity, allowing an encryptor to be pushed to every host at once through Group Policy or remote execution. This page covers how such incidents present, the correct order for Active Directory recovery, and how to decide between cleanup and full rebuild.
Backups Deleted or Destroyed
Modern ransomware follows a fixed sequence: destroy the backups, then encrypt the data — deleting shadow copies, encrypting repositories, disabling jobs, and exploiting backup software flaws to steal credentials. This page covers what can still be inventoried once backups fail, why replication propagates encrypted files off-site, and what offline and immutable copies are really worth.
Related industries
Manufacturing Ransomware Response and Recovery
Ransomware in manufacturing hits information systems and production cadence at the same time: with ERP down there are no orders, with MES down there is no schedule, and an encrypted drawing library takes the process documentation for an entire product line with it. This page covers the asset profile, recovery priorities and targeted defences.
Healthcare Ransomware Response and Recovery
When a hospital is hit, registration, consultation, orders, billing, laboratory and imaging fail at the same moment and care falls back to paper. This page covers the healthcare threat picture, a recovery priority built around clinical continuity, and the handling of patient data and compliance obligations.
Construction and Real Estate Ransomware Response
In construction and real estate the core assets are drawings, models and project documentation — typically scattered across project-site NAS units, shared drives and personal computers with no unified backup. This page covers the sector's threat profile, recovery of drawings and BIM models, and defences for dispersed, multi-project environments.
Similar families
- No public decryptor
World Leaks
World Leaks is the extortion-only brand Hunters International adopted in January 2025: no encryptor, no renamed files, just data theft backed by a Tor leak site. No new victims have been posted since late July 2026 and the leak site has been unreachable, so the operation currently looks dormant.
- No public decryptor
Qilin
Qilin (formerly Agenda) is a Rust-rewritten cross-platform RaaS operation focused on VMware ESXi and Linux estates. It has ranked as the world's most active ransomware group for several consecutive quarters since 2025, with confirmed victims among electronics manufacturers in Taiwan and Hong Kong.
- No public decryptor
BlackCat
BlackCat (ALPHV) was the first major Rust-based ransomware-as-a-service operation, active from November 2021. It seized accounts through help-desk social engineering, encrypted ESXi and Windows estates under double extortion, and shut down in a March 2024 exit scam after keeping the Change Healthcare ransom — its key infrastructure no longer exists.
FAQ
Hunters International Frequently asked questions
Can .locked files be decrypted?
No free public decryptor exists for Hunters International, and the Hive (v1-v4) decryptor on No More Ransom does not apply to the Rust branch this crew inherited. One caveat matters: .locked is a generic extension shared by several unrelated families, so seeing it does not confirm Hunters International - and some of those other strains do have viable decryption paths. The first step is therefore not to hunt for a tool but to identify the family and build from three to five encrypted files plus the note, and to assess structural repair and backup rollback from there.
The gang offered free decryptors in 2025 - can we still get one?
When Hunters International announced the closure of the project on 3 July 2025 it did say it would provide decryption software free of charge to affected organisations through its own site. That is fundamentally different from a tool published by law enforcement or a security vendor: it means contacting the criminals and retrieving an unvetted binary from an onion site, with no way to verify that it works, covers your build, or is safe - and no public report confirms any organisation actually decrypting data that way. We do not recommend that route and do not pursue it for clients. Sample analysis and a backup inventory come first, so recovery rests on a verifiable technical path.
Files will not open but the extension is unchanged and there is no note - is this the same family?
Possibly. From version 6 in August 2024 Hunters International deliberately stopped appending extensions and dropping notes, specifically to delay discovery. The signature symptom is files whose icon and type detection look normal and whose size is essentially unchanged, yet whose content is garbage - because a header region is preserved (with a fixed marker written at 0x45-0x58) and ciphertext begins at offset 0x59. Check whether shadow copies were wiped, whether database and backup services were stopped abnormally, and whether executives recently received unsolicited calls or emails. Attribution here has to come from sample analysis, not from the extension.
Our ESXi virtual machines were encrypted - is anything recoverable?
There is a chance, but start from an unfavourable fact: this family's ESXi encryptor fills the datastore's remaining free space with random data after encrypting, which sharply reduces what low-level carving can achieve. Priorities are therefore: check whether datastore, storage array or NAS volume snapshots survive and whether the backup platform holds VM-level backups - reviewing its login and job-deletion logs too, since the operators usually get there first. Then attempt structural repair of the vmdk files: published analysis indicates this ESXi variant encrypts only a leading portion of each file, so the later part of a large virtual disk may still be plaintext - measure the overwritten range and judge from there whether the disk can be mounted and inner files extracted. Critically, do not create new VMs on the original LUN and do not re-initialise the datastore.
The group has disbanded - do we still need to handle the data exposure?
Yes. Hunters International prioritised theft in the first place, and copies of stolen data do not disappear when a brand closes: the crew rebranded in January 2025 as the data-extortion-only operation World Leaks, which remained active into mid-2026, and historical breach data has repeatedly been resold or recycled by other groups for fresh extortion attempts. Regardless of how file recovery goes, scope the exfiltration and its timeline quickly, drive internal notification and compliance obligations from that, and rotate and notify around the affected accounts, keys, API credentials and customer records.
Sources
- Ransomware Spotlight: Water Ouroboros (Hunters International) — Trend Micro
- Hunters International Ransomware Group — Group-IB
- Case study: how Hunters International and its affiliates target your hypervisors — Synacktiv
- Hunters International ransomware shuts down, releases free decryptors — BleepingComputer
- Hunters International group profile — ransomware.live
- New Hunters International ransomware possible rebrand of Hive — BleepingComputer
- Hunters International ransomware gang targets IT workers with new SharpRhino malware — BleepingComputer
- Storm-0501: Ransomware attacks expanding to hybrid cloud environments — Microsoft
External links are provided for reference only. The content is published by third parties and does not represent our position.
Updated