CVE-2026-84372: Predis: Redis command injection and denial of service via CRLF smuggling in pipelined commands on aggregate connections
Predis is a flexible and feature-complete Redis and Valkey client for PHP. From version 3.0.0-RC1 until version 3.3.0, pipeline handling on aggregate cluster and replication connections reparses an already serialized RESP buffer in AbstractAggregateConnection::write() by splitting it with explode("\r\n") instead of honoring RESP length prefixes. Attacker-controlled keys or values containing CRLF sequences can therefore be interpreted by Command::deserializeCommand() as additional commands. On cluster connections, ClusterStrategy::getFakeKey() can route injected keyless commands using the literal fake key value "key", permitting operations such as shard-wide cache deletion, targeted data modification, data reads, or node disruption. On replication connections, malformed reparsing can throw an uncaught exception and repeatedly terminate affected requests. Only pipeline() reaches this vulnerable path; transaction() and MULTI are not affected. This issue is fixed in version 3.3.0.
Affected Software
Remediation
Recommended actions to resolve this vulnerability, in priority order.
- Upgrade
Upgrade
Predisto a version that resolves this vulnerability.Fixed in 3.3.0
Event History
Frequently Asked Questions
Which deployments are affected?
Predis versions from 3.0.0-RC1 through 3.3.0 are affected when applications use pipeline() over aggregate cluster or replication connections. The vulnerable path is not reached by transaction() or MULTI.
What does an attacker need to exploit this issue?
An attacker needs control over a Redis command key or value that is included in a pipelined command and contains CRLF sequences. No Redis authentication or application privileges are described as prerequisites, but the application must pass attacker-controlled input into the affected pipeline path.
What is the practical impact on cluster and replication connections?
On cluster connections, injected commands may enable shard-wide cache deletion, targeted data modification, data reads, or node disruption. On replication connections, malformed reparsing can cause an uncaught exception that repeatedly terminates affected requests.
What should be done if an immediate upgrade is not possible?
Avoid using pipeline() with aggregate cluster or replication connections for commands containing untrusted keys or values, and reject or sanitize CRLF sequences in those inputs. Upgrading to the fixed version, 3.3.0, is the documented remediation.