Friday, 20 November 2026

Trusted News Discovery – Powered by AI and Verified by Humans

OpenSSL Base64 encoder flaw scored 9.8 out of 10 in US database while its fix awaits review

An OpenSSL heap buffer overflow in EVP_EncodeBlock(), CVE-2026-121481, carries a 9.8 CISA score as its GitHub fix stays under review.

Produced with AI assistance and reviewed by the Badiyaa Editorial Desk. How we work.

Verified by humans. Checked by machines. Open to you.

Grok Fact CheckCheck it yourself in one tapVerify with Grok ↗Ask @grok on X ↗
A corridor between rows of server cabinets in a data center.
Image: BalticServers.com, CC BY-SA 3.0 via Wikimedia Commons

A bug in the function OpenSSL uses to turn raw bytes into Base64 text now carries a severity score of 9.8 out of 10, the “critical” band, in the US National Vulnerability Database. The flaw, CVE-2026-121481, sits in EVP_EncodeBlock() and affects the newest OpenSSL branches, 4.0 and 4.1. Its fix was still under review on GitHub on November 20.

What the bug does

The National Vulnerability Database entry for CVE-2026-121481, published on November 18 with OpenSSL as the reporting authority, says that calling the function with an input length close to INT_MAX can overflow the calculation of the output size and write past the end of the buffer the caller supplied. That is a heap buffer overflow, filed under weakness type CWE-122.

The 9.8 score does not come from OpenSSL. It was added by the US Cybersecurity and Infrastructure Security Agency, which rated the bug as reachable over the network, with low complexity, no login and no user action. The same CISA assessment marks exploitation as “none” so far and the attack as automatable. NVD itself still lists the record as awaiting analysis.

Why the input has to be so large

The math explains why ordinary use is unlikely to trigger it. As Wikipedia’s explainer on Base64 sets out, the encoding turns every three bytes into four printable characters, a growth of about a third. A signed 32-bit integer tops out at 2,147,483,647. Feed the function roughly 2 billion bytes and the encoded result would need about 2.67 billion bytes (2 billion times 4, divided by 3), well past that ceiling, so a size counter held in an int wraps around and the function writes beyond memory it was given. Reaching that point means an application must pass gigabytes of attacker-controlled data into a single call.

The fix is still being argued over

Nicky Mouha opened the repair in OpenSSL pull request 33652 on November 10. It was labeled as a triaged bug with “severity: important” a day later, then put on “hold: discussion” on November 13. The first version simply returned -1 for invalid input; later revisions call OPENSSL_assert() for inputs that would cause undefined behavior, add the same guard to EVP_DecodeBlock(), and correct a length check in the base64encode() helper inside OpenSSL’s HTTP client. As of November 20 the pull request remained open.

NVD’s affected-version data lists OpenSSL 4.1.0 up to 4.1.1 and 4.0.0 up to 4.0.4. OpenSSL’s security advisory of November 17, which covers three other low-severity bugs, says releases 4.1.1 and 4.0.4 are scheduled for December 1, 2026.

Why OpenSSL’s rating may differ

Readers should expect OpenSSL to weigh the bug differently. The project’s security policy states that it does not use CVSS to set severity, because the score “does not properly reflect the broad usage of the OpenSSL Library,” and that scores from outside parties “might severely differ” from its own. The same policy also excludes API misuse in cases where the function is not meant to receive attacker-controlled data directly.

Share this story

Base64 handling has been a weak spot elsewhere this month, too: Badiyaa covered the Java4Shell flaw in Java’s Base64 decoder and an nginx bug that let attackers run code without logging in.

Developers on OpenSSL 4.x who pass large, untrusted buffers to EVP_EncodeBlock() can cap the input length in their own code until the December 1 releases arrive.

Sources

Image: BalticServers.com, CC BY-SA 3.0 via Wikimedia Commons

This article was produced with AI assistance and reviewed by our Editorial Desk. How we work.

Spot an error? Help verify this story

Join our Reader Verifiers. Every report is read by an editor.

Grok Fact CheckCheck it yourself in one tapVerify with Grok ↗Ask @grok on X ↗

Verified by humans. Checked by machines. Open to you.