Use-After-Free in HTTP/2 Upload Leads to Potential RCE in curl: Detailed Analysis of CVE-2026-92392
curl 8.23.0 closed one of the most severe vulnerabilities in recent years, rated HIGH and assigned CVE-2026-92392 (HTTP/2 upload Use-After-Free). The flaw allows a malicious server to trigger remote code execution under specific conditions during data uploads over HTTP/2.
Use-after-free is a classic memory-safety issue in languages without guarantees such as C. A pointer remains valid after the referenced memory has been freed, and an attacker who can control the contents of that memory region can alter program behavior.
HTTP/2 multiplexing requires libcurl to manage multiple concurrent transfers over a single connection. Each transfer uses an easy handle created by the application; a multi handle coordinates any number of easy handles.
To trigger the bug, the client must upload data to an HTTPS server that supports HTTP/2. The server pauses the upload by zeroing the flow-control window, sends a partial response, then resets the stream. libcurl correctly finishes the transfer but leaves a dangling pointer to the easy handle. When the application later frees the handle and closes the multi handle, libcurl reuses the stale pointer while processing incoming data.
The library allocates a receive buffer, reads attacker-controlled data into the former location of the easy handle, and eventually calls a function pointer that has been overwritten by the downloaded payload, resulting in arbitrary code execution. The attack also leaks internal client details that help bypass ASLR.
Successful exploitation demands many conditions: the victim must upload roughly 135 KB, the attacker must know the target architecture and memory layout, and specific allocation patterns involving nghttp2 and OpenSSL TLS 1.3 must align. The payload area is limited to about 23 bytes, making a crash far more likely than code execution.
The vulnerability was reported by Luna Tong of Anthropic via their OSS Scanner. The curl security team confirmed it within hours on 15 September and shortened the release cycle from eight to six weeks to ship the fix promptly.
Developers are reducing the number of internal pointers to handles, replacing them with transfer identifiers that are looked up on demand. Additional tests and static analyzers are being introduced to catch similar errors earlier.
Related articles
PHP2Shell Vulnerabilities Enable Pre-Application Code Execution on Unpatched Servers
The article explains how attackers can exploit two known PHP vulnerabilities, CVE-2026-109618 and CVE-2026-109619, collectively referred to as PHP2Shell, to run arbitrary code inside the PHP process before any application-level code executes. This grants the attacker the same privileges as the PHP user, typically www-data on Ubuntu systems running Apache with mod_php, allowing access to configuration files, database credentials, and the ability to write malicious files if permissions permit. The author demonstrates practical attack paths including reading config.php files, connecting directly to MySQL databases, and creating persistent backdoors such as hack.php that execute shell commands via URL parameters. Even when standard hardening measures like open_basedir, disabled dangerous functions, and separate system users are applied, the attacker can still execute code within PHP limits until the vulnerabilities are patched. The recommended immediate actions include updating PHP, restricting write permissions, disabling PHP execution in upload directories, and placing a hardened reverse proxy such as Cloudflare in front of legacy servers while blocking all direct traffic via firewall rules.
Citrix Issues Critical XenServer Patches for Host Compromise Vulnerabilities
Citrix published security bulletin CTX697256 on November 3, 2026, disclosing eight vulnerabilities affecting XenServer. The most severe issue, tracked as CVE-2026-98432 and rated critical, enables an unauthenticated attacker who reaches the management network to compromise the virtualization host and associated appliances. A second flaw, CVE-2026-102703, allows a privileged user inside a guest virtual machine to break out and take control of the underlying host, violating isolation guarantees. Six additional vulnerabilities impact network boot and secure boot processes, potentially allowing attackers to compromise VMs during startup or bypass code-signing policies. Affected products include XenServer 8.4 and 9 along with Workload Balancing and Conversion Manager appliances. Citrix recommends immediate application of the fixes and temporary isolation of the management network while updates are deployed.
Microsoft Paint Patch for CVE-2026-83844 Fails to Reach Some Windows 11 Users via Microsoft Store
Microsoft fixed a remote code execution vulnerability in the Paint application for Windows 11, but the update is not reaching all users through the Microsoft Store. The flaw, tracked as CVE-2026-83844, carries a CVSS score of 8.1 and involves an out-of-bounds memory read that could allow unauthenticated network attacks. Although exploitation is rated as highly complex, the issue is classified as important by Microsoft. The patch is delivered separately from Windows Update, requiring users to verify the Paint version manually in the Microsoft Store. The clean version is listed as 11.2609.401.0, yet multiple systems remain on 11.2603.251.0 despite automatic updates being enabled. Windows Latest reported the distribution problem after checking its own systems and collecting similar reports from administrators. No real-world exploitation of the vulnerability has been observed so far.
Mozilla Releases Firefox 159.0.1 to Patch High-Severity Layout Vulnerability CVE-2026-120572
Mozilla Foundation has issued Firefox 159.0.1 to address a high-severity vulnerability tracked as CVE-2026-120572. The flaw resides in the Layout: Block and Inline component responsible for web page rendering. It stems from improper handling of boundary conditions during layout processing. CISA assigned the issue a CVSSv3.1 base score of 8.8, confirming high impact. The update was published on November 3, 2026, and is rated high severity by both Mozilla and CISA. Users are advised to apply the patch immediately to mitigate potential exploitation risks.