| CVE |
Vendors |
Products |
Updated |
CVSS v3.1 |
| Deserialization of Untrusted Data vulnerability in Apache Camel Camel-Jgroups component.
This issue affects Apache Camel: from 4.0.0 before 4.18.5, from 4.19.0 before 4.22.1.
JGroupsEndpoint.createExchange(Message), which CamelJGroupsReceiver calls for every message the jgroups consumer receives before the exchange is routed, set the exchange body from org.jgroups.Message.getObject(). For an ObjectMessage, which is the message type the Camel jgroups producer sends, JGroups Java-deserializes the payload through an ObjectInputStream on its own receive thread as the message arrives, and getObject() returns the resulting object. No java.io.ObjectInputFilter was applied on that path unless the deployment had set a JVM-wide serialization filter: camel-jgroups installed none, and the JGroups versions shipped with the affected releases apply none of their own, so any serializable class on the application classpath could be instantiated and its deserialization callbacks run. Unless a channel or the channelProperties option is configured, the consumer creates its channel with new JChannel(), the stock JGroups protocol stack, which multicasts over UDP without the AUTH protocol and without encryption, so any host able to reach the cluster's multicast group or bind port can join the cluster and send messages to the consumer without credentials. Such a host could send a crafted serialized object that leads to remote code execution in the Camel process when a usable deserialization gadget chain is present on the classpath, and no non-default option was required on the consumer. Because JGroups materializes the object before Camel's receiver runs, a check performed by Camel can only inspect the deserialized result and cannot stop code that runs during deserialization, so closing the issue also requires a control that applies before the payload is read, as described in the mitigation. Exploitation requires network access to the cluster's multicast group or bind port, which with the default UDP multicast transport means a host on an adjacent network, and a usable gadget chain on the classpath, which is why the severity is assessed as high rather than critical.
Users are recommended to upgrade to version 4.23.0, which fixes the issue. If users are on the 4.22.x releases stream, then they are suggested to upgrade to 4.22.1. If users are on the 4.18.x releases stream, then they are suggested to upgrade to 4.18.5. The 4.14.x releases stream and earlier, and the 4.15.x to 4.17.x and 4.19.x to 4.21.x release lines, reached end of life before this issue was resolved and will not receive a fix, so deployments on those lines must move to a supported release. |
| Improper Limitation of a Pathname to a Restricted Directory ('Path Traversal') vulnerability in Apache Camel FTP component.
This issue affects Apache Camel: from 4.0.0 before 4.18.5, from 4.19.0 before 4.22.1.
The Camel remote-file consumers poll a directory on a remote server and build the absolute remote path of each file by concatenating the configured directory with the file name reported in the server's directory listing. That concatenation (FtpUtils.absoluteFilePath and GenericFileConsumer) applied no lexical normalization and no check that the result stayed inside the directory being polled, and the resulting path became the operand for every subsequent remote operation on the file: retrieval (FtpOperations.retrieveFile / SftpOperations, reached from GenericFileConsumer where the retrieve name is taken verbatim from GenericFile.getAbsoluteFilePath), deletion when delete=true (GenericFileDeleteProcessStrategy), and renaming when move, preMove, moveFailed or readLock=rename is configured (GenericFileProcessStrategySupport and GenericFileRenameExclusiveReadLockStrategy). Neither isValidFile nor isMatched performed any containment check. A listing entry that is not a single path segment therefore resolved to a path outside the configured polling directory, causing Camel to read, delete or rename files outside it using the endpoint's own credentials. A listing name is not route-controlled data. It is supplied by the remote server, and the FTP control channel carries no integrity protection at all, so an on-path attacker can rewrite a LIST response without holding any credentials; a compromised, malicious or multi-tenant server, or a server backed by a virtual filesystem, can emit such a name directly. Retrieval is reachable in a default consumer configuration, since the remote consumers download every listed file; deletion and renaming additionally require one of the disposition options above, all of which are common. Escape from the directory Camel was configured to poll is unconditional, while escape from the server's own account boundary additionally depends on whether that server confines the account. The jailStartingDirectory option, which is documented as restricting Camel to the starting directory and is enabled by default, governed producer writes and, since the localWorkDirectory containment work, consumer downloads to the local work directory - but it did not constrain this remote operand, so its name and description claimed a boundary that was not enforced here. The same entry point had already been recognised on the local side: the localWorkDirectory download path was constrained so that a remote file name containing parent-directory segments could not resolve outside the local work directory. The remote operand was the remaining half of that surface. This artifact provides the ftp, ftps and sftp schemes, all of which are affected. The reporter demonstrated the issue end to end against an ftp endpoint, driving a Camel route with a malicious server whose LIST response named a file outside the polled directory.
Upgrade to Apache Camel 4.23.0, which fixes the issue. Users on the 4.22.x releases stream should upgrade to 4.22.1. Users on the 4.18.x LTS releases stream should upgrade to 4.18.5. The 4.14.x stream reached end of life and does not receive this fix; users on it should move to a supported releases stream. |
| Improper Handling of Case Sensitivity vulnerability in Apache Camel Netty HTTP component.
This issue affects Apache Camel: from 4.0.0 before 4.18.5, from 4.19.0 before 4.22.1.
The camel-netty-http security constraint compared the request path against the configured context path using a case-sensitive match, while the routing that dispatches the request to the endpoint did not apply the same rule. The two comparisons therefore disagreed for any path whose case differed from the configured value. An external caller could alter the case of a character in a path segment so that the security constraint no longer recognised the path as protected, while the request still dispatched to the same endpoint. The result is that the constraint is skipped and the protected route is reached without satisfying it. No credential, prior access or special configuration is required beyond being able to send a request to the consumer.
Upgrade to Apache Camel 4.23.0, which fixes the issue. Users on the 4.18.x LTS releases stream should upgrade to 4.18.5. Users on the 4.22.x LTS releases stream should upgrade to 4.22.1. The constraint now matches the request path with the same case rules the dispatch uses, so the two can no longer disagree. For deployments that cannot upgrade immediately, place the authorization decision in front of the Camel consumer, for example in a reverse proxy that normalises the request path before forwarding. |
| Server-Side Request Forgery (SSRF) vulnerability in Apache Camel AS2 component.
This issue affects Apache Camel: from 4.0.0 before 4.18.5, from 4.19.0 before 4.22.1.
In the AS2 protocol the sender of a message may request that the Message Disposition Notification be returned asynchronously to an address the sender supplies. camel-as2 took that address from the inbound message and used it as the destination for the outgoing receipt without constraining it to anything the deployment had configured. A remote partner therefore chose the host the Camel instance would connect to when delivering the MDN. That gives server-side request forgery from the position of the AS2 receiver, including reaching hosts that are only routable from inside the deployment's network, and where the outbound connection carries credentials configured for MDN delivery those credentials are presented to the attacker-chosen destination.
Upgrade to Apache Camel 4.23.0, which fixes the issue. Users on the 4.18.x LTS releases stream should upgrade to 4.18.5. Users on the 4.22.x LTS releases stream should upgrade to 4.22.1. The asynchronous MDN destination is now constrained rather than taken verbatim from the inbound message. For deployments that cannot upgrade immediately, restrict outbound connectivity from the AS2 receiver to the partner endpoints it legitimately delivers receipts to, and do not attach credentials to MDN delivery that are valid for any host other than those partners. |
| Permissive Cross-domain Policy with Untrusted Domains vulnerability in Apache Camel Platform HTTP Vertx component.
This issue affects Apache Camel: from 4.0.0 before 4.18.5, from 4.19.0 before 4.22.1.
createCorsHandler() set Access-Control-Allow-Credentials: true outside the origin check, so the header went out on every response to a request carrying an Origin header, including responses to origins the handler had just decided not to allow. When camel.server.cors.origins is unset, the handler treats every origin as allowed and echoes the caller's own Origin back as Access-Control-Allow-Origin. Together those produce the credentialed any-origin configuration that the fetch specification refuses to express as a literal wildcard, which is precisely why reflecting the origin is the usual way around that rule. A page on any origin could therefore make credentialed cross-origin requests to the Camel endpoint and read the responses.
Upgrade to Apache Camel 4.23.0, which fixes the issue. Users on the 4.18.x LTS releases stream should upgrade to 4.18.5. Users on the 4.22.x LTS releases stream should upgrade to 4.22.1. Access-Control-Allow-Credentials is now sent only when the request origin matched an origin the operator actually configured. With no origin list the origin is still reflected, as before, but credentials are not granted, and Vary: Origin is now set whenever the origin is reflected so that a shared cache cannot serve one origin's response to another. For deployments that cannot upgrade immediately, set camel.server.cors.origins to the explicit list of origins that need access rather than leaving it unset. |
| This issue affects Apache SIS from version 0.7 through 1.6 inclusive. When a data file contains links to other files, Apache SIS opened these files regardless their paths. It was possible, for example, to specify a path to a local file. However, these files were always expected to be in some specific formats and the reading stops at the first parsing error, which makes this vulnerability unlikely to leak sensitive information. The impacted formats are:
* Geographic Markup Language (GML) with `xlink:href` attributes for other files in GML format.
* Well Known Text (WKT) with parameter elements for datum shift grids in NADCON, NTv2 or RGF93 formats.
* Landsat scene with paths to files for each band in the GeoTIFF format.
Users are recommended to upgrade to version 1.7, which fixes the issue by restricting the paths to the same directory as the file containing the references to other files, or to a predetermined directory for datum shift grids. |
| Deserialization of Untrusted Data vulnerability in Apache Camel.
This issue affects Apache Camel: from 4.0.0 before 4.18.5, from 4.19.0 before 4.22.1.
The camel-hazelcast component applies a default Hazelcast JavaSerializationFilterConfig to the instances it creates from its own configuration, a protection introduced by CVE-2026-43865. That change did not reach two of the configuration paths Camel builds itself, so those paths continued to create Hazelcast instances whose Java deserialization is unrestricted. ReplicatedHazelcastAggregationRepository extends HazelcastAggregationRepository but overrides doStart() without calling super.doStart(), keeping its own copy of the bootstrap block, and therefore never received the filter: a route using it without supplying a hazelcastInstance starts a Hazelcast member that deserializes values read from the backing ReplicatedMap with no class restriction. HazelcastDefaultComponent#getOrCreateHzClientInstance() builds its own ClientConfig when hazelcastMode=client is selected and neither a referenced ClientConfig nor hazelcastConfigUri is supplied, and likewise applied no filter, so client-mode endpoints behaved differently from node-mode endpoints for an otherwise identical configuration. In both cases an attacker able to publish into the Hazelcast cluster that the affected instance is attached to can cause a serialized Java object of their choosing to be deserialized inside Hazelcast's own serialization layer, before Camel processes the exchange, resulting in remote code execution. Hazelcast clusters have neither authentication nor TLS enabled by default. Selecting client mode is a topology choice rather than a security option, so neither path required the deployment to opt in to unsafe behaviour.
Users are recommended to upgrade to version 4.23.0, which fixes the issue. If users are on the 4.22.x releases stream, then they are suggested to upgrade to 4.22.1. If users are on the 4.18.x releases stream, then they are suggested to upgrade to 4.18.5. The 4.14.x releases stream and earlier reached end of life before this issue was resolved and will not receive a fix, so deployments on those lines must move to a supported release. For deployments that cannot upgrade immediately, supply an explicit Hazelcast configuration carrying a JavaSerializationFilterConfig, or inject a pre-built HazelcastInstance, since a user-supplied configuration is never overwritten by Camel. Alternatively set the JVM-wide filter -Djdk.serialFilter=!java.net.;java.;javax.;org.apache.camel.;!* and enable Hazelcast cluster authentication and TLS to restrict who can reach the cluster. |
| **UNSUPPORTED WHEN ASSIGNED** Unbounded allocation in XML parsing in Apache Vysper; all versions, may lead to an out of memory condition.
This issue affects all versions of Apache Vysper. Apache Vysper is an XMPP (Jabber) server library.
As this project is retired, we do not plan to release a version that fixes this issue. Users are recommended to find an alternative. |
| **UNSUPPORTED WHEN ASSIGNED** Authorization bypass in Apache Vysper, all versions.
Apache Vysper is an XMPP (Jabber) server library. Its XEP-0077 (In-Band Registration) implementation is missing an authentication check, allowing any authenticated user to compromise other accounts.
As this project is retired, we do not plan to release a version that fixes this issue. Users are recommended to find an alternative. |
| Improper Neutralization of Special Elements used in an Expression Language Statement ('Expression Language Injection') vulnerability in Apache Camel AWS2-s3 component.
This issue affects Apache Camel: from 4.8.0 before 4.18.5, from 4.19.0 before 4.22.2.
For the object it uploads, reads or deletes, the camel-aws2-s3 producer takes the key from the CamelAwsS3Key message header and falls back to the keyName option configured on the endpoint only when that header is absent. Since 4.8.0, AWS2S3Utils.determineKey() also passed the resulting value to the Simple language whenever it contained ${ or $simple{, so that a route could use a dynamic key, and the evaluation applied to a value taken from the header as well as to the configured option. The header is not route configuration, however: the aws2-s3 consumer sets CamelAwsS3Key on the exchanges it creates to the key of the object it read, that is, the name of the object in the bucket. In a route that does nothing more than consume from one aws2-s3 endpoint and produce to another, the name of each consumed object therefore reached the producer, and a name containing either marker was evaluated as an expression, although the route author never asked for any evaluation, and configuring keyName on the producer did not prevent it, because the header took precedence. A key that a route copied into the header from other message content was evaluated in the same way. Anyone able to choose the name of an object in a bucket that such a route consumes, typically a principal allowed to upload to that bucket, could therefore have Simple expressions of their choosing evaluated in the Camel process. The Simple language can read JVM system properties, environment variables and configuration properties, which commonly carry credentials, and, where camel-bean is on the classpath, invoke methods on beans in the registry. The evaluated result became the key the producer used, so reading a value back additionally required the ability to see object names in the destination bucket, while a method invocation took effect without it; together with the need to name an object in a consumed bucket, this is why the severity is assessed medium. The bucket name went through the same evaluation, but the consumer only ever places the bucket configured on its own endpoint in a bucket header, never a value derived from a remote object, so the bucket name was not reachable in this way. Releases before 4.8.0 used the key from the header as given and are not affected.
Users are recommended to upgrade to version 4.23.0, which fixes the issue. If users are on the 4.22.x releases stream, then they are suggested to upgrade to 4.22.2. If users are on the 4.18.x releases stream, then they are suggested to upgrade to 4.18.5. The 4.8.x to 4.17.x and 4.19.x to 4.21.x release lines reached end of life before this issue was resolved and will not receive a fix, so deployments on them must move to a supported release. After upgrading, a key or bucket name supplied through a message header is used literally, so a route that put an unevaluated Simple expression in the CamelAwsS3Key or CamelAwsS3OverrideBucketName header must evaluate it in the route instead, for example with setHeader and simple(), or configure it as the keyName or bucketName option of the endpoint. For deployments that cannot upgrade immediately, do not let a key derived from an object name or from other untrusted message content reach an aws2-s3 producer unchecked: in routes that pass exchanges from an aws2-s3 consumer to an aws2-s3 producer, either remove the CamelAwsS3Key header before the producer with removeHeader(“CamelAwsS3Key”), so that the keyName configured on the endpoint is used, or reject exchanges whose CamelAwsS3Key header contains ${ or $simple{. Configuring keyName alone is not a workaround, because the header takes precedence over it. As defence in depth, allow only trusted principals to write to buckets that Camel routes consume. |
| The fix for CVE-2019-0199 was incomplete and did not address HTTP/2 connection window exhaustion on write in Apache Tomcat versions 9.0.0.M1 to 9.0.19 and 8.5.0 to 8.5.40 . By not sending WINDOW_UPDATE messages for the connection window (stream 0) clients were able to cause server-side threads to block eventually leading to thread exhaustion and a DoS. |
| When running on Windows with enableCmdLineArguments enabled, the CGI Servlet in Apache Tomcat 9.0.0.M1 to 9.0.17, 8.5.0 to 8.5.39 and 7.0.0 to 7.0.93 is vulnerable to Remote Code Execution due to a bug in the way the JRE passes command line arguments to Windows. The CGI Servlet is disabled by default. The CGI option enableCmdLineArguments is disable by default in Tomcat 9.0.x (and will be disabled by default in all versions in response to this vulnerability). For a detailed explanation of the JRE behaviour, see Markus Wulftange's blog (https://codewhitesec.blogspot.com/2016/02/java-and-command-line-injections-in-windows.html) and this archived MSDN blog (https://web.archive.org/web/20161228144344/https://blogs.msdn.microsoft.com/twistylittlepassagesallalike/2011/04/23/everyone-quotes-command-line-arguments-the-wrong-way/). |
| The SSI printenv command in Apache Tomcat 9.0.0.M1 to 9.0.0.17, 8.5.0 to 8.5.39 and 7.0.0 to 7.0.93 echoes user provided data without escaping and is, therefore, vulnerable to XSS. SSI is disabled by default. The printenv command is intended for debugging and is unlikely to be present in a production website. |
| A bug in the handling of the pipelined requests in Apache Tomcat 9.0.0.M1 to 9.0.0.M18, 8.5.0 to 8.5.12, 8.0.0.RC1 to 8.0.42, 7.0.0 to 7.0.76, and 6.0.0 to 6.0.52, when send file was used, results in the pipelined request being lost when send file processing of the previous request completed. This could result in responses appearing to be sent for the wrong request. For example, a user agent that sent requests A, B and C could see the correct response for request A, the response for request C for request B and no response for request C. |
| Missing Authorization vulnerability in Apache Answer.
This issue affects Apache Answer: through 2.0.2.
An unauthenticated or unauthorized user can retrieve the invited-user list associated with a non-public question due to missing visibility authorization checks.
Users are recommended to upgrade to version 2.0.3, which fixes the issue. |
| The Apache Pulsar broker exposes a transaction administration REST API under /admin/v3/transactions. When authentication and authorization are enabled, most of the read endpoints of this API verify only that the request is authenticated and that the broker serves the requested topic or transaction coordinator. They do not check whether the caller is permitted to access the tenant, namespace or topic whose data is returned. The endpoints that modify state, which abort a transaction or scale the transaction coordinators, correctly require super-user access and are not affected.
As a result, any authenticated user, including one that has been granted no permissions, can read transaction coordinator and transaction information belonging to other tenants. This includes the list of transaction coordinators with the broker service URLs that serve them, transaction coordinator statistics and internal statistics, transaction metadata including the names of the topics and subscriptions a transaction involves and the principal that opened it, the list of long-running transactions on a transaction coordinator, transaction buffer and pending acknowledgement statistics and internal statistics for any topic and subscription, and whether a given message position of a subscription is currently held in a pending transactional acknowledgement.
The disclosure of transaction metadata and statistics can only occur when transaction support is enabled in the broker (transactionCoordinatorEnabled, which is disabled by default). When transaction support is disabled, these endpoints return no transaction data; the only residual effect is that one endpoint's error response can reveal whether a named topic exists in another tenant.
The impact is limited to disclosure of transaction-related metadata and statistics across tenant boundaries. The issue does not allow reading message contents, modifying or aborting transactions, or changing any configuration.
This issue affects Apache Pulsar versions from 2.8.0 before 4.0.14 and from 4.1.0 before 4.2.5.
- 4.0.x version users should upgrade to at least 4.0.14.
- 4.1.x and 4.2.x version users should upgrade to at least 4.2.5.
Users operating versions prior to those listed above should upgrade to the aforementioned patched versions or newer versions.
The currently supported Apache Pulsar release lines are 4.0.x and 5.0.x. The supported versions schedule is available at https://pulsar.apache.org/contribute/release-policy/.
Users who cannot upgrade immediately and do not use Pulsar transactions should keep transaction support disabled. Where transactions are required, restricting network access to the broker's admin HTTP endpoints to trusted administrative clients reduces the exposure. |
| The Apache Pulsar WebSocket service allows a WebSocket consumer to configure a dead letter policy through query parameters of the consumer endpoint, including the deadLetterTopic parameter that names the topic to which messages are written after the maximum number of redeliveries is exceeded. When authorization is enabled, the WebSocket service checks only that the connecting user holds the CONSUME permission on the subscribed topic. It does not check whether the user holds the PRODUCE permission on the topic named in deadLetterTopic. The dead letter messages are then written by the WebSocket service using its own broker client identity rather than the identity of the connecting user.
As a result, an authenticated user who holds the CONSUME permission on at least one topic can cause messages to be written to a topic of their choosing in any existing namespace, including namespaces belonging to other tenants, provided the role used by the WebSocket service to connect to the brokers is authorized to produce to that topic and the topic accepts a producer that registers no schema. The dead letter producer created by the WebSocket service registers no schema, so it is rejected when the target topic has a schema and schema validation is enforced for that topic or its namespace; this setting is not enabled by default. The same user is refused when attempting to produce to that topic directly. Where automatic topic creation is enabled, which is the default, the named topic is created if it does not already exist. The issue can only be triggered when the WebSocket service is enabled, either in the broker (webSocketServiceEnabled, which is disabled by default) or as a standalone WebSocket proxy.
The impact is unauthorized injection of messages into topics the user is not permitted to produce to. The user can determine the content of the injected messages when they also control the messages on the subscribed topic. The issue does not allow reading, modifying or deleting existing messages in the target topic. Injected messages carry the dead letter message properties and a dead letter producer name, so they can be distinguished from messages produced normally; however, the values of these properties are not authenticated and should not be relied upon to establish the origin of a message. Whether an injected message has further consequences depends on how the applications consuming the target topic process it. Repeated injection also consumes storage and backlog capacity in the target namespace.
This issue affects Apache Pulsar versions from 2.3.0 before 4.0.14 and from 4.1.0 before 4.2.5.
- 4.0.x version users should upgrade to at least 4.0.14.
- 4.1.x and 4.2.x version users should upgrade to at least 4.2.5.
Users operating versions prior to those listed above should upgrade to the aforementioned patched versions or newer versions.
The currently supported Apache Pulsar release lines are 4.0.x and 5.0.x. The supported versions schedule is available at https://pulsar.apache.org/contribute/release-policy/.
Users who cannot upgrade immediately and do not require the WebSocket service should disable it. Where the WebSocket service is required, restricting network access to it and limiting the permissions granted to the role the WebSocket service uses to connect to the brokers to the namespaces WebSocket clients are intended to access reduces the exposure. |
| The Pulsar broker serves two schema commands on its binary protocol: GetSchema, which returns the schema registered for a topic, and GetOrCreateSchema, which registers a new schema version for an existing topic. The broker handled both commands after checking only that the client connection was authenticated. Unlike the broker's other topic commands, such as those for creating producers and consumers or for looking up topics, these two commands did not check whether the authenticated client was authorized for the target topic. The Pulsar REST API endpoints for schemas were not affected.
An authenticated client without any permission on the target topic, namespace or tenant could use these commands to:
- read the schema registered for a topic in another namespace or tenant, disclosing the schema definition such as record field names and types. Message payloads are not disclosed.
- register a new schema version on an existing topic in another namespace or tenant. The topic's schema compatibility and schema auto-update policies still apply. The issue does not allow message data to be read or modified, but a registered schema version can cause legitimate producers and consumers of that topic to fail with schema incompatibility errors until the schema is corrected by an administrator.
The standard Pulsar client performs an authorized topic lookup before it sends these commands, so the missing check could not be observed when testing with the standard client. A client that implements the binary protocol directly is not required to perform that lookup first.
This issue is relevant only to deployments that have authentication and authorization enabled. The GetSchema command is affected from Apache Pulsar 2.2.0, and the GetOrCreateSchema command from Apache Pulsar 2.5.0. After the fix, the broker requires the lookup permission on the topic for GetSchema and the produce permission on the topic for GetOrCreateSchema. This impact analysis assumes that Pulsar has been configured with the default authorization provider. For custom authorization providers, the impact could be slightly different.
This issue affects Apache Pulsar versions from 2.2.0 before 4.0.14, from 4.1.0 through 4.1.3, and from 4.2.0 before 4.2.5.
- 4.0.x version users should upgrade to at least 4.0.14.
- 4.1.x and 4.2.x version users should upgrade to at least 4.2.5.
Users operating versions prior to those listed above should upgrade to the aforementioned patched versions or newer versions.
The currently supported Apache Pulsar release lines are 4.0.x and 5.0.x. The supported versions schedule is available at https://pulsar.apache.org/contribute/release-policy/. |
| Apache Pulsar's MultiRolesTokenAuthorizationProvider authorizes a client with all roles listed in a claim of its JSON Web Token (JWT). The provider is intended for use with token authentication, which verifies the token before authorization. The provider did not verify the token itself: it parsed the token's claims with the signature removed and trusted that authentication had already checked it.
When authentication and authorization are enabled with a non-empty anonymousUserRole setting, some HTTP requests whose credentials fail authentication are not rejected, but are handled as requests of the anonymous user role. For the broker's HTTP endpoints, such as the admin REST API, this fallback applies to requests that do not name an authentication method and requires strictAuthMethod to be false (the default) where that setting is available. For the WebSocket API, failed credentials can also fall back to the anonymous user role; this path does not enforce strictAuthMethod. The authentication data kept for such a request still contained the request's Authorization header. MultiRolesTokenAuthorizationProvider then read the bearer token from that header without verifying its signature and authorized the request with the roles in the token's role claim.
As a result, in deployments that use MultiRolesTokenAuthorizationProvider together with a non-empty anonymousUserRole and that permit this fallback, an unauthenticated client can send an HTTP request, for example to the admin REST API or the WebSocket API, with a forged or unsigned token that lists arbitrary roles, and be authorized with those roles. The impact can include super-user access when the token lists a role configured in superUserRoles, allowing the client to read and change tenants, namespaces, topics, permissions and broker configuration. No valid credentials are required.
Deployments are not affected if they use the default authorization provider (PulsarAuthorizationProvider) or another authorization provider, or if anonymousUserRole is not set (the default). Requests that authenticate successfully with a valid token are not affected, because the token that MultiRolesTokenAuthorizationProvider reads is the token that authentication verified. Connections that use the Pulsar binary protocol are not affected by this issue, because the broker does not keep client authentication data for connections that it assigns to the anonymous user role.
In fixed versions, MultiRolesTokenAuthorizationProvider obtains roles only from tokens that the configured token authentication provider has verified, using its configured keys and validation settings. Broker HTTP requests that are assigned to the anonymous user role are authorized only as that role, and invalid tokens cannot supply additional roles to WebSocket requests. The broker fails to start if MultiRolesTokenAuthorizationProvider is configured with authorization enabled but without a token authentication provider.
This issue affects Apache Pulsar versions from 2.8.1 before 4.0.14 and from 4.1.0 before 4.2.5.
- 4.0.x version users should upgrade to at least 4.0.14.
- 4.1.x and 4.2.x version users should upgrade to at least 4.2.5.
Users operating versions prior to those listed above should upgrade to the aforementioned patched versions or newer versions.
The currently supported Apache Pulsar release lines are 4.0.x and 5.0.x. The supported versions schedule is available at https://pulsar.apache.org/contribute/release-policy/.
Users who cannot upgrade immediately can remove the anonymousUserRole setting from the broker and WebSocket service configurations, or restrict network access to the HTTP endpoints of these services to trusted clients. |
| An integer underflow in the mcpack2pb component of Apache bRPC (version < 1.18.0) on all platforms allows remote attackers to crash a bRPC server by sending a request containing a string field with a zero value_size. The UnparsedValue::as_string() function computes `_size - 1` without validating `_size`; when value_size is 0 the expression underflows to SIZE_MAX, resize() throws std::length_error, and because the exception is not caught on the request parsing path the process aborts (SIGABRT).
Root Cause:
In src/mcpack2pb/parser.cpp, UnparsedValue::as_string() resizes the output string to `_size - 1`, where `_size` is the value_size of a string field taken directly from the network data. A value_size of 0 underflows the unsigned size_t, and the subsequent resize() throws an exception that is not caught.
Affected Scenarios:
bRPC servers that accept nshead frames wired to the ubrpc mcpack2 adaptor (for example the NsheadService + UbrpcMcpack2Adaptor deployment pattern) can be crashed by a single unauthenticated request. Stock open-source bRPC registers the ubrpc/nshead_mcpack protocols as client-only, so default server configurations are not affected.
How to Fix:
1. Upgrade bRPC to a version that includes this fix.
2. Apply the patch in PR https://github.com/apache/brpc/pull/3450 |
