Case 002 · Incident: May–Jun 2026
CVE-2026-0257: A firewall flaw let ransomware gangs log in uninvited
CVE-2026-0257 is an authentication bypass in the GlobalProtect portal and gateway of Palo Alto Networks PAN-OS and Prisma Access. It lets attackers set up unauthorized VPN sessions when authentication override cookies are enabled with a particular certificate configuration. Palo Alto patched it in May 2026, and exploitation in the wild followed.
A Login That Should Not Exist
After the firewalls were breached, the defenders saw something strange in their VPN logs. New encrypted sessions were opening from machines that called themselves kali. Some of the addresses behind those sessions had also been seen attacking the same firewalls.[1] Arctic Wolf investigated several separate intrusions in June 2026. Every one ended in Qilin ransomware, and every one started at the same point: an exposed Palo Alto Networks firewall.[1] The weakness is tracked as CVE-2026-0257. According to Beazley Security, Palo Alto first released it on May 13, 2026. Security Affairs reports that Palo Alto addressed the flaw on May 13, and that about two weeks later the security firm Rapid7 confirmed active exploitation across multiple customer environments.[1][2][3]
The Open Door
GlobalProtect is the VPN that lets remote staff reach the corporate network through the firewall. The flaw is an authentication bypass in its portal and gateway. Palo Alto says it lets an attacker bypass security restrictions and establish an unauthorized VPN connection.[4] It was not every firewall. A device was only exposed when a feature called authentication override cookies was switched on, with a particular certificate setup. Beazley Security noted the risk appeared when that cookie certificate was reused elsewhere, such as for the portal's web service.[2][4] Palo Alto says its management platform Panorama and its Cloud NGFW product are not affected by these issues.[4]
Palo Alto's threat researchers say the attackers used the flaw to gain an initial foothold in corporate networks. They then got around multi-factor authentication, the second login step meant to stop stolen passwords, and set up a working VPN session.[5] Not every attempt worked. Palo Alto's researchers say only a small portion of the probed devices actually ended up with established VPN sessions. Palo Alto also acknowledged limited exploit attempts against unpatched devices that had no mitigations applied.[4][5]
Signals That Went Unread
The next part looks at what investigators found at the edge of the network. These are the traces they say defenders can search for. On May 29, 2026, a public proof-of-concept exploit was released. That is demonstration attack code, now in the open. Unit 42 published a separate set of indicators for activity before that date, so the earliest attacks left their own traces.[5] The exploitation came from numerous external IP addresses. Arctic Wolf took that as a sign of broad scanning and exploitation aimed at vulnerable GlobalProtect deployments.[1]
Some of those addresses showed up twice: first as the source of the attack, then as the origin of the VPN session that followed. The same infrastructure did both jobs, a clear link to connect in the logs.[1] The VPN sessions themselves carried one more detail. Arctic Wolf found they came from systems that identified themselves with the hostname kali.[1] Two days later, on May 31, 2026, Singapore's Cyber Security Agency went public. Its warning was plain: the flaw was being actively exploited in the wild.[6]
Inside The Network
Behind those warnings, investigators were reconstructing what happened once the attackers were inside. From that VPN session, the attackers moved inward one stage at a time. Palo Alto's researchers report internal reconnaissance with dual-use remote monitoring and management tools. That is software built for administrators, which attackers can also abuse. Arctic Wolf saw the SoftPerfect Network Scanner used to find reachable systems, and a tool called NetExec used to map the network and check credentials.[1][5]
Palo Alto's researchers say the attackers moved laterally over RDP, the Windows remote desktop protocol, and over SMB, its file-sharing protocol. In Arctic Wolf's cases, PsExec was the main tool for moving between machines. RDP gave interactive access to domain controllers, backup infrastructure and file servers.[1][5] Then came the credentials. The attackers dumped the memory of the LSASS process and pulled the NTDS database that stores Windows domain accounts. One credential dump was saved with a .odt extension, a disguise likely meant to slip past rules that look for the usual dump files.[1] Arctic Wolf saw the attackers use compromised administrative accounts to move through Windows administrative shares. Palo Alto's researchers describe the same phase as staging more tools for credential harvesting, data exfiltration and, in the end, ransomware deployment.[1][5]
To keep their grip, the attackers set up backup ways in. Arctic Wolf says they deployed several remote access tools on top of built-in persistence mechanisms. Security Affairs reported registry Run keys and scheduled tasks, alongside remote administration tools such as AnyDesk, Ngrok, LogMeIn and MeshAgent.[1][3] They also worked to go dark. A PowerShell routine cleared every Windows event log that held records, and in some cases Microsoft Defender's real-time protection was switched off before the ransomware ran.[1] Arctic Wolf's guidance captured the stakes: confirmed exploitation should be treated as a full domain compromise.[1]
Beneath The Operating System
In one case, Palo Alto's researchers say the attackers moved from the compromised VPN gateway to the vCenter management appliance. That is the console that manages a company's virtual machines.[5] From vCenter they disabled a security control on the ESXi hypervisors and ran a custom tool to read credentials directly from a domain controller's virtual disk, offline.[5] In that one vCenter case, the work happened underneath the guest operating systems. Palo Alto's researchers say the tactic bypasses endpoint detection and response, the security software running inside those machines. It gave the attackers domain-wide credentials.[5]
Palo Alto's researchers also observed actors chaining this initial access with exploit tools aimed at Veeam backup infrastructure. That backup flaw is tracked as CVE-2023-27532, and the goal was to maximize their impact.[5] Arctic Wolf found that this focus on backups was deliberate. The attackers targeted the Veeam backup infrastructure before deploying ransomware. The aim was to eliminate recovery options and pressure the victim to pay.[1]
The Ransom
Then came the ransomware. In every intrusion Arctic Wolf observed, the Qilin payload was named win dot exe. It was staged in deliberately chosen locations, such as the PerfLogs folder.[1] The binary would only run when fed a 32-character password at the moment of execution, which makes automated sandbox analysis and quick triage harder.[1] Encrypted files received an extension unique to each operation, a fingerprint typical of Qilin campaigns.[1]
Where data was stolen, Arctic Wolf says the main destination was MEGA, a legitimate cloud storage service. Security Affairs reported data theft using Rclone, ProtonDrive, FileZilla and MEGA. But not every intrusion involved theft. Forensic evidence showed cases of encryption without data exfiltration.[1][3] Qilin, also known as Agenda, runs a double-extortion model, encrypting data and threatening to leak it. It has operated as a ransomware-as-a-service business since at least 2022.[1] For context, Security Affairs reports that in 2025 Qilin claimed more than 40 victims a month, with a peak of 100 in June of that year.[3]
Arctic Wolf saw the tradecraft range from rapid encryption-only operations to full double extortion. It says this could suggest several affiliates working under the Qilin ransomware-as-a-service umbrella. That remains unconfirmed.[1] Unit 42 attributes the exploitation to two ransomware operators, Spikey Scorpius, known as Qilin, and Tarnished Scorpius, known as INC. Public reporting from ReliaQuest also links a group using Settra ransomware, which Unit 42 has not confirmed.[5][7] Unit 42 says its investigations, ongoing since early June 2026, spanned numerous industries. The Hacker News lists targeted sectors including education, financial services, healthcare, government, manufacturing, and utilities and energy. No source discloses how many organizations were hit, or who they were.[5][7]
Arctic Wolf assessed, with moderate confidence, that intrusions using this flaw and ending in Qilin ransomware are likely ongoing. That is an assessment, not a confirmed finding.[1]
What The Sources Dispute
The sources do not fully agree on how serious the flaw is. Palo Alto rates its severity HIGH, with a suggested urgency of HIGHEST. The Hacker News reports a CVSS score of 7.8, while Singapore's agency calls it critical, with a score of 9.1 out of 10. And according to Beazley Security, it was first published as medium severity.[2][4][6][7] The date the flaw entered the United States catalog of known exploited vulnerabilities also splits the record. Unit 42 says May 13, 2026. Beazley Security says May 29, and Security Affairs says early June.[2][3][5] Several questions are still open. Nobody has named the unidentified threat actor that Palo Alto's researchers first saw exploiting the flaw. The sources give no details of any INC intrusion beyond the attribution. And it is unclear when the activity after the attackers got in actually began.[5]
The Checklist
Unit 42 updated its threat brief on October 19, 2026 to add the ransomware group activity. In the cases investigated, what began as an authentication bypass at the network edge ended with ransomware.[1][5]
What to do
- Unit 42's advice was blunt: hunt for the published indicators, and treat any successful gateway-connected event tied to them as a reason to start incident response.
- If you run this kind of VPN portal, the first steps are concrete. Update affected products without delay. Once you have patched, end all active sessions so any unauthorized ones are cut off. If you cannot patch right away, pick one of two mitigations. Either turn off authentication override, or generate a new certificate used only for override cookies, and store it securely.
- Shrink what the internet can reach. Beazley reports that CISA recommends either turning off the vulnerable technology or limiting VPN access to trusted IP addresses. Check whether any GlobalProtect profile is mapped to management interface services, which Unit 42 warns can expose the management interface to the internet without administrators knowing. And expect users to sign in again once after the upgrade, because the fix regenerates their cookies.
- Assume the worst where it is warranted. Hunt for the published indicators, and start incident response for any successful gateway-connected event linked to them. If exploitation is confirmed, treat it as a full domain compromise and rotate credentials. Protect your backups as if they are a target. Patch Veeam backup servers against CVE-2023-27532, and restrict network access to the backup service so that only your management systems can reach it. Then keep at least one copy isolated from the main network or unchangeable, so it survives an attacker who goes looking for it.
Open questions
- The KEV addition date conflicts across sources: May 13, 2026 (Unit 42 and The Hacker News), May 29 (Beazley) and early June (Security Affairs).
- The severity scores differ: CVSS v4.0 7.8 / HIGH from Palo Alto, CVSS v3.1 9.1 / critical from Singapore's CSA, and Beazley says it was first published as medium severity.
- The total number of victim organizations and their identities are not disclosed in any source.
- Settra ransomware's involvement rests only on public reporting (ReliaQuest) and has not been confirmed by Unit 42.
- The sources give no INC (Tarnished Scorpius) intrusion details beyond the attribution.
- It is unclear whether the Qilin intrusions are one actor or several affiliates sharing infrastructure; Arctic Wolf presents both possibilities.
- The identity of the 'unidentified threat actor' in Unit 42's earliest observations is still unknown.
- Unit 42's brief still says no post-access behavior or lateral movement had been identified, which contradicts its later ransomware findings; the timeline of when post-access activity began is unclear.
- Which organization was hit in the vCenter/ESXi virtualization-layer case, and what the impact was, is not disclosed.
- Affected Prisma Access version ranges differ slightly between Arctic Wolf, Palo Alto's CPE data and Singapore's CSA.
Sources
- [1]Arctic Wolf
- [2]Beazley Security
- [3]Security Affairs
- [4]Palo Alto Networks
- [5]Unit 42 (Palo Alto Networks) — Threat Brief: Active Exploitation of PAN-OS CVE-2026-0257 (Updated October 20)
- [6]Cyber Security Agency of Singapore
- [7]The Hacker News — Qilin and INC Ransomware Groups Exploit Palo Alto Networks Flaw to Bypass MFA