Disclaimer: This content was generated using AI. While we strive for accuracy, please verify critical information with official sources.
Sources: National Vulnerability Database, CISA KEV, Exploit-DB
CVE-2026-110759 Overview
CVE-2026-110759 affects Text::CSV_XS versions before 1.65 for Perl. The parser returns a 0xFF byte that was never present in the input when an empty last field is followed by a lone carriage return (CR). Calling parse("a,\r") with binary mode enabled yields the fields "a" and "\\xFF". On UTF-8 flagged input, the trailing field holds a malformed UTF-8 sequence containing that single byte. CRLF files read line by line and chomped trigger this condition whenever the last column is empty, so downstream applications process a byte that does not exist in their source data [CWE-1286].
Critical Impact
Applications parsing untrusted CSV data receive injected 0xFF bytes in empty trailing fields, producing malformed UTF-8 strings and corrupting downstream data integrity checks.
Affected Products
- Text::CSV_XS for Perl, all versions prior to 1.65
- Perl applications relying on Text::CSV_XS for CRLF-terminated CSV parsing
- Pipelines consuming UTF-8 flagged CSV input through Text::CSV_XS
Discovery Timeline
- 2026-10-16 - CVE-2026-110759 published to NVD
- 2026-10-20 - Last updated in NVD database
Technical Details for CVE-2026-110759
Vulnerability Analysis
The flaw lives in the parse() function inside CSV_XS.xs. When the parser reaches end-of-input immediately after a field separator and a lone CR, it treats the EOF sentinel as the next data byte and stores it as 0xFF. The resulting field value never appeared in the input stream. With binary enabled, the byte is silently returned. With UTF-8 flagged input, the field becomes a UTF-8 flagged string holding a lone 0xFF byte, which is not valid UTF-8 and can trigger decode errors or validation bypasses later in the pipeline.
The condition is reachable through routine CSV handling. Any application reading a CRLF file line by line, chomping the newline, and passing each row to parse() will hit this path whenever the final column is empty. The same input with a non-empty last field is correctly rejected as a parse error, so defensive callers that only check the error flag do not see any signal.
Root Cause
The parser's restart logic did not distinguish between a CR that is part of an end-of-line sequence and a CR at the absolute end of input. The internal seenSomething branch fell through to a restart that promoted the EOF marker into field content. The fix adds an explicit check: when not driven by an I/O handle, a trailing lone CR is now reported through ERROR_INSIDE_FIELD (2032) rather than silently stored.
Attack Vector
An attacker who controls CSV content, or any upstream file source, can craft input ending in ,\r to force a 0xFF byte into the final field. Impact depends on how the consuming application uses the parsed value: UTF-8 validators, hash computations, signature checks, database inserts, and downstream serializers can all diverge from the original input.
#endif
unless (seenSomething)
break;
+ /* parse (): a CR at the very end of the input is not part
+ * of an EOL, as for a non-empty last field; do not go on
+ * with EOF as if it were a data byte */
+ unless (csv->useIO)
+ ERROR_INSIDE_FIELD (2032);
goto restart;
}
Source: GitHub Commit Patch
Detection Methods for CVE-2026-110759
Indicators of Compromise
- Parsed CSV fields containing a trailing 0xFF byte that does not exist in the raw source file.
- UTF-8 decode errors or Malformed UTF-8 warnings emitted by Perl when iterating Text::CSV_XS output.
- Hash or checksum mismatches between the original CSV file and reserialized rows.
Detection Strategies
- Inventory Perl environments and extract installed Text::CSV_XS versions, flagging any below 1.65.
- Add regression tests that call parse("a,\r") with binary => 1 and assert the second field is empty, not \\xFF.
- Scan log aggregation systems for Perl UTF-8 validation errors tied to CSV ingestion jobs.
Monitoring Recommendations
- Monitor CSV ingestion pipelines for anomalous 0xFF bytes in parsed output using data quality rules.
- Alert on Perl Encode warnings or utf8::decode failures correlated with CSV processing hosts.
- Track CPAN module version drift across production servers to catch unpatched Text::CSV_XS installations.
How to Mitigate CVE-2026-110759
Immediate Actions Required
- Upgrade Text::CSV_XS to version 1.65 or later across all Perl runtimes.
- Audit batch jobs and ETL scripts that parse externally supplied CSV files for the affected code path.
- Reprocess recently ingested CSV data if downstream systems may have stored injected 0xFF bytes.
Patch Information
The fix landed in Text::CSV_XS 1.65. See the MetaCPAN Release Changes and the upstream GitHub Commit Patch. The patch raises error 2032 for a trailing lone CR when parsing without an I/O handle, preventing the EOF marker from being stored as data.
Workarounds
- Preprocess input with a regex that strips trailing lone CR characters before calling parse().
- Prefer the getline() streaming interface over parse() on chomped lines, since useIO suppresses the faulty branch.
- Validate parsed fields for 0xFF bytes and reject rows that contain them when the raw input does not.
# Configuration example
cpanm Text::CSV_XS@1.65
perl -MText::CSV_XS -e 'print $Text::CSV_XS::VERSION, "\n"'
Disclaimer: This content was generated using AI. While we strive for accuracy, please verify critical information with official sources.
Sources: National Vulnerability Database, CISA KEV, Exploit-DB

