CVE-2026-121757: CP949 decoder may read past input
Converting crafted CP949 input to UTF-8 character encoding and ignoring invalid characters, in the GNU C Library version 2.1 or later may result in the converter reading past the end of the input buffer, causing the calling application to crash.
When the CP949 decoder encounters an invalid input byte sequence like '\xa2\xe8' it ends up advancing the pointer tracking the read buffer position twice. If //IGNORE is set on the destination encoding and this sequence is at the end of the input buffer, the decoder may then end up reading past the bounds of the input buffer. To exploit this, the target iconv function needs to set //IGNORE on the destination of the conversion and the input and encoding needs to be attacker controlled.
Affected Software
Event History
Frequently Asked Questions
Which applications are exposed?
Applications are exposed if they use iconv to convert attacker-controlled CP949 input to UTF-8, set //IGNORE on the destination encoding, and may process an invalid byte sequence at the end of the input buffer.
What must an attacker control to trigger the issue?
The attacker needs control of both the input data and the input encoding used for the conversion. The conversion must use CP949 input and a destination encoding configured with //IGNORE.
What is the practical impact of successful exploitation?
The out-of-bounds read can cause the calling application to crash. The provided vector and description indicate an availability impact, with no confidentiality or integrity impact stated.
How can teams determine whether they are affected?
Review iconv call sites for CP949-to-UTF-8 conversions that append or otherwise set //IGNORE on the destination encoding. Prioritize paths where untrusted data can reach the converter and invalid CP949 sequences may occur at the end of a buffer.