CYOPS · Cyber Operations

k0t: an edge-device access operation with 70,000 credentials and a heavy APAC footprint

Group-IB and NyxLab describe an operator that scans for exposed FortiGate, FortiWeb and GlobalProtect appliances, decrypts what it finds and leaves remote-management tooling behind. Asia-Pacific accounts for 37.6% of identified victim domains.

A single operator working the network edge has accumulated more than 70,000 exposed credentials from firewalls, web application firewalls and VPN gateways, according to a joint investigation by Group-IB and NyxLab published on 21 October 2026 [1]. The operator is tracked as k0t. Group-IB found that the credential set includes network-device logins and LDAP dumps, and that it does not match data held by other tracked actors [1]. Asia-Pacific accounts for 37.6% of the unique victim domains identified [1].

The finding matters because it lands four months after the FortiBleed leak, which prompted national advisories including one from Singapore’s Cyber Security Agency [9][10]. k0t is a separate dataset with a different toolset. Teams that closed out FortiBleed should not assume they have closed out this.

What the investigation found

Group-IB describes a large-scale campaign that combines vulnerability exploitation with brute force against internet-exposed devices [1]. Targets were selected by product, not by organisation, using internet-scanning platforms and underground IP-list services [1][3]. Many of the FortiGate compromises date from February to June 2026 [1]. Group-IB says the targeting was opportunistic and driven by exposure rather than sector [1].

Group-IB states that the unique-domain count is a lower bound on the number of affected organisations, because victim identifiers ranged from domain names to internal addresses [1][2]. It adds that the APAC share “should not be interpreted as proof of deliberate regional targeting”, and that Asia-Pacific and Latin America together hold the largest share of victim domains [1]. No sector breakdown is published [1].

The report names k0t as the likely main operator and treats CryptoFeel as a potential alias. A second handle is described as a likely collaborator. Two tool developers named in the code are not directly linked to the operations and may be freelancers selling to several groups [1].

How access was gained and kept

The intrusion chain uses known flaws. A custom scanner first tries CVE-2025-64446 against FortiWeb and falls back to brute force against FortiGate [1][3]. Fortinet disclosed CVE-2025-64446 on 14 November 2025 as a relative path traversal that lets an unauthenticated attacker run administrative commands, and confirmed exploitation in the wild [4]. Sources differ on its severity: Fortinet’s advisory scores it 9.8, while Tenable’s analysis reports 9.1 [4][13].

Once logged in, the tooling downloads the device configuration and decrypts stored secrets with the default FortiOS key associated with CVE-2019-6693 [1][3]. Fortinet’s advisory for that issue explains that anyone holding a configuration or an unprotected backup can decrypt the protected credential fields unless private-data-encryption is enabled with a user-defined key [5]. That setting is off by default [5]. The practical effect is that LDAP and Active Directory bind accounts stored on the firewall become the route into the domain [1][3].

For GlobalProtect, the operators used CVE-2026-0257, an authentication bypass based on forged authentication-override cookies, then opened VPN sessions [1][3]. Palo Alto Networks published the advisory on 13 May 2026 and now rates it high severity with the highest suggested urgency [6]. Security Affairs, reporting Rapid7’s findings, notes it was initially rated medium and that exploitation began within days of disclosure [7].

Inside the network, Group-IB observed common post-exploitation tooling for credential theft and privilege escalation, including the noPac chain for domain administrator access [1]. Persistence came through OpenConnect VPN sessions, MeshCentral agents, AdaptixC2 and commercial remote-management products, plus new local administrator accounts on the appliances [1]. One MeshCentral server showed 84 compromised devices across 28 active groups, with records of 100 removed groups [1][2]. In the worst cases, data was moved to cloud storage with Rclone [1].

The ransomware question

Many k0t victims later appeared on the Qilin and The Gentlemen leak sites [1]. Group-IB assesses that the operators “are likely to be acting as Initial Access Brokers”, while stating that the relationship with the ransomware groups “is still unconfirmed” [1]. Barracuda describes The Gentlemen as a ransomware-as-a-service operation heavily associated with compromised FortiGate appliances and stolen or purchased VPN credentials [12]. Leak-site listings are claims by criminals. They are not confirmation of how any individual victim was breached.

The link to FortiBleed is weaker than headlines suggest. Group-IB found negligible overlap between the two credential sets and assesses with moderate confidence that k0t is associated with broader FortiBleed-related activity, while stating the evidence does not show the same operators controlled both [1][2]. Attribution of FortiBleed itself also differs by source. SOCRadar attributes it to the Lynx/INC group [10]. The FBI and US Secret Service name INC/Lynx and Payload as ransomware groups receiving access [8]. Help Net Security describes the leaking group as Russian-speaking [11].

Counts differ as well. Help Net Security reports Hudson Rock’s figure of 73,932 firewall URLs [11]. SOCRadar and the FBI cite 86,644 compromised devices across 194 countries [8][10][14]. Singapore’s Cyber Security Agency refers to credentials for over 70,000 FortiGate devices [9]. These figures describe FortiBleed, not k0t, and should not be added together.

What to do

  • Confirm FortiWeb is on a fixed release for CVE-2025-64446; if not, disable HTTP and HTTPS administration on internet-facing interfaces until it is [4].
  • Confirm PAN-OS is fixed for CVE-2026-0257, or apply the vendor workaround of a dedicated certificate for authentication-override cookies or disabling the feature [6].
  • Enable private-data-encryption with an operator-defined key on FortiGate, and password-protect every configuration backup [1][5].
  • Treat any configuration that has left a device as a credential disclosure. Rotate LDAP and Active Directory bind accounts stored on it and use low-privilege bind accounts [1].
  • Remove management interfaces from the internet, enforce MFA on administrative and VPN access, and investigate repeated lockouts rather than clearing them [1][9].
  • Hunt for unapproved MeshCentral agents, AdaptixC2, remote-management tools and unexpected Rclone use, and review new appliance administrator accounts and configuration changes [1][8].
  • Preserve edge-device, VPN, identity and endpoint logs before any factory reset, as Singapore’s advisory recommends [1][9].

Indicators for this campaign are published in the Group-IB report and should be taken from there [1].

The cyber operations view

k0t is a reminder that edge access has become an industrial input to ransomware. The operator did not need a new vulnerability. It needed a product-filtered target list, a scanner that waits out FortiGate’s default lockout timer, and configuration files protected by a default key [1][2]. FortiBleed showed the same economics at a larger scale [8][10].

For APAC defenders, the 37.6% figure reflects where vulnerable appliances sit, not a regional campaign [1]. That makes it a measure of exposure across the region. Teams that already ran FortiBleed checks should extend them: check FortiWeb and GlobalProtect, decrypt-proof the configuration, and assume that any directory account stored on an edge device has been read. The fix is identity hygiene behind the firewall as much as firmware on it.

Sources

Every R3KONX article cites its primary material. 14 sources, in order of first citation. Links open the original publication.

  1. Stray Cat, Open Doors: How k0t Wanders Through the Network Edge Group-IB · 2026-10-21
  2. Suspected Access Broker k0t Hits FortiGate and GlobalProtect Devices, Exposing 70,000 Credentials Daily CyberSecurity (SecurityOnline) · 2026-10-23
  3. K0T Hackers Target FortiGate, FortiWeb and Palo Alto GlobalProtect in Credential Theft Campaign GBHackers · 2026-10-22
  4. FG-IR-25-910: FortiWeb relative path traversal (CVE-2025-64446) Fortinet PSIRT · 2025-11-14
  5. FG-IR-19-007: Hardcoded symmetric key in fips.c (CVE-2019-6693) Fortinet PSIRT · 2020-06-30
  6. CVE-2026-0257 PAN-OS: GlobalProtect authentication bypass Palo Alto Networks · 2026-05-13
  7. CVE-2026-0257: Rapid7 Caught Attackers Abusing Forged VPN Cookies Against Multiple Customers Security Affairs · 2026-05-31
  8. FortiBleed Operations Continue Targeting Exposed Systems Leading to Reports of Lockouts (JCSA-20261006-01) FBI and US Secret Service · 2026-10-06
  9. Advisory on Credential Compromise of FortiGate Devices ("FortiBleed") AD-2026-007 Cyber Security Agency of Singapore · 2026-06-22
  10. FortiBleed: The Campaign That Cracked 86,644 Firewalls SOCRadar · 2026-06-16
  11. 74,000 Fortinet firewall credentials exposed in FortiBleed data leak Help Net Security · 2026-06-18
  12. The Gentlemen ransomware: Inside one of the fastest-growing extortion operations Barracuda · 2026-08-17
  13. CVE-2025-64446: Fortinet FortiWeb zero-day path traversal vulnerability exploited in the wild Tenable · 2025-11-14
  14. FBI Warns FortiBleed Remains Active After Amassing 86,644 Fortinet Device Credentials The Hacker News · 2026-10-07

Researched and written by the R3KONX analysis desk from the cited primary material. The k0t findings rest on a single joint Group-IB and NyxLab investigation; trade-press reports summarise it and add no independent data, and no Malaysian government or CERT source on this campaign was found. Corrections to event@r3konx.asia.

R3KONX 2027

Work in this domain? So does the programme

4–6 May 2027, World Trade Centre Kuala Lumpur. Curated talks, hands-on training and The BattleGrid.