CTIPilot
AI-generated · no human review · verify critical claims against the linked source. how it works →
‹Wed · 20 May 2026›
All daily briefs →
Daily brief · UTC day

Wednesday, 20 May 2026

13 verified findings from 1 run · 2 updates to prior coverage · the settled record for this UTC day, in the classic brief order.

CriticalUpdate Drupal core sites on PostgreSQL to 11.3.10, 11.2.12, 11.1.10, 10.6.9, 10.5.10 or 10.4.10CVE-2026-9082 · exploited · corrected 21 Oct 16:06Z
Criticality
Kind
Topic
Region
TL;DR · the day in one read
  1. 01Drupal core CVE-2026-9082: anonymous SQL injection on PostgreSQL sites, exploited in the wild from 2026-05-22. Drupal's SA-CORE-2026-004 (2026-05-20), pre-announced in PSA-2026-05-18, fixes CVE-2026-9082, an anonymous SQL injection in Drupal core's database abstraction API that affects only sites using PostgreSQL and can lead to information disclosure and in some cases privilege escalation or remote code execution. Drupal confirmed exploit attempts in the wild on 2026-05-22, raising its risk score from 20 to 23 of 25, and CISA added the flaw to its KEV catalog the same day; the fixed releases are 11.3.10, 11.2.12, 11.1.10, 10.6.9, 10.5.10 and 10.4.10. →
  2. 02Storm-2949 hijacked privileged Entra ID accounts through SSPR abuse, then mined Azure management features for data. Microsoft Threat Intelligence reported on 2026-05-18 that Storm-2949 abused Microsoft's Self-Service Password Reset to take over four privileged Entra ID accounts at one organization, re-enrolled MFA on its own devices, and used their Azure RBAC roles to pull App Service publishing profiles, Key Vault secrets, SQL and Storage data and VM access through Run Command and VMAccess. The actor relied mostly on legitimate cloud management features, installed ScreenConnect on VMs, aimed at mass data exfiltration and, per Microsoft, did not extort and shows no link to ShinyHunters. →
  3. 03Exploited Microsoft Defender flaws: CVE-2026-41091 gives a local attacker SYSTEM, CVE-2026-45498 denies service. CVE-2026-41091 (CVSS 7.8) is a link-following flaw in the Microsoft Malware Protection Engine that lets an authorized local attacker gain SYSTEM, and CVE-2026-45498 (CVSS 3.1 base score 4.0) is a denial of service in the Microsoft Defender Antimalware Platform. Microsoft marked both exploited on 2026-05-19 and CISA added both to its KEV catalog on 2026-05-20; the fixes are Engine version 1.1.26040.8 and Platform version 4.18.26040.7, which the default configuration installs automatically. The researcher who published the RedSun and UnDefend exploits in April says the two CVEs are those flaws. →
  4. 04The compromised actions-cool GitHub Actions were re-enabled in September with malicious tags and ran the payload again. On 2026-05-18 all 53 version tags of the actions-cool/issues-helper GitHub Action, and all 15 of actions-cool/maintain-one-comment, were moved to imposter commits whose payload reads the Runner.Worker process memory for decrypted workflow secrets and sends them over HTTPS to an attacker-controlled domain (StepSecurity); Socket places the compromise in the Mini Shai-Hulud campaign. GitHub disabled both repositories on 2026-05-19 but they became reachable again on 2026-09-16 with the malicious tags intact, so tag-referencing workflows ran the payload again and SafeDep counted six newly infected repositories by 2026-09-24, until GitHub disabled both again by 2026-09-25 (Socket, The Hacker News). Only workflows pinned to a full commit SHA from before 2026-05-18 were unaffected. →

01Active threats, incidents & disclosures6 items

NOTABLEupdatedNATOB2

Previously unknown flaw in Huawei enterprise router software caused POST Luxembourg's nationwide telecom outage of July 2025; no CVE ten months later

An attack exploiting a previously unknown vulnerability in Huawei enterprise router software caused POST Luxembourg's nationwide outage on 23 July 2025, according to multiple sources briefed on the matter, disrupting mobile, landline and emergency communications for more than three hours (The Record, 2026-05-19). POST's head of communications Paul Rausch called it a denial-of-service attack on a network device that exploited "a non-public, non-documented behaviour, for which no patch was available at the time", and anonymous sources described it as a zero-day attack (The Record, 2026-05-19). Specially crafted network traffic sent the Huawei enterprise routers into a continuous restart loop: the public prosecutor said corrupted data relayed through POST in its role as internet service provider made its systems stop and reboot instead of relaying it, and a spokesperson for the High Commission for National Protection said investigators found "no evidence that an attack was specifically directed at POST Luxembourg as a chosen target" (The Record, 2026-05-19). POST said neither of the earlier, publicly disclosed denial-of-service flaws in Huawei's VRP operating system was involved, Huawei told POST it had never seen the attack among its customers, and no CVE or public warning has been issued (The Record, 2026-05-19).

An attack exploiting a previously unknown vulnerability in Huawei enterprise router software caused a nationwide telecoms outage in Luxembourg last year, according to multiple sources briefed on the matter, disrupting mobile, landline and emergency communications for more than three hours.

It was caused by specially crafted network traffic that sent Huawei enterprise routers into a continuous restart loop, crashing critical parts of POST’s infrastructure.

no evidence that an attack was specifically directed at POST Luxembourg as a chosen target

The Record 2026-05-19
Correctionrun 2026-10-21T1148Z-auditbodyclassificationentitiesevidenceheadlinesourcessourcing_notesummarytagstechniquestitle

The sources do not name Huawei's VRP operating system as the vulnerable software: The Record reports a previously unknown vulnerability in Huawei enterprise router software and mentions VRP only for earlier denial-of-service flaws that POST says were not involved (The Record, 2026-05-19). The finding that no attack was specifically directed at POST came from a spokesperson for Luxembourg's High Commission for National Protection, not from prosecutors, and the zero-day label rests on anonymous sources; no source attributes the attack to a state actor (The Record, 2026-05-19).

threat20 May 05:00Zsingle-sourceOpen finding →
NOTABLEupdatedNATOB2

Microsoft DCU disrupts Fox Tempest, a malware-signing service that enabled Rhysida deployments and is linked to INC, Qilin and Akira affiliates

Microsoft's Digital Crimes Unit unsealed a legal case in the US District Court for the Southern District of New York against Fox Tempest, a malware-signing-as-a-service (MSaaS) operation active since May 2025. It seized the service's website, took hundreds of the virtual machines running the operation offline and blocked access to a site hosting its code (Microsoft On the Issues, 2026-05-19). Fox Tempest abused Microsoft Artifact Signing (formerly Azure Trusted Signing) to obtain Microsoft-issued code-signing certificates valid for only 72 hours, which let malware masquerade as software such as AnyDesk, Teams, PuTTY and Webex and bypass security controls. Microsoft assesses that the actor very likely used stolen identities based in the United States and Canada to pass Artifact Signing's identity validation (Microsoft Threat Intelligence, 2026-05-19).

Customers uploaded files through a web portal and, from February 2026, through pre-configured virtual machines hosted on Cloudzy infrastructure, and received signed binaries in return. They chose plans priced at $5,000, $7,500 or $9,000, with higher plans getting priority in the queue (Microsoft Threat Intelligence, 2026-05-19). Microsoft has revoked over a thousand certificates attributed to Fox Tempest (Microsoft Threat Intelligence, 2026-05-19) and reports the operators trying to move operations and customers to another code-signing service (Microsoft On the Issues, 2026-05-19).

Microsoft links Fox Tempest to Vanilla Tempest, Storm-0501, Storm-2561 and Storm-0249, which all used Fox Tempest-signed malware in active intrusions delivered through purchased advertisements, malvertising and SEO poisoning. Separately, cryptocurrency analysis ties the actor to affiliates delivering the INC, Qilin and Akira ransomware families (Microsoft Threat Intelligence, 2026-05-19). In Microsoft's case study, Vanilla Tempest bought ads that sent users searching for Microsoft Teams to fraudulent download pages serving a signed, trojanized Teams installer, which deployed the Oyster backdoor (also known as Broomstick) and in some cases Rhysida ransomware. Microsoft lists healthcare, education, government and financial services among the affected sectors, and the United States, France, India and China among the affected countries (Microsoft Threat Intelligence, 2026-05-19).

Triage: a genuine Teams, AnyDesk, PuTTY or Webex installer carries its publisher's signature. A Fox Tempest-signed copy carries a valid Microsoft-issued Artifact Signing certificate in the name of an entity that is not the product's publisher, obtained, Microsoft assesses, with stolen identities (Microsoft Threat Intelligence, 2026-05-19).

Correctionrun 2026-09-30T0639Z-auditsummarybodyclassificationtechniquesprioritytitleheadline

Microsoft assesses, rather than states, that Fox Tempest very likely used stolen identities based in the United States and Canada, and its customers paid for plans of $5,000, $7,500 or $9,000 with queue priority, not per signing run (Microsoft Threat Intelligence, 2026-05-19). INC, Qilin and Akira are tied to Fox Tempest through cryptocurrency links to affiliates delivering those ransomware families, not named as confirmed customers (Microsoft Threat Intelligence, 2026-05-19). No cited source reports European public-sector or healthcare victims of these affiliates. Earlier advice about installers spawning shells and about alerting on Key Vault certificate creation had no source. The earlier text also named the installer's file name and the seized domain; both are removed. Microsoft describes a trojanized Teams installer that deployed the Oyster backdoor (Microsoft Threat Intelligence, 2026-05-19). The court case is described by Microsoft as a legal case unsealed in the Southern District of New York (Microsoft On the Issues, 2026-05-19).

threat20 May 05:00Zmulti-sourceOpen finding →
ROUTINEupdatedNATOC3

The Gentlemen ransomware group lists Czech university VSFS and Swiss engineering firm DEVO-Tech AG on its leak site

The Gentlemen ransomware group claimed the Czech private University of Finance and Administration (VSFS) and the Swiss engineering firm DEVO-Tech AG on its leak site on 2026-05-18, threatening to publish stolen data unless the victims made contact or negotiated (DeXpose, 2026-05-19) (DeXpose, 2026-05-19); DEVO-Tech is a family-owned engineering company from Ziefen (ransomware.live, 2026-05-18). The listings carry no technical detail. Check Point Research's analysis of the group's leaked internal chats describes a locker written in Go and initial access mostly through exposed VPN appliances, firewalls and other edge devices, chiefly Fortinet FortiGate and Cisco, by brute-forcing credentials, exploiting known vulnerabilities or buying access (Check Point Research, 2026-05-13).

The full leak will be published soon, unless a university representative contacts us via the channels provided.

The full leak will be published unless DEVO-Tech AG begins negotiations with us.

DeXpose 2026-05-19

The group claims to usually gain initial access through exposed edge devices such as VPN appliances, firewalls, and other internet-facing systems, with a particular focus on platforms like Fortinet FortiGate and Cisco.

Check Point Research 2026-05-13
Correctionrun 2026-10-21T1148Z-auditbodyclassificationconfidenceentitiesevent_dateevidenceheadlinepriorityreferencessourcessourcing_notesummarytagstechniquestitleverification

Both victims were listed on 2026-05-18 (DeXpose, 2026-05-19) (DeXpose, 2026-05-19). The listings carry no technical detail, so nothing in them shows the group's methods unchanged; the group profile is Check Point Research's analysis of 2026-05-13, which names exposed VPN appliances and firewalls, chiefly FortiGate and Cisco, as the usual entry (Check Point Research, 2026-05-13), not SD-WAN gateways.

Builds on: Leaked chats of The Gentlemen ransomware group show its edge-device access, NTLM relay and…

incident20 May 05:00Zsingle-sourceOpen finding →
HIGHupdatedNATOB2

actions-cool/issues-helper GitHub Action compromised in Mini Shai-Hulud: 53 tags moved to imposter commits that read Runner.Worker memory, re-enabled in September with the tags intact

StepSecurity disclosed on 2026-05-18 that every existing tag of the actions-cool/issues-helper GitHub Action, 53 in total, had been moved to its own imposter commit that is not reachable from the action's default branch, and that all 15 tags of the companion actions-cool/maintain-one-comment action were moved the same way (StepSecurity, 2026-05-18). The payload downloads the Bun JavaScript runtime to the runner, then a python3 child process reads the memory of the Runner.Worker process, which holds the workflow's decrypted secrets, filters the captured bytes for values flagged as secret and sends them over HTTPS to an attacker-controlled domain (StepSecurity, 2026-05-18). The 53 imposter commits carry fake "Build action for vX.Y.Z" messages in the maintainer's style and were all created within 3 minutes 16 seconds on 2026-05-18 (StepSecurity, 2026-05-18). Socket's head of threat intelligence told The Hacker News that the @antv npm compromise in the Mini Shai-Hulud campaign is likely linked to the actions-cool hack because the two share an exfiltration domain, and that "the overlap is strong enough that we're treating them as related" (The Hacker News, 2026-05-19). GitHub's security team disabled both repositories on 2026-05-19 but left the malicious tags in place; both became reachable again on 2026-09-16, so every workflow referencing either action by tag ran the payload again until GitHub disabled them a second time by 2026-09-25 (Socket, 2026-09-24). SafeDep found six more repositories infected through issues-helper between 2026-09-20 and 09-24, with the payload also committing Claude Code and VS Code hooks that run on a developer's machine when the project is opened (The Hacker News, 2026-09-25).

Triage: an issue-housekeeping action has no reason to install a separate JavaScript runtime or run a multi-minute script; a scheduled workflow that switches from failing in seconds to running for minutes without any change to its own code means the tagged action became downloadable again (Socket, 2026-09-24).

GitHub security team disabled both repositories on May 19, 2026, one day after the malicious content was introduced.

On September 16, 2026, both repositories became accessible again. Their release tags were not cleaned up first.

have been disabled on GitHub again, so workflows referencing them now fail at job setup instead of running the payload.

Socket 2026-09-24

The payload steals the CI secrets. Then it commits Claude Code and VS Code hooks into the repository.

The Hacker News (quoting SafeDep)
Correctionrun 2026-09-30T0639Z-auditbodyclassificationtechniquestitleheadlinesummaryevent_date

The Hacker News reports that GitHub disabled access to the companion maintain-one-comment repository, and Socket described the link to the Mini Shai-Hulud npm campaign as likely, based on the shared exfiltration domain, rather than confirmed; no cited source ties it to PyPI (The Hacker News, 2026-05-19). The Nx Console extension compromise named in the earlier summary is a separate incident, and no cited source says EU or Swiss organisations were affected.

Updaterun 2026-10-20T0835Z-auditactionsbodyevidenceheadlinesourcessummarytechniquestitle

Socket reported on 2026-09-24 that GitHub's security team had disabled both actions-cool/issues-helper and actions-cool/maintain-one-comment on 2026-05-19, not only maintain-one-comment, and that both became reachable again on 2026-09-16, between 09:09 and 16:16 UTC, with every release tag still pointing at the content introduced on 2026-05-18, so any workflow referencing either action by tag ran the payload again on its next run; Socket could not determine why they were re-enabled (Socket, 2026-09-24). Socket places the compromise in the May 2026 Mini Shai-Hulud campaign and notes that GitHub's dependency graph lists about 15,000 dependent repositories for issues-helper alone (Socket, 2026-09-24). SafeDep found six more repositories infected through issues-helper between 2026-09-20 and 09-24, and says the payload steals the CI secrets and then commits Claude Code and VS Code hooks into the repository, which run the malware on the machine of a developer who pulls the change and opens the project (The Hacker News, 2026-09-25). GitHub disabled both repositories again by 2026-09-25, so referencing workflows now fail at job setup (Socket, 2026-09-24; The Hacker News, 2026-09-25).

threat20 May 05:00Zmulti-sourceOpen finding →
NOTABLEupdatedNATOA1

Nx Console VS Code extension version 18.95.0 shipped a credential stealer, published with a contributor token stolen in the TanStack npm compromise

An attacker published Nx Console version 18.95.0 to the Visual Studio Marketplace at 12:30 UTC on 2026-05-18, where it stayed until 12:48 UTC, and to Open VSX from 12:33 to 13:09 UTC (Nx GHSA-c9j4-9m59-847w, 2026-05-18), times StepSecurity's account puts at 12:36 UTC for the Marketplace publication (StepSecurity, 2026-05-18); the Marketplace upload passed Microsoft's automated verification, and a maintainer caught it from the routine publisher-notification e-mail (Nx, 2026-05-21). The attacker published as a legitimate core contributor whose GitHub CLI OAuth token had been exfiltrated seven days earlier by a credential stealer that arrived through the 2026-05-11 TanStack npm supply-chain compromise, and no approval gate stopped a single contributor from releasing (Nx, 2026-05-21). The attacker had also planted an unsigned orphan commit in the nrwl/nx repository; the code injected into the extension runs when any workspace opens, fetches the 498 KB obfuscated payload from that commit (StepSecurity, 2026-05-18) and installs the Bun JavaScript runtime to run it (The Hacker News, 2026-05-19). The payload harvests Vault, npm, AWS (including the instance and ECS metadata services), GitHub, 1Password and filesystem secrets, exfiltrates them over HTTPS, the GitHub API and DNS, writes a Python backdoor with a macOS LaunchAgent for persistence and on Linux attempts a sudoers change (Nx, 2026-05-21); the macOS backdoor takes commands signed with a 4096-bit RSA key through the GitHub Search API as a dead drop (StepSecurity, 2026-05-18). Microsoft counts 28 Marketplace installs and Open VSX 41 downloads, while Nx's own analytics registered about 6,000 activations; the Nx CLI, the official plugins and Nx Cloud were not affected (Nx, 2026-05-21).

A malicious version of Nx Console 18.95.0 was published at 12:30 PM UTC and removed soon after at 12:48 PM UTC

Nx (GHSA-c9j4-9m59-847w) 2026-05-18

The attacker published the malicious version as a legitimate Nx core contributor.

Anyone who installed v18.95.0 during the exposure window (2026-05-18, 12:30-13:09 UTC) should treat their machine as compromised and rotate credentials.

Nx (postmortem) 2026-05-21

It also installs a persistent Python backdoor on macOS that uses the GitHub Search API as a dead-drop for receiving further commands signed with a 4096-bit RSA key.

StepSecurity 2026-05-18
Correctionrun 2026-10-21T1148Z-auditaffected_productsbodyclassificationentitiesevent_dateevidenceheadlinesourcessourcing_notesummarytechniquestitle

The malicious version was live on the Visual Studio Marketplace from 12:30 to 12:48 UTC and on Open VSX from 12:33 to 13:09 UTC (Nx GHSA-c9j4-9m59-847w, 2026-05-18); 12:36 to 12:47 UTC is only the span from the maintainer's alert e-mail to the Marketplace unpublish, and Nx gives the exposure window as 12:30 to 13:09 UTC (Nx, 2026-05-21). The publishing credential was a core contributor's GitHub CLI OAuth token, stolen a week earlier through the TanStack npm compromise (Nx, 2026-05-21). The 2.2 million installs are the extension's total (StepSecurity, 2026-05-18); only 18.95.0 was malicious, with 28 Marketplace installs by Microsoft's count and about 6,000 activations by Nx's, and the upload was not too short-lived to scan: it passed Microsoft's automated checks, basic malware scans included (Nx, 2026-05-21).

threat20 May 05:00Zmulti-sourceOpen finding →
NOTABLECVE-2026-42096 +4updatedNATOA2

Sparx Pro Cloud Server and Enterprise Architect: five CVEs including unauthenticated SQL execution and a WebEA race to code execution, public PoC, no vendor fix at disclosure

CERT Polska coordinated the disclosure of five vulnerabilities in Sparx Systems Pro Cloud Server (PCS) and Enterprise Architect (EA) on 2026-05-19, reported by Blazej Adamczyk (br0x) of Efigo (CERT Polska, 2026-05-19). CVE-2026-42097 lets an attacker omit the "model" query parameter and send the model name only in the POST request's binary blob, so PCS executes SQL queries without authentication, and CVE-2026-42096 lets any low-privileged user run arbitrary SQL queries in the database user context because permission checks are missing (CERT Polska, 2026-05-19). CVE-2026-42099 is a race condition in WebEA's /data_api/dl_internal_artifact.php: an attacker with repository access controls the name and contents of the file it writes, so a delayed response keeps a malicious PHP file reachable long enough for a second request to execute it; CVE-2026-42098 lets an authenticated EA user modify the client, for example with a debugger, to log in as any other user or administrator; and CVE-2026-42100 makes the PCS service terminate on a crafted SQL query (CERT Polska, 2026-05-19). The researcher scores the chain CVSS 4.0 10.0 because, with WebEA installed, the flaws give remote unauthenticated code execution in the web server context (sploit.tech, 2026-05-19); CCB lists CVSS 4.0 scores of 9.3 for CVE-2026-42097, 8.7 for CVE-2026-42096 and -42098, 7.7 for CVE-2026-42099 and 7.1 for CVE-2026-42100 (CCB, 2026-05-21). Proof-of-concept code exists for all five and no exploitation was known as of 2026-05-20: CCB's advisory title calls the flaws actively exploited, but its text reports no proof of exploitation as of that date (CCB, 2026-05-21); the published exploit script ships without the working key (GitHub br0xpl/sparx_hack), which sits in the freely downloadable EA client binary (sploit.tech, 2026-05-19). The vendor gave CERT Polska no details or vulnerable version range (CERT Polska, 2026-05-19) and had resolved none of the CVEs at disclosure (sploit.tech, 2026-05-19); Sparx's later Pro Cloud Server 6.5 release notes list WebEA security fixes, among them a corrected vulnerability when downloading internal artifacts and added validation of the model number parameter, without naming these CVEs (Sparx Systems).

Triage: a legitimate EA client names its model in the model query parameter of the request URL; a POST to the endpoint with no or a wrong model parameter is the bypass, which is why the researcher suggests a proxy that drops such requests (sploit.tech, 2026-05-19).

An attacker can omit the "model" query parameter and send the model name only in the binary blob in POST request allowing SQL query execution without authentication.

Only Pro Cloud Server version 6.1 (build 167) and below and Enterprise Architect version 17.1 and below were tested and confirmed as vulnerable, other versions were not tested and might also be vulnerable.

CERT Polska 2026-05-19

In case where PCS is installed with WebEA the vulnerabilities allow further for remote unauthenticated code execution (RCE) within the web server context.

sploit.tech (Blazej Adamczyk) 2026-05-19

There is no proof of exploitation as of the writing of this advisory, as of 2026-05-20, but since a PoC is available, it is highly likely that those vulnerabilities will be exploited.

Centre for Cybersecurity Belgium 2026-05-21
Correctionrun 2026-10-21T1148Z-auditactionsaffected_productsbodyclassificationcvesevidenceheadlinepriorityregionssectorssourcessourcing_notesummarytagstechniquestitleverification

CVE-2026-42100 is improper handling of a syntactically invalid structure (CWE-228), not CWE-835 (CERT Polska, 2026-05-19), and CCB lists a CVSS 4.0 score for every one of the five, 8.7 for CVE-2026-42096 and 7.1 for CVE-2026-42100 included (CCB, 2026-05-21). Only CVE-2026-42097 needs no account: CVE-2026-42096, -42098, -42099 and -42100 need a low-privileged login or repository access (CCB, 2026-05-21), and the two SQL flaws are unauthorized SQL execution rather than SQL injection (CERT Polska, 2026-05-19).

The authentication bypass shows as a POST to /SparxCloudLink.sseap without the model query parameter, not as an _api/data request; the chained CVSS 4.0 10.0 is the researcher's score (sploit.tech, 2026-05-19), and the public exploit code ships without the working key (GitHub br0xpl/sparx_hack). No source states how widely Sparx tools are used in Swiss or EU government. Sparx's Pro Cloud Server 6.5 notes list WebEA security fixes without naming these CVEs (Sparx Systems).

threat20 May 05:00Zsingle-sourceOpen finding →
CRITICALCVE-2026-9082exploitedupdatedNATOA1

CVE-2026-9082: anonymous SQL injection in Drupal core on PostgreSQL, Highly critical, exploited and CISA KEV-listed

The Drupal Security Team pre-announced a Highly critical core release on 2026-05-18, scored 20 of 25 (AC:None, A:None, E:Theoretical, TD:Uncommon), and urged administrators to reserve the 2026-05-20 window from 17:00 to 21:00 UTC because "exploits might be developed within hours or days" (Drupal PSA-2026-05-18). SA-CORE-2026-004 followed on 2026-05-20 with CVE-2026-9082: a flaw in core's database abstraction API lets an anonymous attacker send specially crafted requests that result in arbitrary SQL injection on sites using PostgreSQL, which can lead to information disclosure and in some cases privilege escalation, remote code execution or other attacks (Drupal SA-CORE-2026-004). Searchlight Cyber traced it to the PostgreSQL entity-query override of the case-insensitive IN operator, which builds SQL placeholder names from the keys of the submitted value array, so attacker-chosen JSON object keys reach the SQL text through two anonymous entry points, the JSON login endpoint and the JSON:API filter syntax (Searchlight Cyber, 2026-05-21). The fixed releases are 11.3.10, 11.2.12, 11.1.10, 10.6.9, 10.5.10 and 10.4.10, and Drupal 9 and 8.9 sites are told to try applying a best-effort patch file by hand (Drupal SA-CORE-2026-004); Drupal 7 is not affected (Drupal PSA-2026-05-18). On 2026-05-22 Drupal raised the risk score to 23 of 25 because exploit attempts were being detected in the wild (Drupal SA-CORE-2026-004), and CISA added the flaw to its Known Exploited Vulnerabilities catalog the same day (CISA KEV catalog, 2026-05-22); the Drupal CNA scores it CVSS 9.8 (ENISA EUVD-2026-31153). By 2026-05-21 Imperva had observed over 15,000 attack attempts against almost 6,000 sites in 65 countries, most of it probing (Imperva, 2026-05-21).

Triage: most of this traffic is automated probing that tests whether a site is vulnerable (Imperva, 2026-05-21); a probe against a MySQL or SQLite site, or one already on a fixed release, is noise, while the same request against an unpatched PostgreSQL site is an attempt that calls for the account review above (Drupal SA-CORE-2026-004).

exploits might be developed within hours or days.

This issue is being protected by Drupal Steward.

Drupal Security Team (PSA-2026-05-18) 2026-05-18

A vulnerability in this API allows an attacker to send specially crafted requests, resulting in arbitrary SQL injection for sites using PostgreSQL databases.

The risk score has been updated to reflect that exploit attempts are now being detected in the wild.

Drupal Security Team (SA-CORE-2026-004) 2026-05-20

Since CVE-2026-9082 was released, Imperva has observed over 15,000 attack attempts targeting almost 6,000 individual sites across 65 countries.

Most of the observed activity so far appears to be probing.

Imperva 2026-05-21

There are two anonymous entry points that we have confirmed reach this loop with attacker-controlled array keys: the JSON login endpoint and the JSON:API filter syntax.

Searchlight Cyber 2026-05-21
Updaterun 2026-05-21-77cdc4cdcvesprioritysectorssourcestagsbody

Drupal published SA-CORE-2026-004 on 2026-05-20 with CVE-2026-9082: a vulnerability in core's database abstraction API allows an attacker to send specially crafted requests resulting in arbitrary SQL injection for sites using PostgreSQL, it can be exploited by anonymous users, and it "can lead to information disclosure, and in some cases privilege escalation, remote code execution, or other attacks" (Drupal SA-CORE-2026-004). The pre-announcement had scored it 20 of 25 and warned that "exploits might be developed within hours or days" (Drupal PSA-2026-05-18).

Affected versions run from 8.9.0 up to (not including) 10.4.10, 10.5.0 before 10.5.10, 10.6.0 before 10.6.9, 11.0.0 before 11.1.10, 11.2.0 before 11.2.12 and 11.3.0 before 11.3.10, and the fixed releases are 10.4.10, 10.5.10, 10.6.9, 11.1.10, 11.2.12 and 11.3.10; Drupal 9 and 8.9 sites get best-effort patch files to apply manually (Drupal SA-CORE-2026-004), and Drupal 7 is not affected (Drupal PSA-2026-05-18). Sites using the Drupal Steward web application firewall are already protected from known attack vectors but should upgrade in the near future (CSO Online, 2026-05-20). NCSC.ch had carried the pre-announcement on 2026-05-19 (NCSC.ch Security Hub).

Updaterun 2026-05-23-852c21c8actionscvesevidenceimmediate_actionprioritysourcestagsbody

On 2026-05-22 Drupal updated SA-CORE-2026-004, raising the risk score to 23 of 25 because exploit attempts are now being detected in the wild (Drupal SA-CORE-2026-004); NCSC.ch marked Security Hub post 12584 as actively exploited the same day (NCSC.ch Security Hub), and CISA added CVE-2026-9082 to its Known Exploited Vulnerabilities catalog on 2026-05-22 (CISA KEV catalog, 2026-05-22).

Imperva observed over 15,000 attack attempts targeting almost 6,000 sites across 65 countries, most of it probing of JSON:API routes, with Gaming and Financial Services sites drawing almost half the attacks (Imperva, 2026-05-21). Searchlight Cyber's analysis shows attacker-controlled array keys, sent as JSON object keys to the anonymous JSON login endpoint or through the JSON:API filter syntax, reaching the placeholder names that the PostgreSQL entity-query override builds for the case-insensitive IN operator (Searchlight Cyber, 2026-05-21).

Correctionrun 2026-10-21T1148Z-auditactionsaffected_productsbodyclassificationcvesevidenceheadlineimmediate_actionkindregionssectorssourcessourcing_notesummarytagstechniquestitle

A score of 20 of 25 is Drupal's top tier, Highly critical, which covers 20 to 25, not the second-highest (Drupal security risk levels); it stayed below 25 because no exploit existed yet and only uncommon configurations were affected (The Register, 2026-05-19), and on 2026-05-22 it rose to 23 of 25 with exploitation detected (Drupal SA-CORE-2026-004). A mitigation did exist before the patch: the PSA said the issue was protected by Drupal Steward (Drupal PSA-2026-05-18). Only Drupal 9 and 8.9 sites were left with patch files to apply by hand; the end-of-life 11.1.x and 10.4.x lines received full releases, 11.1.10 and 10.4.10 (Drupal SA-CORE-2026-004).

The Drupal CNA scores CVE-2026-9082 CVSS 9.8 (ENISA EUVD-2026-31153); 6.5 is NIST's score (BleepingComputer, 2026-05-22). Moving a site off PostgreSQL is not a control any source offers; the sourced interim measure is Drupal Steward. Imperva's figures count attack attempts, mostly probing, aimed chiefly at Gaming and Financial Services sites (Imperva, 2026-05-21). No source names Swiss, EU or university Drupal sites as targets or says Drupal dominates Swiss public portals; the Federal Administration's standard web platform is based on Livingdocs (Swiss Federal Chancellery).

vulnerability20 May 05:00Zmulti-sourceOpen finding →
HIGHCVE-2026-41091 +1exploitedupdatedNATOA1

Microsoft published CVE-2026-41091 on 2026-05-19 as exploited and publicly disclosed: improper link resolution before file access (CWE-59) in Microsoft Defender lets an authorized attacker elevate privileges locally and gain SYSTEM, scored CVSS 7.8 (MSRC, 2026-05-19). The flaw sits in the Microsoft Malware Protection Engine (mpengine.dll); the last affected build is Engine version 1.1.26030.3008 and the first fixed build is Engine version 1.1.26040.8 (MSRC, 2026-05-19). Microsoft marked CVE-2026-45498, a Defender denial of service (CWE-400, CVSS 3.1 base score 4.0), as exploited the same day and fixed it separately in Antimalware Platform version 4.18.26040.7 (MSRC, 2026-05-19); CISA added both to its Known Exploited Vulnerabilities catalog on 2026-05-20 based on evidence of active exploitation (CISA, 2026-05-20). The researcher Nightmare Eclipse published the RedSun and UnDefend exploits on 2026-04-16 (Help Net Security, 2026-04-17) and has said that CVE-2026-41091 is RedSun and CVE-2026-45498 is UnDefend (The Hacker News, 2026-05-21); UnDefend lets a standard user block Defender from receiving signature updates (Help Net Security, 2026-04-17). RedSun baits Defender's real-time protection with an embedded EICAR test string and, using the Cloud Files API, oplocks and a directory junction, redirects the file rewrite of the resulting remediation into a protected system path, after which the Cloud Files infrastructure runs the attacker's binary as SYSTEM (Vectra AI, 2026-04-21).

Triage: an EICAR detection on its own is what an antivirus test produces; the RedSun pattern adds a directory junction that redirects the remediation's file rewrite into a protected system path and a SYSTEM-context process launched right after the remediation (Vectra AI, 2026-04-21).

Improper link resolution before file access ('link following') in Microsoft Defender allows an authorized attacker to elevate privileges locally.

First version of the Microsoft Malware Protection Engine with this vulnerability addressed Version 1.1.26040.8

Systems that have disabled Microsoft Defender are not in an exploitable state.

MSRC CVE-2026-41091 2026-05-19

CISA has added seven new vulnerabilities to its Known Exploited Vulnerabilities (KEV) Catalog

CISA 2026-05-20

The researcher has also confirmed that CVE-2026-45498 is UnDefend and CVE-2026-41091 is RedSun.

The Hacker News 2026-05-21

The attacker dropped the exploit files into the user’s Pictures and Downloads folders and renamed them to avoid suspicion.

Help Net Security 2026-04-17

Defender sees a known signature, initiates a remediation cycle, and RedSun wins the race to redirect the resulting file rewrite.

Vectra AI 2026-04-21
Updaterun 2026-05-22-5b90d5a1actionscvesevidencesourcesbody

Microsoft also marks CVE-2026-45498 as exploited: a Microsoft Defender denial of service (CVSS 3.1 base score 4.0) fixed in Antimalware Platform version 4.18.26040.7 (MSRC, 2026-05-19), while CVE-2026-41091 is fixed in Malware Protection Engine version 1.1.26040.8 (MSRC, 2026-05-19). CISA added both to its Known Exploited Vulnerabilities catalog on 2026-05-20 (CISA, 2026-05-20). The Hacker News notes that the two descriptions overlap with RedSun and UnDefend, the Defender zero-days Nightmare Eclipse disclosed in April, and that Huntress has observed exploitation of both (The Hacker News, 2026-05-21).

Correctionrun 2026-10-21T1148Z-auditactionsaffected_productsbodyclassificationcvesevidenceheadlinesourcessourcing_notesummarytagstechniquestitleverification

The two flaws are fixed by separate updates rather than one combined out-of-band release: Malware Protection Engine version 1.1.26040.8 fixes CVE-2026-41091, and Microsoft typically ships engine updates once a month or as needed (MSRC, 2026-05-19); Antimalware Platform version 4.18.26040.7 fixes CVE-2026-45498, which is a denial of service (CWE-400), not a privilege escalation (MSRC, 2026-05-19). The "Turn off routine remediation" policy named earlier only decides whether Defender remediates detected threats automatically or prompts the user (Microsoft Learn); it does not hold back engine updates, so the check is whether engine and platform updates reach each host.

The hunt signals given earlier, junction events on Defender paths and SYSTEM processes spawned by MsMpEng.exe, have no source. The behavior observed in the wild is renamed exploit binaries staged in the user's Pictures and Downloads folders (Help Net Security, 2026-04-17) and, for RedSun, a Defender remediation of an EICAR test string whose file rewrite is redirected so that the Cloud Files infrastructure runs the attacker's binary as SYSTEM (Vectra AI, 2026-04-21). The researcher has since said CVE-2026-41091 is RedSun and CVE-2026-45498 is UnDefend (The Hacker News, 2026-05-21).

vulnerability20 May 05:00Zmulti-sourceOpen finding →
NOTABLECVE-2026-24118 +20updatedNATOA2

vm2 Node.js sandbox: two waves of critical sandbox escapes in May 2026, fixed in 3.11.2 and then 3.11.4

The Hacker News reported on 2026-05-07 a dozen critical vulnerabilities in vm2, the Node.js library that runs untrusted JavaScript in a sandbox by intercepting and proxying objects, which let an attacker break out and execute arbitrary code on the host; fixes landed across versions 3.10.5 to 3.11.2, and users were told to update to 3.11.2 (The Hacker News, 2026-05-07). Among them, CVE-2026-43997 (CVSS 10.0) obtains the host Object despite earlier protections (vm2 GHSA-47x8-96vw-5wg6), CVE-2026-44006 (10.0) reaches BaseHandler.getPrototypeOf to get arbitrary prototypes (vm2 GHSA-qcp4-v2jj-fjx8), and CVE-2026-43999 (9.9) bypasses NodeVM's builtin allowlist whenever the module builtin is allowed, including through the '*' wildcard, so sandboxed code can load excluded builtins such as child_process (vm2 GHSA-947f-4v7f-x2v8); CVE-2026-44008 escapes through neutralizeArraySpeciesBatch() and CVE-2026-44009 through a null-proto exception, both 9.8 and fixed in 3.11.2 (The Hacker News, 2026-05-07). On 2026-05-18 the maintainer published a second set of advisories affecting 3.11.3 and earlier and fixed in 3.11.4 (vm2 GHSA-v6mx-mf47-r5wg), covered by BSI's WID-SEC-2026-1583 with CVE-2026-47131, CVE-2026-47135, CVE-2026-47137, CVE-2026-47139, CVE-2026-47140, CVE-2026-47141, CVE-2026-47208, CVE-2026-47209 and CVE-2026-47210; BSI says a remote, anonymous attacker can use them to execute arbitrary code, bypass security measures and disclose confidential information (BSI WID-SEC-2026-1583). Kodem Security describes the wave as vulnerability research rather than campaign reporting (Kodem Security, 2026-05-14).

Triage: the discriminator is whether the Node.js runtime is expected to start shells at all: Kodem's indicator is a shell parented to a runtime that should not invoke one, or a child_process call traced to a sandboxed context (Kodem Security, 2026-05-14).

A dozen critical security vulnerabilities have been disclosed in the vm2 Node.js library that could be exploited by bad actors to break out of the sandbox and execute arbitrary code on susceptible systems.

The Hacker News 2026-05-07

NodeVM's builtin allowlist can be bypassed when the module builtin is allowed (including via the '*' wildcard).

vm2 (GHSA-947f-4v7f-x2v8)

Untrusted JavaScript reaches vm2. In AI agent environments, the source is often LLM output produced in response to a manipulated prompt.

Kodem Security 2026-05-14
Correctionrun 2026-10-21T1148Z-auditaffected_productsbodyclassificationcvesevent_dateevidenceheadlinesectorssourcessourcing_notesummarytechniquestitleverification

The twelve critical vm2 CVEs were reported on 2026-05-07, with 3.11.2 as the version to update to (The Hacker News, 2026-05-07); BSI's WID-SEC-2026-1583 of 2026-05-18 covers a later, different set, CVE-2026-47131 and eight others, fixed in 3.11.4 (BSI WID-SEC-2026-1583). CVE-2026-43999 bypasses NodeVM's builtin allowlist when the module builtin is allowed, not when child_process is (vm2 GHSA-947f-4v7f-x2v8); CVE-2026-43997 obtains the host Object (vm2 GHSA-47x8-96vw-5wg6) while BaseHandler.getPrototypeOf is CVE-2026-44006 (vm2 GHSA-qcp4-v2jj-fjx8); and CVE-2026-44008 and CVE-2026-44009 are separate escapes, through neutralizeArraySpeciesBatch() and through a null-proto exception (The Hacker News, 2026-05-07). Kodem Security attributes the prompt-to-host-execution framing to Microsoft Security Research and gives no quoted chain (Kodem Security, 2026-05-14).

vulnerability20 May 05:00Zsingle-sourceOpen finding →
NOTABLECVE-2026-31635updatedNATOB2

DirtyDecrypt (DirtyCBC): public root exploit for a Linux kernel RxGK page-cache write, mapped to CVE-2026-31635, on kernels built with CONFIG_RXGK

V12 found and reported DirtyDecrypt on 2026-05-09 and was told by the kernel maintainers that it duplicated a flaw already patched in mainline; Delphos Labs found it independently and calls it DirtyCBC (BleepingComputer, 2026-05-18). V12 describes it as an rxgk page-cache write due to a missing copy-on-write guard in rxgk_decrypt_skb (BleepingComputer, 2026-05-18), and Moselwal, quoted by The Hacker News, explains that the function decrypts incoming socket buffers on memory pages partly shared with other processes' page cache, so without the guard data is written into the memory of privileged processes or the page cache of privileged files such as /etc/shadow, /etc/sudoers or a SUID binary, giving root (The Hacker News, 2026-05-19). A working proof of concept from V12 was public by 2026-05-18, tested only against Fedora and the mainline kernel (BleepingComputer, 2026-05-18), and in containerized environments a vulnerable worker node could give a pod a way out (The Hacker News, 2026-05-19). Neither discoverer published a CVE id: Will Dormann says the details align with CVE-2026-31635, patched on 2026-04-25, and The Hacker News notes that the NVD record links the DirtyDecrypt proof of concept (BleepingComputer, 2026-05-18; The Hacker News, 2026-05-19). The kernel's own description of CVE-2026-31635 differs, an inverted length check that lets oversized RESPONSE authenticators reach rxgk_decrypt_skb() and hit a BUG_ON, scored CVSS 3.1 7.5 as a network denial of service, affecting kernels from 6.16 and fixed in 6.18.23, 6.19.13 and 7.0 (ENISA EUVD-2026-25528).

This limits the attack surface to Linux distributions that closely follow the latest upstream kernel releases, including Fedora, Arch Linux, and openSUSE Tumbleweed.

However, V12's proof-of-concept exploit has only been tested against Fedora and the mainline Linux kernel.

BleepingComputer 2026-05-18

In containerized environments, worker nodes running a vulnerable version of Linux could provide a pathway to escape the pod.

The Hacker News 2026-05-19
Correctionrun 2026-10-21T1148Z-auditactionsaffected_productsbodyclassificationconfidencecvesevent_dateevidenceheadlinesectorssourcessourcing_notesummarytechniquestitle

DirtyDecrypt's tie to CVE-2026-31635 is an attribution by Will Dormann and the NVD record's link to the proof of concept, not the discoverers' own: the kernel's description of CVE-2026-31635 is an inverted length check that lets oversized RESPONSE authenticators reach rxgk_decrypt_skb() and hit a BUG_ON, scored CVSS 3.1 7.5 as a network denial of service (ENISA EUVD-2026-25528; BleepingComputer, 2026-05-18; The Hacker News, 2026-05-19). The affected range is kernels from 6.16 before 6.18.23 or 6.19.13, with 7.0 also fixed, and older kernels fall outside it; no source says Debian Stable, RHEL or Ubuntu LTS are unaffected (ENISA EUVD-2026-25528). V12's proof of concept was already public when BleepingComputer reported it on 2026-05-18 (BleepingComputer, 2026-05-18). The interim mitigation BleepingComputer gives blocks the esp4, esp6 and rxrpc modules and breaks IPsec VPNs and AFS (BleepingComputer, 2026-05-18).

vulnerability20 May 05:00Zmulti-sourceOpen finding →
ROUTINECVE-2026-45584updatedNATOA2

CVE-2026-45584: Microsoft Defender heap buffer overflow gives code execution when Defender scans a malicious quarantined file

Microsoft disclosed CVE-2026-45584 on 2026-05-19: a heap-based buffer overflow (CWE-122) in Microsoft Defender allows an unauthorized attacker to execute code over a network, scored CVSS 8.1 with high attack complexity (MSRC, 2026-05-19). Microsoft says a remote, unauthenticated attacker would have to entice a local user to take multiple actions that result in Defender scanning a malicious file that has been quarantined (MSRC, 2026-05-19). MSRC listed the flaw as neither exploited nor publicly disclosed at release (MSRC, 2026-05-19), and The Hacker News found no evidence of in-the-wild exploitation (The Hacker News, 2026-05-21). The last affected build is Malware Protection Engine version 1.1.26030.3008 and the first fixed build is Engine version 1.1.26040.8 (MSRC, 2026-05-19), the same engine release that fixes the exploited CVE-2026-41091 (The Hacker News, 2026-05-21).

Heap-based buffer overflow in Microsoft Defender allows an unauthorized attacker to execute code over a network.

Successful exploitation of this vulnerability would require a remote, unauthenticated attacker to entice a local user to take multiple actions that results in Defender scanning a malicious file that has been quarantined.

MSRC CVE-2026-45584 2026-05-19

There is no evidence the vulnerability has been exploited in the wild.

The Hacker News 2026-05-21
Correctionrun 2026-10-21T1148Z-auditaffected_productsbodyclassificationcvesevidenceheadlinesourcessourcing_notesummarytechniquestitlepriority

Exploitation is not zero-click: Microsoft says a remote, unauthenticated attacker would have to entice a local user to take multiple actions that result in Defender scanning a malicious file that has been quarantined, and rates the attack complexity high (MSRC, 2026-05-19). The advisory does not say the attacker's code runs in the Defender process context. Malware Protection Engine version 1.1.26030.3008 and earlier are affected and Engine version 1.1.26040.8 fixes the flaw (MSRC, 2026-05-19).

vulnerability20 May 05:00Zsingle-sourceOpen finding →

03Research, reports & policy1 item

NOTABLEupdatedNATOB2

Cisco Talos: the "demo.pdb" BadIIS variant is commodity IIS-module malware for SEO fraud, likely sold among Chinese-speaking crime groups and built by the author "lwxat"

Since 2024 Talos has investigated attacks, mainly in the Asia-Pacific region and a few in South Africa, Europe and North America, that use a BadIIS variant marked by "demo.pdb" strings, which it assesses with moderate confidence is commodity malware sold privately or shared among several Chinese-speaking cybercrime groups under a malware-as-a-service model (Cisco Talos, 2026-05-19). The BadIIS variants of the DragonRank and UAT-8099 groups Talos documented before relied on hardcoded C2 and statically compiled payloads, whereas this variant is configured with a recovered builder that injects parameters for JavaScript redirection of visitors to illicit sites, reverse proxying of illicit content to search-engine crawlers, site-wide content hijacking and backlink injection, and that encodes the C2 address with single-byte XOR (Cisco Talos, 2026-05-19). Service-based installers from the same author impersonate Windows services such as Winlogin, FaxService or AudiosService, authenticate to a C2 server, download further payloads, hide their command lines with custom Base64 and register the BadIIS DLL in the globalModules and modules sections of the IIS configuration, and the latest copies the DLL to a hidden backup location so it is restored every time IIS restarts (Cisco Talos, 2026-05-19). Talos ties development from at least September 2021 to January 2026, including a build to bypass Norton, to the alias "lwxat" with moderate-to-high confidence, from the builder's authentication string, a configuration parameter and a custom user agent seen in samples (Cisco Talos, 2026-05-19). Talos does not describe how the attackers first gain access to the IIS servers.

Based on the recovered builder and PDB strings, Talos assesses with moderate confidence that this "demo.pdb" variant is commodity malware, likely sold privately or shared within underground markets.

This ensures that the malicious BadIIS is automatically restored and launched every time the compromised IIS server is restarted.

Talos assesses with moderate-to-high confidence that the string "lwxat" is the author's alias or handle.

Cisco Talos 2026-05-19
Correctionrun 2026-10-21T1148Z-auditaffected_productsbodyclassificationentitiesevent_dateevidenceheadlinesourcessourcing_notesummarytechniquestitle

The variant is not an ISAPI filter or extension: its installers register the malicious DLL in the globalModules and modules sections of the IIS configuration at server level, and the latest one keeps a hidden backup copy that restores it every time IIS restarts, so the place to look is the server module list and the backup location (Cisco Talos, 2026-05-19). Talos does not place it in DragonRank or UAT-8099 activity; it contrasts it with those groups' BadIIS variants, which used hardcoded C2 and static payloads, assesses it as commodity malware likely sold or shared among several Chinese-speaking cybercrime groups, and says the attacks are difficult to attribute to a single actor (Cisco Talos, 2026-05-19). The "lwxat" alias rests on the builder's authentication string, a configuration parameter and a custom user agent, and Talos does not say how attackers first reach the IIS servers (Cisco Talos, 2026-05-19).

research20 May 05:00Zsingle-sourceOpen finding →
Sources: Cisco Talos

04Updates to prior coverage2 items

HIGHCVE-2026-44128 +6updatedNATOA2

SEPPmail Secure E-Mail Gateway: unauthenticated code execution and five more flaws in NCSC-CH's advisory (CVE-2026-44128 et al., CVSS up to 9.3), fixed by 15.0.4

First published 2026-05-09 · open finding →

Updaterun 2026-05-20-a0f7b07factionscvesregionssectorssourcestagsbody

UPDATE (originally covered 2026-05-09 deep dive on CVE-2026-44128 cluster): InfoGuard Labs (the Baar-based Swiss security firm that performed the original SEPPmail review) published its full technical write-up on 2026-05-18.

InfoGuard Labs, which found CVE-2026-2743 and the GINA v2 flaws, published its technical write-up on 2026-05-18 (InfoGuard Labs, 2026-05-18). CVE-2026-2743 (CVSS 10.0 (The Hacker News, 2026-05-19)) is a pre-authentication path traversal in the Large File Transfer (LFT) component's /v1/file.app endpoint that writes files as the nobody user; because nobody can write /etc/syslog.conf, and newsyslog runs every 15 minutes and makes syslogd reload its configuration, the write becomes command execution (InfoGuard Labs, 2026-05-18). Only instances with LFT, which InfoGuard says most customers enable, are affected; /v1/file.app returning HTTP 404 means not affected (InfoGuard Labs, 2026-05-18). CVE-2026-2743 is not new: SEPPmail's 15.0.3 release notes already close it (SEPPmail release notes). The Hacker News reported the write-up (The Hacker News, 2026-05-19).

Contradiction: SEPPmail's release notes place the CVE-2026-2743 fix in 15.0.3 (SEPPmail release notes), while InfoGuard's table lists 15.0.4 (InfoGuard Labs, 2026-05-18); 15.0.4 or later covers both readings. InfoGuard also places the CVE-2026-44126 fix in 15.0.3 (InfoGuard Labs, 2026-05-18), while NCSC-CH's advisory says the vulnerabilities were fixed in 15.0.4 (NCSC-CH, 2026-05-08), so 15.0.4 is the safe floor for that flaw too.

NOTABLECVE-2026-45585 +1updatedNATOA1

Windows BitLocker "YellowKey" (CVE-2026-45585) and CTFMON "GreenPlasma" (CVE-2026-45586) zero-days: public PoCs in May, fixed in the June 2026 Windows updates

First published 2026-05-15 · open finding →

Updaterun 2026-05-20-a0f7b07factionscvessourcesbody

UPDATE (originally covered 2026-05-15): Microsoft formally assigned CVE-2026-45585 to the BitLocker / WinRE bypass disclosed by "Nightmare Eclipse" on 2026-05-12 and confirmed there is still no security update.

Microsoft formally assigned CVE-2026-45585 to the BitLocker / WinRE bypass disclosed by "Nightmare Eclipse" on 2026-05-12 and confirmed there is still no security update. The MSRC update guide entry, published 2026-05-19, classifies it as CWE-77 (command injection in BitLocker / Windows Recovery Environment), CVSS 6.8 (AV:P/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H), with exploit-code maturity rated E:P (proof-of-concept) and remediation level RL:W (workaround only).

Microsoft's interim mitigation, which it recommends that customers "consider implementing" if they are concerned their devices and data are at risk of being compromised or stolen, is per-device work: mount the WinRE image, remove the autofstx.exe entry from the BootExecute registry value inside the WinRE image, commit the image, then re-establish BitLocker trust for WinRE (MSRC CVE-2026-45585). The MSRC FAQ states: "A successful attacker could bypass the BitLocker Device Encryption feature on the system storage device. An attacker with physical access to the target could exploit this vulnerability to gain access to encrypted data."

A pre-boot PIN avoids depending on the WinRE edit: the public exploit relies on TPM-only automatic unlock and does not work with TPM+PIN (BleepingComputer, 2026-05-13), and NCSC-CH advises that BIOS passwords and a BitLocker PIN mitigate the current public PoC (NCSC-CH Security Hub #12574, 2026-05-14).

05Deep dive1 item

HIGHupdatedNATOB2

Storm-2949: SSPR abuse of four privileged users turned into a data-theft sweep across Microsoft 365, Key Vault, SQL, Storage and Azure VMs

Microsoft Threat Intelligence published its account of an intrusion by Storm-2949 on 2026-05-18: a campaign whose singular focus was to exfiltrate as much sensitive data as possible from one target organization's high-value assets, across Microsoft 365, file hosting and the Azure production environment (Microsoft, 2026-05-18). The actor did not rely on traditional malware but on legitimate cloud and Azure management features that blend into administrative behavior (Microsoft, 2026-05-18). Storm is Microsoft's temporary designation for activity not yet classified, and Microsoft saw no follow-on extortion and no link to ShinyHunters (BleepingComputer, 2026-05-19).

Identity takeover. Microsoft assesses with high confidence that Storm-2949 abused Self-Service Password Reset (SSPR): the actor starts a reset on a targeted user's behalf and persuades the user to approve the resulting MFA prompts, likely through an organized and convincing phishing scheme, for example by posing as IT support (Microsoft, 2026-05-18). It then reset the password, removed the user's phone, e-mail and Authenticator methods and enrolled Microsoft Authenticator on its own device, locking the user out; the victims included IT personnel and senior leadership, and the actor repeated the process until it held four accounts (Microsoft, 2026-05-18).

Directory discovery and Microsoft 365 theft. A custom Python script enumerated users and applications through Microsoft Graph to find privileged identities, and an attempt to add credentials to a service principal failed for lack of permissions (Microsoft, 2026-05-18). The actor searched OneDrive and SharePoint for VPN configuration and remote-access documents and in one instance downloaded thousands of files from the OneDrive web interface in a single action (Microsoft, 2026-05-18).

Azure control plane. With privileged custom RBAC roles on several subscriptions, the actor called microsoft.Web/sites/publishxml/action to retrieve the publishing profiles of secondary App Service apps, which often hold FTP, Web Deploy and Kudu credentials; with an Owner role over one Key Vault it changed the vault's access configuration and read dozens of secrets within four minutes; Microsoft believes they held the credentials the actor then used to take over the primary production web app and change its password (Microsoft, 2026-05-18). It opened SQL server firewall rules with microsoft.sql/servers/firewallrules/write and deleted them after taking data, opened storage account network access to its own addresses with microsoft.storage/storageaccounts/write, and retrieved account keys and SAS tokens with microsoft.Storage/storageAccounts/listkeys/action to download blobs over several days with an Azure SDK script (Microsoft, 2026-05-18).

Virtual machines and endpoints. Through the VMAccess extension the actor created a local administrator on a VM, and through Run Command it tried to use the VM's managed-identity token from the Instance Metadata Service against Key Vault, which failed on permissions (Microsoft, 2026-05-18). A Run Command PowerShell script then attempted to weaken Defender Antivirus by disabling several protections, installed ScreenConnect from actor infrastructure under names resembling Windows components, and attempted to clear event logs, command history and temporary files; over ScreenConnect the actor ran host and domain discovery, exfiltrated .pfx certificate files and scanned file shares for password strings (Microsoft, 2026-05-18).

We assess with high confidence that Storm-2949 leveraged a social engineering technique consistent with known abuses of Microsoft’s Self-Service Password Reset (SSPR) process.

Over the span of four minutes, the threat actor successfully manipulated Key Vault access configurations and accessed dozens of secrets within the said Key Vault.

The script then installed the ScreenConnect remote monitoring and management (RMM) tool obtained from threat actor-controlled infrastructure.

Microsoft Threat Intelligence 2026-05-18

Tanmay Ganacharya, VP of Microsoft Security Research, told BleepingComputer that current Storm-2949 activity does not indicate a connection to ShinyHunters or an affiliate of the group.

BleepingComputer 2026-05-19
Correctionrun 2026-10-21T1148Z-auditbodyclassificationevent_dateevidenceheadlineregionssectorssourcessourcing_notesummarytagstechniquestitleverification

Storm-2949 is not described as financially motivated: Microsoft's stated objective for the campaign is mass data exfiltration (Microsoft, 2026-05-18), Storm is its temporary designation for unclassified activity, and it observed no extortion and sees no link to ShinyHunters (BleepingComputer, 2026-05-19). The intrusion did not stay free of binaries: on VMs the actor installed ScreenConnect, tried to weaken Defender Antivirus and cleared event logs (Microsoft, 2026-05-18). SSPR abuse took over four privileged accounts, not one, and Microsoft hedges the social engineering as consistent with SSPR abuse and likely a phishing scheme, giving an IT-support call only as an example (Microsoft, 2026-05-18).

At Key Vault the actor changed the vault's access configuration and read dozens of secrets, connection strings and identity credentials among them (Microsoft, 2026-05-18); the Key Vault operation names given earlier for hunting are not in Microsoft's write-up, and Entra ID records SSPR activity in the Self-Service Password Management audit category (Microsoft Learn). Microsoft names no victim sector or country.

threat20 May 05:00Zsingle-sourceOpen finding →

06Action items2 items

Verification & coverage notes1 run

2026-05-20-a0f7b07f · Claude Opus 4.7 · 16 entries published

  • Items dropped, out-of-window primary sources (PD-7 recency enforcement).
    • CVE-2026-44277 / CVE-2026-26083, Fortinet FortiAuthenticator + FortiSandbox unauthenticated RCE (Fortinet PSIRT FG-IR-26-128 / FG-IR-26-136 dated 2026-05-12; NCSC.ch CSH post 12569 dated 2026-05-13). Primary advisories sit ~8 days outside the 36-h window with no fresh exploitation evidence in-window; defer to next coverage if Fortinet exploitation evidence emerges. Both CVEs and the patched-versions reference remain in the § 2 CVE summary table as context only.
    • CVE-2026-45185, Exim "Dead.Letter" UAF in BDAT/GnuTLS (oss-security disclosure 2026-05-12, XBOW blog 2026-05-12, NCSC.NL NCSC-2026-0163 published 2026-05-15). Primary source 8 days outside window; NCSC.NL corroboration is 5 days outside. No in-window exploitation evidence to drive an UPDATE; defer.
    • Ofcom UK Online Safety Act hash-matching final decision (The Record, 2026-05-19). UK-domestic regulatory action affecting platform operators; single-source, in-window, but does not clear PD-11 inclusion bar (no Swiss/EU public-sector SOC operational delta in the next 1–7 days). Deferred from § 1 / § 6.
  • Single-source items (PD-5 marked).
    • Huawei VRP / POST Luxembourg zero-day in § 1, Recorded Future News, 2026-05-19. HIGH-reliability investigative journalism with named institutional sources (POST Luxembourg head of communications Paul Rausch, Luxembourg High Commission for National Protection spokesperson Anne Jung). Huawei did not respond to questions; no second independent outlet had broken the story at brief composition. Marked [SINGLE-SOURCE] per policy.
    • CVE-2026-45584 (Microsoft Defender network RCE in § 2) MSRC only; national-CERT carve-out applies (Microsoft is the disclosing vendor; primary).
    • CVE-2026-45585 (YellowKey) UPDATE in § 4, MSRC only for the formal CVE assignment + mitigation publication. National-CERT-style carve-out for vendor-as-primary applies.
    • CVE-2026-41091 in § 2, MSRC only for the active-exploitation confirmation; vendor-as-primary carve-out.
    • Cisco Talos BadIIS "demo.pdb" in § 3; Cisco Talos research is itself the primary; included as substantive primary research per PD-12.
    • Fox Tempest disruption (§ 1): effectively single-organisational-source: two of three cited URLs are Microsoft properties (Microsoft Threat Intelligence security blog + Microsoft On the Issues DCU legal blog); The Record corroborates but does not independently verify the technical specifics. Vendor-as-primary carve-out applies; Microsoft is the disclosing party and the action's filer.
  • Reduced-confidence items.
    • TheGentlemen RaaS UPDATE (§ 4), DeXpose aggregator coverage of leak-site listings. The listings themselves are the primary fact; neither victim (VSFS, DEVO-Tech AG) has publicly confirmed the breach. Framed as listing-by-actor-not-victim-confirmation per PD-6.
  • Contradictions surfaced.
    • CVE-2026-44277 CVSS: NCSC.ch CSH post 12569 lists 9.1; BleepingComputer and Fortinet PSIRT carry 9.8. Item deferred this run (out of window); flag for next-run resolution if the item re-enters.
    • vm2 comprehensive-fix version: BSI WID-SEC-2026-1583 lists the fixed version as <3.11.4 (i.e. 3.11.4 is the safe cut-over); The Hacker News reports 3.11.2 as the patch closing the last two CVEs in the 12-CVE cluster (CVE-2026-44008 / 44009). Brief and action items report ≥ 3.11.4 per BSI as the comprehensive, primary-cited cut-over; defenders that already moved to 3.11.2 close 10 of 12 CVEs but BSI flags 3.11.4 as the full fix.
  • Coverage window: Standard daily, 36 h (gap of 24 h to prior brief briefs/2026-05-19.md).
  • Sub-agents. All four returned (S1 / S2 / S3 / S4). S1 wall-clock 681 s; S2 wall-clock 333 s; S3 wall-clock 468 s; S4 wall-clock 587 s. None stalled at the 30-min cap. Models: Claude Sonnet 4.6 across all four research sub-agents; main agent Claude Opus 4.7. Verification: 4 iterations (Opus 4.7 / Sonnet 4.6 / Opus 4.7 / Sonnet 4.6 per the model-rotation table). Iter 1 found F4 + F3×2 + F14 + F12 (5 findings; all remediated). Iter 2 found one residual H3-heading drift on vm2 patch version (3.11.2 → 3.11.4); remediated. Iter 3 found F4 (Fox Tempest "~1,000 accounts" hallucinated count) + F3 (Storm-2949 Key Vault role misattributed Contributor → Owner); both remediated. Iter 4 found F13 (TL;DR over-attributed Nx Console to Mini Shai-Hulud cluster); remediated in-place, published per early-exit rule (NEEDS_FIXES with truth + editorial ≤ 2 AND no F1/F4 → apply remediations, publish). verification_residual_count = 1 reflects iter-4's flagged truth count even though the remediation has been applied; the Ops dashboard cap-breach signal surfaces the residual for after-the-fact review.
  • Coverage gaps: inside-it-ch (Cloudflare Managed Challenge, affected S2, S3, S4; WebSearch fallback yielded no Swiss-only in-window items distinct from NCSC.ch and BSI captures); databreaches-net (Cloudflare-gated, S4 rotation-priority candidate; no usable Wayback snapshot in window); cert-fr-actu (feed stale, returned only Sep–Oct 2025 items); ncsc-uk (no in-window advisory body content); cert-eu (last advisory 2026-006 is from 2026-05-06, outside window); ico-uk (no fresh enforcement in-window); sophos-xops (RSS feed parse failure, rotation-priority); trendmicro-research (not attempted within S3 time budget, rotation-priority); drupal.org (PSA-2026-05-18 page Cloudflare-gated for sub-agents; primary captured via NCSC.ch Security Hub corroboration + SecurityWeek + The Register + BSI WID); fortiguard.fortinet.com (SPA returned empty body; coverage via NCSC.ch CSH + BleepingComputer); drupal.org/security feed (Cloudflare client challenge; sub-agents could not fetch directly, used corroborating sources); sec-edgar-8k (0 Item-1.05 filings in 2026-05-19 / 2026-05-20 window).

Unmatched action items (migrated)

  • Schedule emergency Drupal patch window for today 17:00–21:00 UTC. Freeze unrelated changes, monitor the Drupal Security Advisories feed at 17:00 UTC, apply within hours of release; Drupal Security Team warns exploits expected within hours of disclosure on a 20/25-scored pre-auth flaw. Public-sector portals (Swiss federal / cantonal / municipal, EU-agency, university) are the priority (Drupal PSA-2026-05-18 · NCSC.ch Security Hub 12584).
  • Audit GitHub Actions workflows for any unpinned third-party reference. Enforce full-commit-SHA pinning in repository policy; deploy Harden-Runner or equivalent egress controls. Any pipeline that ran actions-cool/issues-helper@v* between 2026-05-18 and the action's takedown, rotate every secret accessible to that workflow (GitHub PATs, npm, AWS, SSH, deployment keys) (StepSecurity, 2026-05-18).
  • Apply phishing-resistant MFA + SSPR Conditional Access on privileged Azure / M365 roles. Storm-2949's kill chain starts with SSPR voice-phishing of IT and senior-leadership accounts; FIDO2 / certificate-based MFA on Owner, Key Vault Contributor, Storage Account Contributor, SQL Server Contributor, App Service Contributor, and Global / Privileged admin roles closes the entry vector. Restrict SSPR registration of new recovery methods to pre-registered devices via Conditional Access. Enable Defender for Cloud on Key Vault, App Service, Storage, and SQL (Microsoft Threat Intelligence, 2026-05-18).
  • Hunt Microsoft-signed binaries with ≤ 72 h certificate validity from Trusted Signing issuers after 2025-05-01. Where the signing CN does not match a known organisational EV identity, treat as suspect; correlate with Teams / AnyDesk / Webex / PuTTY installer process trees spawning cmd.exe / powershell.exe / rundll32 / regsvr32 without Microsoft installer ancestry (Sysmon EID 1 with parent-image filter), Vanilla Tempest / Rhysida and Oyster/Broomstick (Microsoft Threat Intelligence, 2026-05-19).
  • SBOM-scan Node.js dependencies for vm2 < 3.11.4 across CI/CD runners, automation platforms (n8n etc.), and AI-agent stacks. Upgrade to 3.11.4 per BSI WID-SEC-2026-1583 as the comprehensive fix; 3.11.2 closes 10 of 12 CVEs but BSI flags 3.11.4 as the safe cut-over. No configuration workaround. Multiple CVSS 10.0 sandbox-escape CVEs with public PoC; AI agents that pass model-generated code through vm2 are direct host-RCE vectors (BSI WID-SEC-2026-1583).
  • Huawei VRP enterprise-router operators (telco / large enterprise): escalate the Luxembourg outage advisory with Huawei account team and demand status on whether the underlying flaw is patched and applicable to your fleet. No public CVE 10 months on (The Record, 2026-05-19).

Migrated from briefs/2026-05-20.md (v2).