On Oct 19 the Canadian Centre for Cyber Security added Update 1 to its Redis advisory AV26-423: "Open-source reporting indicates that CVE-2026-25243 is being exploited in the wild." The flaw (CVSS 4.0 7.7, High; GHSA-c8h9-259x-jff4) is in RESTORE, the command that loads a serialized value back into Redis. RESTORE doesn't properly validate the payload, so an authenticated user allowed to run RESTORE can trigger invalid heap memory access that Redis says may lead to remote code execution. Redis fixed it on May 5, 2026, alongside four related bugs (CVE-2026-23479, CVE-2026-25588 in RedisTimeSeries, CVE-2026-25589 in RedisBloom, CVE-2026-23631). Credit goes to Emil Lerner (Wiz Zeroday Cloud) and Joseph Surin.

The public exploitation report comes from Defused, which says its Redis decoys recorded repeated scripted attempts overnight Oct 16-17. One source hit a decoy five times on a timer with the same sequence: fingerprint the server, shape the heap with 26 key writes, send one malformed-zipmap RESTORE that triggers the over-read, run a throwaway Lua compute probe, then FLUSHALL, which wipes the keyspace and its own traces. That's honeypot activity. No victim has been named, and CISA hasn't added the CVE to KEV.

Do now: Redis OSS/CE fixed releases are 6.2.22, 7.2.14, 7.4.9, 8.2.6, 8.4.3 and 8.6.3. Redis Software fixed builds are 7.2.4-153, 7.4.6-279, 7.8.6-253, 7.22.2-79 and 8.0.10-64. Redis Cloud is already patched. Redis's documented workaround, if you can't patch yet, is an ACL rule that stops users from running RESTORE. Because exploitation needs a login, the real exposure is Redis that is reachable with no password, a shared default user, or a password sitting in app configs. Also patch distro builds and Valkey packages per your vendor's advisory.