The 16 November batch holds four records. None sits in the KEV catalog, and neither the vendor reports nor the NVD entries mention exploitation in the wild. I rank them by what each bug gives an attacker. The first is an Apache Shiro authentication bypass that turns a cached login into a standing yes for any password. The second is a use-after-free in FRRouting’s bgpd that a BGP peer can repeat until the daemon gives up.

A Shiro LDAP realm that accepts any password

CVE-2026-129349 is the one I would queue first. Apache rates it critical and scores it 9.1, with no privileges required 1.

Two settings together produce the bug. An application authenticates against LDAP through DefaultLdapRealm or JndiLdapRealm, and someone turns on authenticationCachingEnabled, which Apache says is off by default 2. With caching on, Shiro stores the result of a successful login and skips the LDAP bind on later requests for that user. DefaultLdapRealm also installs AllowAllCredentialsMatcher by default, and that matcher approves any credentials it receives. The cached success skips the bind, so the matcher approves whatever password arrives, and a valid username is enough to log in 2. Every version through 3.0.0 carries the bug, and 3.0.1 fixes it by no longer installing that matcher by default 2.

The exposure question is narrow enough to answer: which Shiro deployments use an LDAP realm, and which set the caching flag. Apache’s alternative to upgrading is to disable authentication caching or to configure a matcher such as SimpleCredentialsMatcher that checks submitted credentials against the cached authentication 2.

Access review is where this one gets awkward. The cache answers before the bind, so a reset changes what LDAP accepts on the next bind and leaves a stored success alone. The application’s cache settings decide how wide that gap gets.

bgpd frees the same chain twice

CVE-2026-129456 lands in FRRouting’s bgpd and scores 7.7 on the NetDEF CVSS 4.0 rating 3. A cache added to speed up attribute interning decides whether an interned attribute set may be reused by comparing the parsed attributes with themselves, so the comparison always succeeds. Interning the parsed attributes for the Adj-RIB-In removes the Tunnel Encapsulation sub-TLVs from them, and the copy bgpd interns for the route then reuses the cached set, which still holds those sub-TLVs, without taking a reference on them. Once bgpd replaces, withdraws or clears that route, it frees the sub-TLV chain and later reads it, decrements a reference count inside it and frees it a second time 4.

In the reporter’s build the second free tripped an assertion in bgpd’s memory accounting, which aborted the daemon and dropped every BGP session on the router. The neighbor can repeat it each time bgpd restarts. The record adds that bgpd writes to the freed heap before freeing it again, so consequences beyond a crash cannot be ruled out 4.

Two conditions decide who is exposed. The session needs soft-reconfiguration inbound, because the broken path sits in Adj-RIB-In handling. The peer also has to send attribute 23, and the record says bgpd processes that attribute from every neighbor, eBGP included, unless an operator discards it, even though RFC 9012 requires filtering it on eBGP by default 4. The affected semver range runs from 10.6.0 up to but not including 10.6.2 3. CISA’s SSVC assessment on the record marks exploitation as none and technical impact as total 3.

The routers belong to the network team, so my output is the question plus the build number. The question is which sessions carry soft-reconfiguration inbound and accept attribute 23 from a peer outside our control, and the check is a read of the neighbor output against the build string. A 10.6.1 router with that flag set on an eBGP session goes to the top of the list I send them, because an UPDATE from an established peer triggers it without a malformed attribute.

Two more Shiro records, one release later

Apache scores CVE-2026-74763 at 6.3 5, and the bypass leaves no failed check to alert on. Shiro did not enforce @RequiresRoles and @RequiresPermissions declared on an interface when an application uses the shiro-guice, shiro-aspectj, shiro-cdi or shiro-jakarta-ee modules. The same gap applied to class and superclass declarations under shiro-guice and shiro-aspectj 2. A method that should have required a role ran for an unauthorized caller. Version 3.0.2 fixes it by giving the annotation closest to the invoked method precedence, and declaring the annotations on each protected implementing method works as an interim step 2.

CVE-2026-129347 is the small one, at 2.3 on Apache’s rating 6. DefaultPasswordService and SimpleCredentialsMatcher could treat an empty stored credential as a match for an empty submitted password, in versions through 3.0.1, and 3.0.2 fixes it. Apache’s advice until then is to stop storing empty credentials and to reject empty passwords in the application 2.

The Shiro upgrade line splits. Release 3.0.1 closes the authentication bypass and leaves both lower records open, and 3.0.2 closes them. Apache also put Shiro v2 out of support on 29 June 2026, so a 2.x deployment crosses a major version to reach either release 2. The upgrade target is 3.0.2 for anything that can take it, and 3.0.1 is a floor rather than the target.


  1. National Vulnerability Database. (2026, November 15). CVE-2026-129349 detail. National Institute of Standards and Technology. https://nvd.nist.gov/vuln/detail/CVE-2026-129349 ↩︎

  2. Apache Software Foundation. (2026, November). Apache Shiro security reports. https://shiro.apache.org/security-reports.html ↩︎ ↩︎ ↩︎ ↩︎ ↩︎ ↩︎ ↩︎ ↩︎

  3. CVE Program. (2026, November 16). CVE-2026-129456 record. https://www.cve.org/CVERecord?id=CVE-2026-129456 ↩︎ ↩︎ ↩︎

  4. National Vulnerability Database. (2026, November 14). CVE-2026-129456 detail. National Institute of Standards and Technology. https://nvd.nist.gov/vuln/detail/CVE-2026-129456 ↩︎ ↩︎ ↩︎

  5. National Vulnerability Database. (2026, November 16). CVE-2026-74763 detail. National Institute of Standards and Technology. https://nvd.nist.gov/vuln/detail/CVE-2026-74763 ↩︎

  6. National Vulnerability Database. (2026, November 16). CVE-2026-129347 detail. National Institute of Standards and Technology. https://nvd.nist.gov/vuln/detail/CVE-2026-129347 ↩︎