Three CVSS 9 or higher Cisco bugs this year, in three different protocols. Every one of them is a parser that answers an unauthenticated network peer before the authentication check runs. Every one of them left an advisory that someone else, who did not find the bug, built a storefront, a scanner, or a scoring verdict on top of.
The Nefarious Plan of Cisco
The Cisco Product Security Incident Response Team (PSIRT) is not aware of any public announcements or malicious use of the vulnerability that is described in this advisory.
That line is Cisco's own, from advisory cisco-sa-ngoam-rce-LWKQ4BU, quoted in CVE-2026-76501: Cisco's NX-OS NGOAM RCE Surfaces With Two "PoCs". Both Are Mastodon Statuses From @Blackpoint_APG. The Second Is The First's Self-Reply.. Cisco wrote it about a pre-authenticated remote code execution scored CVSS 9.8, the top band of the industry's severity scale, across sixty NX-OS release trains from 9.3(3) through 10.6(3s). NX-OS is the operating system on Cisco's Nexus switches. Cisco found the bug themselves. No one in the public record had exploited it. The sentence is boilerplate. It is also, across this year's three Cisco case files on this site, the only sentence a reader can trust without checking it.
The sentence Cisco writes is the one everyone else writes against
I have filed three Cisco cases in six months. Three different products. Three different protocols. Three different years in each release train. Every one is a pre-authenticated network-reachable primitive in a Cisco management or perimeter service, and every one produced the same downstream shape: someone who did not find the bug takes Cisco's advisory text and builds a product on top of it.
The first was a Bitcoin paywall. CVE-2026-20182: The PoC Is the README. The Exploit Is a Bitcoin Address. documents a May 2026 peering-authentication bypass in Cisco Catalyst SD-WAN Controller and Manager, scored CVSS 10. CISA, the US government's cybersecurity agency, added it to its Known Exploited Vulnerabilities Catalog, which obliges federal agencies to patch within a listed deadline. The deadline was three days. The next morning, a twenty-day-old GitHub account published a README naming the service, the port, the handshake, the message type, the internal account, and the post-exploit protocol. The README ended with a TinyURL to a listing on SatoshiDisk, a Bitcoin file-sale platform, selling the exploit for 0.00693 BTC.
The seller does not need a working exploit. The seller needs a deadline.
The scammer reproduced Cisco's advisory. The fix table in the README is copied verbatim from Cisco. The tested-versions footer in the README is a cut-and-paste of two patched releases from Cisco's own table, which the seller did not notice they had done. Cisco wrote the raw material. The scammer put a storefront around it.
The second case was a scanner. CVE-2025-20362: The Scanner Flags Every Cisco WebVPN That Serves Its Logon Page documents a missing-authorization fault in Cisco Secure Firewall ASA and Threat Defense WebVPN, the appliance's SSL VPN portal. Cisco published the advisory in September 2025. Two months later Cisco updated it.
The 5 November 2025 advisory update adds a separate exploitation primitive against the same input-validation gap: the gap can be coerced into crashing the device, escalating the impact past the original CVSS vector.
Two hundred and thirty-three days after the original advisory, a GitHub user shipped a Python scanner that assigns a HIGH RISK verdict. The scanner sends one hundred and thirty requests. The verdict fires after two 200 responses whose bodies contain any of webvpn, cscoe, anyconnect, username, password, config. The Cisco WebVPN logon page returns 200. The Cisco WebVPN logon page contains all of those strings by design, because Cisco's own portal HTML references its own product names. Every Cisco WebVPN appliance that answers the door is HIGH RISK. A freshly patched one, verified by its own operator, is HIGH RISK. The scanner's README lists Cisco's severity, affected products, and the chained CVE from the advisory. Cisco wrote the raw material. The scanner put a verdict around it.
The third is the case at the top of this piece. A vulnerability-scoring pipeline lifted CVE-2026-76501 above its breaking-publication threshold because a corporate threat-intelligence account with three followers posted its weekly summary across two Mastodon statuses.
The pipeline lifted CVE-2026-76501 to score 9 because an account restating Cisco's own no-exploit statement ran into Mastodon's character limit.
The two statuses the pipeline counted as separate proof-of-concept URLs are one thread. The thread quotes Cisco's advisory. The thread's body is a weekly bulletin restating five Cisco CVEs. No exploit code exists anywhere in it. Cisco wrote the raw material. The pipeline put a score around it.
Cisco keeps writing the same parser in three different protocols
Cisco publishes an advisory. In the body of every advisory, Cisco lists the affected release trains. The list is long. For CVE-2026-76501 the list runs sixty trains, from 9.3(3) to 10.6(3s). 9.3(3) shipped in 2019. The code path has been in the product since then. For CVE-2025-20362 the list spans ASA and FTD across both the September 2025 and November 2025 advisory passes, because the November pass added a denial-of-service primitive against the same code. For CVE-2026-20182 the list covers both the 20.12 and 20.15 SD-WAN trains most customers run.
The bug in each case is the same kind of bug. SD-WAN accepts a malformed peering handshake, the packet two SD-WAN boxes use to authenticate to each other, from an unauthenticated peer on UDP/12346. WebVPN accepts an HTTP request whose path normalization routes authenticated-only endpoints to an unauthenticated client. NGOAM accepts a management-plane OAM packet from an unauthenticated sender and overflows the stack. The protocols are different. The product lines are different. The architectural decision is the same: the input handler runs before the authentication handler. The parser says yes before the auth says no.
When the parser breaks, the workaround is the feature's off switch. The NGOAM advisory, as the case file quotes it, states that no workarounds address the vulnerabilities, then notes that no feature ngoam removes the attack vector by taking the feature offline. That is not a workaround. That is a disable. It is the only thing Cisco can offer a customer who cannot deploy the patch fast enough, because the parser is the fucking feature.
Cisco is right about three things
Cisco PSIRT self-disclosed CVE-2026-76501 from internal testing. No outside researcher is named in the advisory. No pending coordinated disclosure is referenced. Cisco found a CVSS 9.8 pre-authenticated remote code execution in their own product and published it before any public actor built anything on top of it. That is the system working.
Cisco updated the CVE-2025-20362 advisory when a second exploitation primitive was found in the same gap. They did not hide the escalation. They did not reissue the CVE under a different number to reset the deadline clock. They published Version 2 against the same identifier and added the new primitive to the public record. That is the system working.
Cisco writes advisories that are specific enough to patch and general enough not to publish a payload. The CVE-2026-20182 record names the service, the port, the handshake class, and the fixed releases. It does not reproduce the packet that triggers the bypass. That is the correct trade. A defender can plan a patch cycle. The scammer in the first case still had to write the Python, which is why the Python was not what they shipped.
The scammer in case one is not Cisco's fault. The scanner in case two is not Cisco's fault. The pipeline in case three is not Cisco's fault. The TinyURL, the SatoshiDisk listing, the HIGH RISK banner, the multi-poc×2+1 bonus: Cisco did not write any of them. The incentives that select for all three belong to the ecosystem. CISA's KEV calendar. GitHub's identifier-matching indexer. Mastodon's five-hundred-character envelope. A scoring pipeline that reads two status IDs as two sources. None of it is Cisco.
The charge is narrower. Cisco keeps shipping the parser. Everyone else shows up every time the advisory lands because the advisory is good enough to copy.
The next case comes from an advisory already on Cisco's site
Open a Cisco advisory the day it lands. Find the affected-release list. Count the trains. If the count is high, the code path is old. Find the workaround section. If the only option is a no feature command that takes the service offline, the parser is the bug and the bug is older than the current train. Those two facts together are the signature of the shape this piece names.
The product surface is large. SD-WAN Manager has the API server and the peering daemons. ASA and FTD have WebVPN, the SSL VPN endpoints, and the management interface. NX-OS has NGOAM, NX-API, MOAM, the SRv6 handlers, and the sibling OAM features listed in cisco-sa-ngoam-rce-LWKQ4BU and its two companion advisories from October 7. IOS XE has its own web UI. Each sits in front of its own authentication path. Each is an advisory away from the next 9.8.
Pick the advisory with both signatures and watch the KEV catalog. The next downstream artifact will arrive within seventy-two hours. It will be a README, a scanner, or a bulletin, in some order this piece has already named.
The next Cisco advisory lands before Q1 2027 closes
The next Cisco entry on KEV, before the end of Q1 2027, will be a pre-authenticated network-reachable primitive in an NX-OS OAM-family feature, in IOS XE's web UI, or in SD-WAN Manager's control-plane API. CVSS 9.0 or higher. Affected release trains will number at least twenty. The workaround section will say no workarounds address the vulnerability and will point to a feature-disable command that takes the service offline. Within seventy-two hours of publication, at least one of the three downstream shapes in this file will appear against the CVE number: a paywall README on a throwaway GitHub account, a scanner that flags the product's own front page, or a pipeline-surfaced multi-poc thread with zero engagement.
You wrote sixty release trains on one advisory and told the customer the workaround is to turn off the feature. Move the authentication in front of the parser.