Sunday, October 4, 2026|Issue 16|6 min read|Join the discussion
The trusted relationship is the new boundary
The most important security boundary today isn't necessarily the vulnerable product. It's the trusted relationship behind it. Denmark's national registry was exposed through a company's legitimate access, NetScaler appliances are under attack again through a flaw last week's patches didn't cover, and automated vulnerability reports have become noisy enough that Google has paused part of a bug-bounty intake.
The short version
- Citrix confirms another exploited NetScaler zero-day, CVE-2026-88779, in SAML deployments; confirmed impact is denial of service. KEV deadline October 7.
- Denmark's Central Person Register: names, addresses and CPR numbers of about 8.8 million people, taken through a company's lawful access.
- Rejetto HFS CVE-2026-61500, found with Anthropic's Mythos model, is now being probed in the wild. Fixed in 3.2.1.
- Google has stopped taking product-vulnerability reports through its OSS VRP after a flood of mostly invalid automated submissions.
- FortiGuard's ClingSTUN Linux backdoor hides its connectivity in public STUN traffic and spreads itself with built-in exploits.
Issue at a glance
2 critical1 high1 notable
- criticalVulnerabilityNetScaler gets hit again, through a flaw last week's patch didn't fix
- criticalBreachDenmark's population register exposed through a company's legitimate access
- highAIAn AI-found Rejetto HFS flaw is now being probed
- notableSupply chainGoogle pauses part of its open-source bug bounty over automated noise
Patch clock
| CVE | Product | What it is | Severity |
|---|---|---|---|
| CVE-2026-88779 | Citrix NetScaler ADC and Gateway (SAML SP or IdP) | Memory overflow; Citrix-confirmed impact is denial of service. RCE not confirmed. Due October 7 | High (CVSS v4 8.7), KEV, exploited |
Also on the vulnerability desk
- Rejetto HFS CVE-2026-61500 (CVSS v4 9.3): weak PRNG lets an unauthenticated attacker recover the session signing key, forge an admin session and reach code execution via server_code. Fixed in 3.2.1 (July). VulnCheck honeypots have seen small-scale probing; it is not on KEV.
Horizon3.ai|GitHub Advisory GHSA-xxrm-3f86-v97j|BleepingComputer
- FortiMail CVE-2026-104286's federal deadline passed October 4. Organizations that were exposed should investigate whether compromise came before remediation.
Citrix security bulletin CTX697174|BleepingComputer|GovInfoSecurity|CISA KEV catalog
Breaking
Citrix / NetScaler / KEV
NetScaler gets hit again, through a flaw last week's patch didn't fix
Citrix disclosed CVE-2026-88779, a memory overflow in NetScaler ADC and Gateway configured as a SAML service provider or identity provider. Citrix has observed targeted attacks causing denial of service and says it has not identified any impact on customer-data integrity. The fixes for CVE-2026-88771 and CVE-2026-88772 do not cover it. Researchers are investigating whether it can go further, but remote code execution is not confirmed. CISA added it to KEV on October 4 with an October 7 deadline. Fixed builds: 14.1-73.41, 13.1-64.28, 14.1-73.41 FIPS and 13.1-37.282 FIPS/NDcPP, and later.
Why it matters. A separate zero-day hitting appliances that took last week's emergency patch breaks the assumption that the NetScaler emergency is over.
What to do. Check for samlAction or samlIdPProfile in your configuration, install the newest builds, preserve evidence and investigate unexpected reboots and unusual authentication traffic. Last week's patch doesn't mean this week's appliance is clean.
CVE-2026-88779
Citrix security bulletin CTX697174|BleepingComputer|GovInfoSecurity|CISA KEV catalog
breach / government / third-party access
Denmark's population register exposed through a company's legitimate access
Denmark says unauthorized people obtained the names, addresses and CPR numbers of about 8.8 million registered people, including deceased and emigrated people, by misusing a Danish company's lawful access to search the register. Irregular activity was detected on Friday; the access happened during September. The Danish Data Protection Agency describes a very large number of automated searches aimed at finding valid CPR numbers. People with name and address protection were not affected. The company's access was cut off and police are investigating; no perpetrator has been named.
Why it matters. The central database didn't need a zero-day. The attackers needed an organization it already trusted.
What to do. Inventory third-party identities with query access to sensitive data, set per-partner baselines and rate limits, and detect bulk enumeration through valid credentials. An authorized connection is not the same as legitimate behavior.
Rejetto / AI research / session forgery
An AI-found Rejetto HFS flaw is now being probed
CVE-2026-61500 in Rejetto HFS 3.0.0 through 3.2.0 derives its session-cookie signing key from Math.random() and leaks outputs of the same generator during login. An unauthenticated attacker can reconstruct the generator state, recover the key, forge an admin session and run code through the server_code feature. CVSS v4 9.3; fixed in 3.2.1 on July 13. Horizon3.ai says Anthropic's Mythos model identified the issue and published details and a proof of concept on September 30. VulnCheck has since seen small-scale probing from a single IP.
Why it matters. AI-assisted discovery, then disclosure, patch, public write-up and attacker interest, all on a compressed timeline.
What to do. Confirm HFS 3.2.1 or later and investigate any exposed older instances for forged admin sessions or new server_code endpoints.
CVE-2026-61500
Horizon3.ai|GitHub Advisory GHSA-xxrm-3f86-v97j|BleepingComputer
bug bounty / AI / open source
Google pauses part of its open-source bug bounty over automated noise
As of October 1, Google's OSS VRP no longer accepts product-vulnerability submissions, citing a significant rise in automated submissions, the vast majority not valid. Reports filed before October 1 and supply-chain reports are unaffected; some Cloud repository issues can still go to the Cloud VRP, and the Patch Rewards Program remains. Google commits to an update in Q1 2027. Google's post didn't say whether the submissions were AI-generated.
Why it matters. Automation lowers the cost of finding real vulnerabilities and of producing plausible garbage.
What to do. Disclosure programs need deduplication, reproducibility checks, exploitability validation and researcher context before a human spends time on a report.
Trends
Exposure × Authority × Time × Adversary Speed × Trust
Yesterday we added trust. Denmark shows why: the path was valuable because it was trusted. NetScaler brokers authentication, FortiMail processes email, SD-WAN Manager controls networks, a supplier queries a government database, an AI agent may modify infrastructure. Attackers increasingly need to become something controls already trust. The defensive question shifts from "Is this identity authenticated?" to "Is what this authenticated identity is doing normal, necessary and proportionate to the authority we've given it?" That's Zero Trust moving from authentication into continuous authorization.
Third-party access is part of the identity perimeter
Employees, service accounts, vendors, contractors, partner apps, API consumers and delegated identities all sit inside the perimeter now. Each needs its own behavioral baseline.
Discovery is getting cheaper than validation
Rejetto is the upside of AI-assisted research; Google's OSS VRP pause is the downside. Finding candidate bugs is cheap. Proving they're real still takes people.
Horizon3.ai|GitHub Advisory GHSA-xxrm-3f86-v97j|BleepingComputer|Help Net Security|The Hacker News
Fully patched is a point-in-time claim
NetScaler was patched last week and has a new exploited CVE this week. Security infrastructure, from NetScaler and FortiMail to Cisco SD-WAN, Zimbra and SharePoint, needs continuous reassessment.
Citrix security bulletin CTX697174|BleepingComputer|GovInfoSecurity|CISA KEV catalog
ClingSTUN: exploit, implant, scan, repeat
FortiGuard Labs describes ClingSTUN, a Linux back-connect proxy backdoor that exploits about two dozen known flaws in internet-facing devices, persists through init scripts, uses public STUN servers to maintain NAT bindings so its traffic blends with VoIP and WebRTC, and carries hardcoded exploits to spread. Hunt for STUN traffic from servers and devices that have no reason to use it.
Clearance corner
KEV and BOD 26-04: exploitation beats severity
CVE-2026-88779 scores 8.7, lower than last week's NetScaler flaws, but it entered KEV with a three-day deadline because Citrix confirmed targeted attacks. That is BOD 26-04's model: vulnerability plus exposure plus exploitation plus asset consequence, not CVSS alone.
CISA BOD 26-04|Citrix security bulletin CTX697174|BleepingComputer|GovInfoSecurity|CISA KEV catalog
What we are watching next
NetScaler CVE-2026-88779
Technical analysis showing whether exploitation goes beyond denial of service. Don't assume RCE.
Denmark CPR
The compromised company, initial access and any use or sale of the records.
Rejetto HFS
Movement toward KEV and wider exploitation beyond probing.
AI vulnerability triage
How Google redesigns OSS VRP intake before its Q1 2027 update.
Zammad
Further victims or KEV listing for CVE-2026-102489 and CVE-2026-102490.
Sources
Join the discussion
Loading comments…