From: Hannes Mehnert <hannes@mehnert.org>
To: "caml-list@inria.fr" <caml-list@inria.fr>
Subject: [Caml-list] OSEC-2026-01 in the OCaml runtime: Buffer Over-Read in OCaml Marshal Deserialization
Date: Tue, 17 Feb 2026 15:26:02 +0100 [thread overview]
Message-ID: <82a75ba9-e897-49a1-ae9d-3bdf9cafd2fd@mehnert.org> (raw)
Dear everyone,
it is my pleasure to announce the first security announcement of this
year, and the first coordinated by the new OCaml security response team
(https://ocaml.org/security).
Please subscribe to the OCaml security announcement mailing list
(https://sympa.inria.fr/sympa/info/ocsf-ocaml-security-announcements) to
receive all security advisories. To this mailing list I'll only copy
those affecting the core of OCaml distribution.
It should any moment now also appear at https://osv.dev/list?q=OSEC-2026-01
Human link:
https://github.com/ocaml/security-advisories/tree/main/advisories/2026/OSEC-2026-01.md
```
id: OSEC-2026-01
modified: "2026-02-17T13:30:00Z"
published: "2026-01-24T13:30:00Z"
aliases: [ GHSA-j26j-m5xr-g23c GHSA-m34r-cgq7-jhfm ]
severity: "CVSS:3.1/AV:L/AC:L/PR:N/UI:N/S:U/C:H/I:L/A:N"
severity_score: "6.8"
affected: "ocaml" {< "4.14.3" | (>= "5" & < "5.4.1")}
events: [
[
git "https://github.com/ocaml/ocaml" [
[fixed "b0a2614684a52acded784ec213f14ddfe085d146"]
]
]
[
git "https://github.com/ocaml/ocaml" [
[fixed "e3919fef436f89271bc30bbe8592851f7289fb68"]
]
]
]
references: [
[report
"https://github.com/ocaml/security-advisories/security/advisories/GHSA-j26j-m5xr-g23c"]
]
credits: [
[reporter "Justin Timperio"]
[remediation_developer "Nicolás Ojeda Bär"]
[remediation_developer "Xavier Leroy"]
[remediation_developer "Gabriel Scherer"]
[remediation_reviewer "Xavier Leroy"]
[remediation_reviewer "Olivier Nicole"]
[remediation_verifier "Mindy Preston"]
[remediation_verifier "Edwin Török"]
[coordinator "Hannes Mehnert"]
]
cwe: [ CWE-126 CWE-502 CWE-754 ]
```
# Buffer Over-Read in OCaml Marshal Deserialization
## Summary
A critical buffer over-read vulnerability in OCaml's Marshal
deserialization (runtime/intern.c) enables remote code execution through
a multi-phase attack chain. The vulnerability stems from missing bounds
validation in the readblock() function, which performs unbounded
memcpy() operations using attacker-controlled lengths from malicious
Marshal data.
Please note that Marshal is not type safe, and you have to be careful if
you use the deserialization on untrusted input (due to type confusion,
and remote code execution by design - you can use Marshal for code).
Affected functions: `Marshal.from_channel`, `Marshal.from_bytes`,
`Marshal.from_string`, `Stdlib.input_value`, `Pervasives.input_value`
when reading data from an untrusted source.
## Vulnerability Attack Vector
Corrupted or malicious marshaled data that causes undefined behaviour in
the runtime system when unmarshaled.
`input_value` should either fail cleanly or produce a well-formed OCaml
object, without corrupting the runtime system.
Consequently, this excludes:
* well-formed marshaled data that produces an OCaml object that is not
of the type expected by the OCaml code and causes the Ocaml code to
crash or misbehave
* misuses of the OCaml runtime system by the program performing
input_value, such as setting `Debugger.function_placeholder` to the
wrong function.
The former issue may be addressed at some point by validating the
unmarshaled OCaml value against the expected type, using the functions
from module `Obj` and some kind of run-time type description.
The latter issue is a bug in the program that unmarshals the data.
## Fix
### OCaml runtime
The OCaml runtime has been hardened with additional bounds checks. An
exception is raised on bad input.
### Third party libraries
Third party libraries that want to harden their custom Marshal
deserialization code can follow the example fix for bigarrays from the
standard library.
There are new macros in `custom.h` called `Wsize_custom_data` and
`Bsize_custom_data` that return the size in words or bytes of the
allocated custom destination block. The deserializer needs to ensure it
only writes data within those bounds.
This only needs to be done if the library defines a custom type in a C
binding, and `struct custom_operations`'s `deserialize` field is not set
to `NULL` or `custom_deserialize_default`, and `struct
custom_operations`'s `fixed_length` field is set to `NULL` or
`custom_fixed_length_default`
Since `Marshal.from*` and `input_value` remain unsafe to use, the fix
for the OCaml runtime is released, and we wouldn't attempt to coordinate
updating all deserialization functions in the ecosystem.
## Timeline
- Nov 4th 2025: Discovery Date: Discovered first in OxCaml
- Nov 5th 2025: First Disclosure Date (Jane Street Team): Emailed top
maintainers, no response.
- Nov 9th 2025: Second Disclosure Date (OCaml Team): Submitted to
OCaml/ocaml GitHub Repo as a Security Advisory.
- Nov 11th 2025: Emailed OCaml Security Mail List: Submitted to OCaml
over email, responded asking for details.
- Nov 11th 2025: Third Disclosure (OCaml Security Response Team):
Submitted to ocaml/security-advisories GitHub Repo as a Security Advisory.
- Dec 16th 2025: Initial patch is developed
- Dec 17th 2025: Fuzz testing found further issues
- Dec 24th 2025: Final patch for OCaml is developed
- Dec 25th 2025: Fuzz testing couldn't find any further issues
- Jan 2nd 2026: Patch got reviewed by OCaml maintainers
- Jan 4th 2026: Benchmarking of the patch with good results
- Jan 6th 2026: Reporter got contacted to confirm
- Jan 25th 2026: Further related issues discovered by fuzzing
- Feb 17th 2026: fixed OCaml releases are published, security advisory
is published
next reply other threads:[~2026-02-17 14:26 UTC|newest]
Thread overview: 2+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-02-17 14:26 Hannes Mehnert [this message]
2026-02-17 14:33 ` Hannes Mehnert
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=82a75ba9-e897-49a1-ae9d-3bdf9cafd2fd@mehnert.org \
--to=hannes@mehnert.org \
--cc=caml-list@inria.fr \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox