
Cloud Vulnerability DB
A community-led vulnerabilities database
CVE-2026-89160 is an out-of-bounds read vulnerability in PCRE2 (Perl Compatible Regular Expressions 2) affecting versions 10.34 through 10.47 (including 10.48-rc1). The flaw exists in pcre2_match() and is triggered when matching invalid UTF subjects using the PCRE2_MATCH_INVALID_UTF flag, causing the engine to read heap memory before the start of the subject buffer. It was reported by Ilia Alshanetsky and publicly disclosed on August 31, 2026, with a fix released in PCRE2 10.48. The CVSS v3.1 base score is 6.5 (Medium) per Feedly/NVD, though the official GHSA advisory rates it 3.7 (Low) using a higher attack complexity assumption (GHSA Advisory, Red Hat).
The root cause is CWE-125 (Out-of-bounds Read): two backward-walking code paths in pcre2_match.c (Site A: OP_VREVERSE at line 6236) and pcre2_extuni.c (Site B: Regional Indicator count-back at line 124) use mb->start_subject as their lower bound instead of the correct mb->check_subject. Under PCRE2_MATCH_INVALID_UTF, check_subject marks the boundary between the unvalidated prefix [start_subject, check_subject) and the UTF-validated region [check_subject, end_subject). When the subject begins with a UTF-8 continuation byte (0x80–0xBF), the unbounded BACKCHAR(eptr) macro — defined as while ((*eptr & 0xc0u) == 0x80u) eptr-- — steps past start_subject and reads 1 to N bytes before the allocated heap buffer. Exploitation requires: (1) the application links libpcre2-8 or libpcre2-16; (2) patterns are compiled with both PCRE2_UTF and PCRE2_MATCH_INVALID_UTF; (3) the pattern contains variable-length lookbehind ((?<=...)) or extended grapheme matching (\X); and (4) the application accepts attacker-controlled subject strings. The 32-bit code-unit width and DFA matcher are not affected (GHSA Advisory).
Successful exploitation results in a heap out-of-bounds read of 1 to N bytes of adjacent heap memory, enabling potential information disclosure of sensitive data (e.g., session tokens, cryptographic keys, or process secrets) residing in adjacent heap allocations. A secondary impact is denial of service: if the out-of-bounds read lands in an unmapped memory page, the process crashes. Integrity is not affected. The vulnerability is particularly relevant to applications that use PCRE2 with PCRE2_MATCH_INVALID_UTF for processing untrusted input, such as log scanners, HTTP request parsers, protocol matchers, and anti-spam/DLP regex engines (GHSA Advisory).
Proof-of-concept (PoC) exploit code is publicly available in the GHSA advisory, consisting of two complete, standalone C programs (poc-ss001-op_vreverse.c and poc-ss002-extuni-ri.c) that compile and run against PCRE2 10.47 to demonstrate the OOB read with AddressSanitizer detection. There is no evidence of in-the-wild exploitation at this time, and no threat actor attribution has been reported. The EPSS score is 0.0022 (approximately 0.22%), indicating low probability of near-term exploitation. The vulnerability is not listed in the CISA KEV catalog. NVD SSVC classifies exploitation status as "poc" with "no" automatability (GHSA Advisory, Feedly).
libpcre2-8 or libpcre2-16 and compile patterns with both PCRE2_UTF and PCRE2_MATCH_INVALID_UTF flags — common in log scanners, HTTP parsers, and protocol matchers accepting untrusted input.0x80), followed by content that triggers the vulnerable backward-walking code path.(?<=a{1,2})X) compiled with PCRE2_UTF | PCRE2_MATCH_INVALID_UTF. Example subject: [0x80, 'X'].0x80 followed by two Regional Indicator codepoints (e.g., U+1F1E6 U+1F1E7) to an application using a pattern containing \X\X compiled with the same flags.pcre2_match() processes the subject, the backward-walk loop crosses check_subject into the unvalidated prefix, and the unbounded BACKCHAR macro decrements the pointer past start_subject, reading 1+ bytes before the allocated heap buffer.subject[-1] onward are processed as UTF character data, potentially disclosing adjacent heap contents (information disclosure) or crashing the process if the read lands in unmapped memory (denial of service) (GHSA Advisory).pcre2_match.c (around line 6236–6243) or pcre2_extuni.c (around line 124–129); AddressSanitizer output containing heap-buffer-overflow with READ of size 1 located 1 bytes to the left of the subject buffer.libpcre2-8.so or libpcre2-16.so.Upgrade PCRE2 to version 10.48 or later, which applies a lower buffer bound (check_subject) to prevent the two out-of-bounds reads during backward scanning through invalid UTF data. As a workaround for systems that cannot be immediately patched, avoid using PCRE2_MATCH_INVALID_UTF with untrusted subjects, or validate/sanitize all subject strings before passing them to pcre2_match(). SUSE has released a security update (SUSE-SU-2026:4201-1) for affected distributions. Microsoft has also published guidance for affected products (PCRE2 Release, GHSA Advisory, Microsoft MSRC).
The PCRE2 maintainer (NWilson) published the advisory and fix simultaneously with the PCRE2 10.48 release on August 31, 2026, describing it as a security fix for backward scans crossing the check_subject boundary. The reporter, Ilia Alshanetsky, provided a detailed technical report including two PoC programs and suggested CVSS scoring. SUSE issued a security update for its distributions, and the vulnerability was picked up by oss-security mailing lists and security aggregators including AusCERT and pro-linux.de. No significant social media controversy or broader community debate has been observed (GHSA Advisory, PCRE2 Release).
Fix availability across major Linux distributions and their releases.
bookworm
pcre2: 10.42-1+deb12u1
sid
pcre2: 10.48-1
trixie
pcre2: 10.46-1~deb13u2
bionic (esm-apps)
pcre2
devel
pcre2
focal (esm-infra)
pcre2
jammy
pcre2
noble
pcre2
resolute
pcre2
xenial (esm-apps-legacy)
pcre2
OpenShift
openshift/ose-rhel-coreos-9
RHEL 8
mariadb:10.11/mariadb.src
RHEL 9
mariadb:10.11/mariadb.src
RHEL 10
mariadb10.11.src
Source: This report was generated using AI
Free Vulnerability Assessment
Evaluate your cloud security practices across 9 security domains to benchmark your risk level and identify gaps in your defenses.
Get a personalized demo
"Best User Experience I have ever seen, provides full visibility to cloud workloads."
"Wiz provides a single pane of glass to see what is going on in our cloud environments."
"We know that if Wiz identifies something as critical, it actually is."