Security researchers published a working remote code execution exploit for Redis versions 6.2.22, 7.4.9, 8.6.4, and 8.8.0 on July 22, and the PoC repo crossed 200 GitHub stars in under 24 hours. The exploit targets two separate vulnerabilities: a double-free bug in the stream consumer-group shared-NACK mechanism (the CVE-2026-25589 family) affecting three major versions, and a newly discovered TDigest heap overflow in RedisBloom, a bundled module in Redis 8.8.0.
If you run Redis in production, this is the kind of disclosure that keeps infrastructure teams awake. The exploit requires no special privileges beyond a valid Redis connection with EVAL, RESTORE, and XGROUP commands available. Most Redis deployments exposed to the internet qualify.
Background: What the Exploit Actually Does
The PoC, published by a group of security researchers under the berabuddies GitHub org, includes exploits for four Redis versions using two entirely different attack vectors.
For Redis 6.2.22, 7.4.9, and 8.6.4, the exploit weaponizes an incomplete fix for CVE-2026-25589, a stream consumer-group shared-NACK double-free vulnerability. The researchers found that Redis Corp's earlier patch did not fully close the hole. The double free allows an attacker to corrupt heap memory and gain code execution.
For Redis 8.8.0, the exploit takes a different approach. Redis 8.8 ships with a bundled RedisBloom module, which includes a TDigest implementation. The researchers found a heap overflow in the TDigest percentile calculation code. By spraying tdigests to build a deterministic jemalloc corridor, they achieve reliable code execution.
The exploit is designed to be non-destructive on first run. It writes a proof file under /data/pwned and leaves a handful of inert keys behind. But the README warns that wrong offsets will crash the target server.
Key Insights: What This Means for Redis Users
The most concerning takeaway is that Redis 8.8.0 was supposed to be the safe version. Redis Corp fixed the original shared-NACK bug in PR #15081, which shipped in 8.8.0. But the researchers simply pivoted to a completely different vulnerability in the RedisBloom module, which ships enabled by default. This means no version of Redis currently deployed is immune.
Here is why this matters to every team running Redis:
- Redis 6.x is now a liability. Version 6.2.22 is three years old and has no business being exposed in 2026. If you are still on Redis 6, the exploit works trivially. No debug flags needed.
- Default module bundling creates attack surface. RedisBloom is useful, but shipping it bundled by default in 8.8.0 gave the exploit a free second vector. The TDigest heap overflow is in code most users never interact with directly.
- The PoC quality is production-grade. The code includes calibration scripts for non-official builds, CRC64 computation for valid RESTORE payloads, and boot-retry wrappers. This is not a researcher proof-of-concept. It is a weapon-ready exploit.
The broader pattern here goes beyond Redis. Core infrastructure databases are becoming favorite targets because they are long-lived, rarely patched, and widely exposed. The Redis PoC follows a similar trajectory to the Apache Log4j and Elasticsearch RCE disclosures of 2022-2023: a foundational tool, a years-old partial fix, and researchers who poked harder.
What This Means for Founders
If you run Redis in production, here is your immediate action plan.
First, check your Redis version. Run redis-server --version on every instance. If you are on 6.2.22, 7.4.9, or 8.6.4, you are exploitable through the shared-NACK path. If you are on 8.8.0, you are exploitable through the TDigest path. There is no version of Redis currently deployed where this PoC does not work on at least one exploit path.
Second, audit your network exposure. The exploit requires network access to the Redis port. If Redis is not behind a firewall or VPN, close the port now. Every Shodan-searchable Redis instance is at risk. And if you are using cloud-managed Redis like Upstash or Redis Cloud, check with your provider whether they have deployed mitigation patches.
Third, watch for the official fix. Redis Corp will likely ship patched releases this week. The moment those land, patch immediately. Do not wait for the next maintenance window. Infrastructure teams that delay Redis patches are the ones dealing with cryptominers in their Redis clusters a week later.
For founders building products on Redis: this is also a supply chain risk for you. If your infrastructure vendor, caching layer, or queue system runs Redis, this vulnerability flows downstream to your service. Ask your cloud provider for their mitigation status today, not tomorrow.
The bottom line: Redis is not your problem. Ignoring Redis is. A published, weapon-grade PoC for your database cache layer is a ticking clock. Patch, isolate, or migrate. The researchers gave the industry a free warning shot before the dark web gets the easier versions.