All CVEs

OpenSSL CVE-2026-121481 (EVP_EncodeBlock heap overflow): OpenSSL 4.0.0-4.0.3 and 4.1.0 can write past a buffer when asked to base64-encode about 1.6 GB or more in one call. The only trigger described is openssl s_client with a huge -proxy_pass set by the person running it; not a remote bug in TLS servers or clients as far as any source shows. The 9.8 Critical score is CISA's, not OpenSSL's (no OpenSSL rating yet; Ubuntu says Medium). No public exploit, no attacks, not on KEV. Fix expected in 4.1.1 and 4.0.4 (scheduled 1 Dec 2026); patch on your normal cycle. https://tldrcve.com/openssl-encodeblock/

OpenSSL · CVE-2026-121481

OpenSSL base64

Next cycleNeeds ~1.6 GB of input you supply. No remote path known.

A bug in OpenSSL 4.0 and 4.1 can write past the end of a buffer when a program asks it to base64-encode about 1.6 GB or more in one go. The only trigger described so far is the openssl s_client command given a huge proxy password by the person running it. The Critical 9.8 score comes from CISA, which assumes a remote attacker with no login; OpenSSL hasn't rated it yet, and Ubuntu rates it Medium.

How scary is it?

updated 20 Nov
  • No: Needs control over ~1.6 GB of data an app base64-encodes first2
  • No: No public exploit1
  • No: No attacks seen yet1
  • No: Not on CISA's exploited list6

Scare-o-meter

Shrug

Real, but not worth losing sleep over.

Are you affected? only if all are true

1OpenSSL 4.0.0 to 4.0.3, or 4.1.0command

OpenSSL's CVE record lists only these versions; 3.x and older are not listed. Ubuntu disagrees in one place: it marks the Node.js package on Ubuntu 22.04, which bundles OpenSSL 1.1.1m, as vulnerable. Apps can also bundle their own OpenSSL.1,7

openssl version
2Something encodes ~1.6 GB at once

A program has to pass more than (INT_MAX / 4) * 3 bytes, about 1.6 GB, to EVP_EncodeBlock() in one call. Inside OpenSSL the known path is the proxy login of openssl s_client -proxy_pass, with the password read from an environment variable because a command line is too short to hold it. The fix changes only the base64 code and the HTTP proxy login code.1,2

3An attacker controls that data

The s_client proxy password is a setting chosen by whoever runs the command, so an attacker would already need to control your command or environment. Apps that call EVP_EncodeBlock() themselves on huge untrusted input would also be exposed. CISA's 9.8 assumes a remote attacker with no login; neither OpenSSL nor NVD has scored it.2,1,3,5

What to do

1Update when 4.1.1 / 4.0.4 shipPending · expected 1 Dec

OpenSSL's CVE record names 4.1.1 and 4.0.4 as the first fixed versions, and OpenSSL has scheduled those releases for 1 Dec 2026. On 20 Nov the fix was still under review, on hold for discussion and not merged, so check the release notes. Debian and Ubuntu had not assessed their packages yet on 20 Nov.1,4,2,9,7,8

2Only use proxy passwords you set

If you script openssl s_client -proxy_pass, don't let other people supply the password or the environment variable it's read from.2

3Cap input in your own code

If your code calls EVP_EncodeBlock() on data from others, reject inputs over (INT_MAX / 4) * 3 bytes. That's the limit OpenSSL's fix enforces.2

This page shows you how to check. It can't tell you you're safe.

Sources (9) and changelog
  1. CVE record: OpenSSL lists 4.0.0-4.0.3 and 4.1.0; CISA adds 9.8, SSVC exploitation none 18 Nov
  2. OpenSSL PR #33652: the fix, s_client -proxy_pass trigger, still open on 20 Nov 10 Nov
  3. OpenSSL: vulnerabilities list, no entry or rating for this CVE on 20 Nov
  4. OpenSSL advisory: 4.1.1, 4.0.4 releases scheduled for 1 Dec 2026 17 Nov
  5. NVD: awaiting analysis, only CISA's 9.8 score
  6. CISA KEV catalog: not listed
  7. Ubuntu: priority Medium, openssl needs evaluation, Node.js on 22.04 vulnerable 19 Nov
  8. Debian tracker: no packages assessed yet
  9. OpenSSL releases: newest are 4.1.0 and 4.0.3, no 4.1.1 or 4.0.4 yet