CVE-2025-46817

This is classified as Moderate because exploitation requires an authenticated client with permission to execute Lua scripts (for example via EVAL/EVALSHA) rather than an unauthenticated network attacker, narrowing the exposure to already trusted principals or compromised application identities. The flaw is an integer overflow in Lua script handling that can corrupt memory, but reliable code execution is non-trivial and there are no public proofs of concept or known in-the-wild exploits as of the initial advisories, which lowers immediate operational risk despite the theoretical impact. Practically, triggering a controlled overflow through the embedded Lua runtime adds exploit complexity, as the attacker must shape arithmetic and memory behavior inside Redis’s scripting environment to gain control flow, increasing the likelihood of crashes over deterministic RCE in typical configurations. Even in a worst case, the effect is within the Redis server process context rather than a guaranteed system-wide compromise, further reducing systemic blast radius if the service is not over-privileged.

This flaw exists only in the Redis server implementation;

Redis client libraries (Python, Node.js, Rust, etc.) are not affected by this vulnerability, and it only exists in the Redis server’s embedded Lua engine where scripts execute. Client libraries merely transmit EVAL/EVALSHA to the server.