CVE-2026-129448: Apache Roller XML-RPC deserialization gets CISA KEV deadline — unauthenticated remote code execution in versions before 6.1.5

Apache Roller before 6.1.5 runs unauthenticated XML-RPC deserialization before authentication; it is now on CISA's exploited list with a patch deadline of November 23, 2026.
CoeLab Real-World Impact: 94/100 — Urgent. Editorial score v1.0. Evidence checked 2026-11-21T10:10:56.181830+00:00. CVSS and EPSS are separate measures, not this editorial score.
Rating breakdown: exploitation: 30, reachability: 25, consequences: 24, operational relevance: 15.
Who is affected
Apache Roller, Apache's Java-based blog server, is affected in every release from 4.0.0 through 6.1.4. That is the corrected range published under CVE-2026-129448, which Apache's own CVE Numbering Authority issued on November 17, 2026. Rollers at 6.1.6 or later are fixed.
The version range matters because this is the second identifier for the same underlying flaw. A related record, CVE-2026-82384, had earlier listed only Roller 6.1.5 as affected, which implied that 6.1.5 was a safe stopping point. The new CNA description says explicitly that CVE-2026-82384 described the same vulnerability and that the affected population is much larger: anything from 4.0.0 through 6.1.4. In practice that means the original disclosure was incomplete about which releases contain the bug, and organizations that triaged CVE-2026-82384 based on its narrower list may have concluded they were not exposed.
Roller is not as widely deployed as mainstream web content platforms, but it does run on internet-facing infrastructure: it is used as a blog and blogging backend, sometimes by enterprises, developer teams and universities, and it exposes administrative and service endpoints over HTTP. Any Roller instance that is reachable on the network and remains on 4.0.0 through 6.1.4 needs urgent attention. If the XML-RPC servlet is reachable, that is the exposed surface.
Real-world impact and exploitation evidence
The flaw is CWE-502, deserialization of untrusted data, in Roller's XML-RPC endpoint. The endpoint historically enabled a vendor extension mechanism that lets XML-RPC requests carry typed, serialized objects, and Roller deserialized those request bodies before checking whether the caller had authenticated. That ordering is the whole problem: an unauthenticated network client can send a crafted XML-RPC request and get the server to instantiate attacker-influenced objects from the server's classpath, which Apache describes as leading to arbitrary code execution by an unauthenticated remote attacker.
CISA added CVE-2026-129448 to the Known Exploited Vulnerabilities catalog on November 20, 2026, with a federal remediation due date of November 23, 2026 and a 'yes' flag for forensic triage. Apache's CNA text also states the issue is being actively exploited. At this stage public reporting establishes that exploitation is occurring; it does not yet attribute a campaign, identify a specific attacker, or quantify how widespread exploitation has been. CISA lists known ransomware campaign use as unknown. Treat observed attacker capability as demonstrated, and campaign scope as unmeasured.
The CoeLab Real-World Impact score for this candidate is 94 out of 100, labeled Urgent, based on CoeLab Real-World Impact v1.0. That score is an editorial heuristic, not a CVSS or EPSS metric, and it is not a measurement of any specific deployment. Its components here are: 30 of 30 for exploitation evidence (the KEV listing is the maximum signal), 25 of 25 for reachability (a published vector of network, low complexity, no privileges, no user interaction), 24 of 25 for consequences (a 9.8 CVSS base score from the CNA, doubled to a ceiling of 25 on the consequence scale), and 15 of 20 for operational relevance from recognizing Roller as an infrastructure-adjacent server category. Two limitations matter: the operational-relevance term is a category heuristic that cannot see how many Roller instances actually exist or how many are internet-facing, and EPSS currently sits around 0.012 (roughly the 69th percentile), which is low but reflects exploitation probability in a narrow window rather than confirmed harm. If exploitation continues to spread, EPSS would rise but the KEV-based score would not change.
Attack prerequisites
The attacker needs network access to the XML-RPC servlet on a running Roller instance. The published vector is AV:N/AC:L/PR:N/UI:N under CVSS 3.1: no authentication, no user interaction, no special conditions. The request has to reach the /roller-services/xmlrpc endpoint, so a Roller instance that is exposed over the internet and still has that mapping in place is the worst case; an internal instance reachable only from a restricted network is the next tier of exposure.
Because deserialization happens on the server side of a single request, the practical prerequisites are minimal. No valid account is needed, no configuration option needs to be enabled by the operator, and the CNA notes that even older versions where the XML-RPC feature has been disabled through the administration interface are still deserializing requests before that setting is checked. That detail matters: existing admin-UI toggles are not a defense on vulnerable versions.
What is not publicly documented here is the exact payload class chain or whether specific classpath configurations are required for reliable code execution. That absence does not change the urgency, since CISA's catalog entry depends on observed exploitation, but it does mean public detail about which deployments are easiest to compromise is thinner than the vendor's summary level.
Actions and fixes
The clean fix is to upgrade Apache Roller to 6.1.6 or later. Apache describes 6.1.6 as disabling the XML-RPC extension types so vendor-extension deserialization no longer happens on that path, and the servlet rejects requests when the XML-RPC feature is administratively disabled. That is the durable change, and it lines up with the patch pull request tracked in the project repository.
If an upgrade cannot land before the deadline, Apache's advisory offers two workable mitigations for vulnerable releases: set the web.xml parameter enabledForExtensions to false in WEB-INF/web.xml, or remove the /roller-services/xmlrpc servlet mapping entirely, then restart Roller. Either change stops the extension path that accepts deserialized vendor-extension types on versions below 6.1.6. Doing nothing to configuration is not a defensible option, because the admin-UI XML-RPC toggle alone does not suppress the vulnerable behavior on older versions, per the CNA text.
Operationally, three concrete steps apply this week. First, inventory Roller deployments and record the exact version; anything 4.0.0 through 6.1.4 is in scope, including instances someone previously classified as safe after reviewing only the CVE-2026-82384 record. Second, on internet-facing instances, apply the XML-RPC removal or web.xml change immediately and schedule the 6.1.6+ upgrade; on internal instances, apply the same change on the normal critical-path timeline. Third, because CISA flagged forensic triage, systems that had the XML-RPC endpoint exposed and cannot rule out prior compromise should be reviewed for unexplained deserialization-related request activity rather than simply patched forward. Federal agencies have until November 23, 2026 under BOD 26-04; private operators should treat that date as a sensible external deadline as well, not as a mandate that applies to them.
Uncertainty and what could change
Several facts are pinned down by primary sources and should not drift. The affected range (4.0.0 through 6.1.4), the fixed version (6.1.6+), the unauthenticated deserialization-before-authentication ordering in the XML-RPC endpoint, and the two documented mitigations all come straight from Apache's CNA-recorded description and the vendor mailing-list advisories. The KEV addition date, due date and forensic-triage flag come from CISA's catalog.
What we do not yet know includes which attacker is exploiting this, whether exploitation is commodity or targeted, and whether any specific deployment pattern is more susceptible. CISA's ransomware-campaign-use field is unknown, so the catalog is not currently asserting a ransomware link. EPSS will move over the coming days and may rise if automated scanning increases; that is a probability signal, not confirmation of new attacker behavior.
The relationship to CVE-2026-82384 is a separate ongoing point of interest. Because the newer record is the broader, corrected statement, future tracking should treat CVE-2026-129448 as the canonical identifier for this flaw and should not rely on the earlier, narrower version list. If Apache ships further advisories or the KEV description is amended, that could sharpen or soften details here; nothing in this article should be read as a claim about versions Apache has not confirmed.
Sources and evidence date
Primary vendor evidence comes from Apache's CVE numbering authority description and the two related Apache mailing-list advisory threads, plus the patch pull request in the apache/roller repository. The affected and fixed versions above are taken only from those vendor sources. CISA's Known Exploited Vulnerabilities catalog supplies the KEV addition date of November 20, 2026, the November 23, 2026 due date, and the forensic-triage flag, with the underlying BOD 26-04 context and NVD reference linked from CISA's notes.
Distinct dates worth keeping straight: the flaw was reserved by Apache on November 14, 2026; the CNA record was published November 17, 2026; CISA added it to the exploited-vulnerabilities catalog on November 20, 2026; the record itself was last updated November 21, 2026. This article was written November 21, 2026. No claim in this piece predates the evidence above, and no exploit code, payload chain, or exploitation detail beyond what Apache and CISA have published is being introduced.
Primary-source links
https://lists.apache.org/thread/cc0jck13nb9vvcmzxzwotr6n16mx28z9
https://lists.apache.org/thread.html/lnz4s26tv6hdo03wngtdbpfj4hg4gtds
https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-129448
https://github.com/apache/roller/pull/171
https://www.cve.org/CVERecord?id=CVE-2026-129448
AI-assisted research and drafting, with a separate model editorial review. CoeLab ratings are editorial judgments, not a substitute for checking your own versions and vendor advisories. Methodology: https://thecoelab.com/blog/security/methodology
Related Posts
Your USB-C Dock Is Part of Your Network. Test It That Way.
A trending Ethernet investigation shows why docks belong in your network diagram. Here is how to isolate bottlenecks before replacing routers or adapters.
Valkey’s Sorted Set Upgrade Is a Capacity Planning Opportunity
Valkey’s sorted set redesign promises lower index overhead. Here is how to test the real memory benefit, protect latency, and turn savings into usable capacity.
Jujutsu Makes Git Workflows Flexible. Keep Reviews Rigorous.
Jujutsu offers a different way to organize Git-backed work. Here is how to evaluate stacked changes, recovery, and integrations without weakening reviews.