How Denial of Service via Invalid UTF-8 Input happens in Go and how to fix it

How Denial of Service via Invalid UTF-8 Input Happens in Go and How to Fix It

The Quiet Danger of Unicode Normalization

The parser/go.mod file in this repository lists golang.org/x/text as an indirect dependency — a library pulled in to handle Unicode normalization. It's the kind of dependency that sits quietly in a module file, rarely thought about, until a malformed byte sequence turns a routine text-processing call into an infinite loop that never returns.

That's exactly what CVE-2026-56852 enables. A single crafted input containing invalid UTF-8 bytes can cause norm.Iter — the iterator at the heart of golang.org/x/text's normalization pipeline — to spin forever, consuming 100% of a CPU core and making the parser completely unresponsive. No crash, no error, just silence and a hung goroutine.

The Vulnerability Explained

What Is norm.Iter and Why Does It Loop?

golang.org/x/text/unicode/norm provides Unicode normalization forms (NFC, NFD, NFKC, NFKD). The norm.Iter type is an iterator that walks through input bytes, yielding one normalized segment at a time. Internally, it tracks its position in the input buffer and advances after processing each segment.

The bug in versions prior to v0.39.0 is a failure to advance the iterator's position when it encounters an invalid UTF-8 byte sequence. Instead of skipping the unrecognized byte and moving forward, the iterator re-processes the same position repeatedly — a classic CWE-835: Loop with Unreachable Exit Condition.

The vulnerable dependency declaration in parser/go.mod was:

// parser/go.mod (before fix)
golang.org/x/text v0.34.0 // indirect

Any code path in the parser that calls into golang.org/x/text's normalization functions — directly or transitively — inherits this bug.

How Could an Attacker Exploit This?

Consider the typical flow in a game demo parser (which this repository implements): a player's name, chat message, or any string field in a .dem file is read from disk or over the network and passed through text processing. If that string contains a carefully crafted invalid UTF-8 sequence — even a single malformed byte like 0xFF or a truncated multi-byte sequence like 0xC0 — and it reaches a normalization call backed by norm.Iter, the iterator stalls.

Concrete attack scenario:

1. An attacker crafts a CS2 demo file with a player name field containing the byte sequence "\xC0" (an incomplete two-byte UTF-8 sequence).
2. The parser reads the name and passes it through a normalization step that uses norm.Iter internally.
3. norm.Iter attempts to process "\xC0", cannot decode a valid rune, fails to advance its internal offset, and loops back to the same byte.
4. The loop never terminates. The goroutine is stuck. The parser hangs.
5. If the parser is called repeatedly (e.g., in a web service that accepts demo uploads), the attacker can exhaust all available goroutines or CPU cores with a handful of requests.

This is a zero-cost denial of service: no authentication required, no memory corruption, just one bad byte.

The Fix

Upgrading golang.org/x/text from v0.34.0 to v0.39.0

The fix is a one-line change in parser/go.mod:

# parser/go.mod
-   golang.org/x/text v0.34.0 // indirect
+   golang.org/x/text v0.39.0 // indirect

Version 0.39.0 of golang.org/x/text corrects the loop termination logic inside norm.Iter. When the iterator encounters a byte sequence that cannot be decoded as valid UTF-8, it now advances past the offending byte(s) rather than stalling. The exit condition becomes reachable for all inputs, including malformed ones.

The corresponding parser/go.sum update replaces the old checksum entries:

# parser/go.sum (relevant lines)
-github.com/markus-wa/demoinfocs-golang/v5 v5.1.2 h1:YbC23degEUIini8Qe051wDgLM47AqHPwBKeHNPApyxw=
-github.com/markus-wa/demoinfocs-golang/v5 v5.1.2/go.mod h1:cnrd9QDLk2XroPtujR46xAKGEROHxEZgEw9Wy0Pido8=
 github.com/markus-wa/demoinfocs-golang/v5 v5.2.0 h1:hvSXyE9AUvqO4t25a9bqyMIvcwM/Wx9jO/7gPejTSkE=
 github.com/markus-wa/demoinfocs-golang/v5 v5.2.0/go.mod h1:JG2eu06s72JijIJDR7wnCSqgLtuOjhHQMtT8piem0Lw=

The go.sum changes also reflect a bump in the markus-wa/demoinfocs-golang dependency from v5.1.2 to v5.2.0, which itself likely carries the updated golang.org/x/text transitively. Both files must be updated together — go.mod declares the version constraint, and go.sum records the cryptographic hashes that Go's module system uses to verify download integrity.

Why two files?
go.mod is the human-readable version manifest. go.sum is the tamper-evident lock file. Updating one without the other would either fail go mod verify or leave the build using a cached (vulnerable) version of the module.

Key Takeaways

• norm.Iter in golang.org/x/text < v0.39.0 cannot safely process invalid UTF-8 — a single malformed byte is enough to trigger an infinite loop and hang the parser process.
• The fix in parser/go.mod is a one-line bump from v0.34.0 to v0.39.0, but both go.mod and go.sum must be updated together to maintain module integrity.
• Indirect dependencies in Go are not insulated from exploitation — golang.org/x/text is listed as // indirect, yet its bug directly affects the parser's availability.
• utf8.Valid() is a cheap, effective guard to add at any input boundary before text normalization, providing defense-in-depth even against future normalization bugs.
• Trivy's static analysis of go.mod identified this vulnerability without needing to trace the full call graph — version-based scanning is fast and catches issues like this before they reach production.

How Orbis AppSec Detected This

• Source: User-influenced input (player names, string fields) read from demo files processed by the parser component.
• Sink: norm.Iter inside golang.org/x/text v0.34.0, reached transitively through the parser's text-processing pipeline — specifically the normalization functions exposed by golang.org/x/text/unicode/norm.
• Missing control: No UTF-8 validity check before input reaches the normalization iterator; no version constraint preventing the use of the vulnerable golang.org/x/text release.
• CWE: CWE-835 — Loop with Unreachable Exit Condition.
• Fix: golang.org/x/text was upgraded from v0.34.0 to v0.39.0 in parser/go.mod and parser/go.sum, replacing the vulnerable iterator implementation with a version that correctly handles invalid UTF-8 byte sequences.

Orbis AppSec automatically detected this vulnerability and opened a pull request with the fix. Try Orbis AppSec on your repositories to find and fix issues like this automatically.

Conclusion

CVE-2026-56852 is a reminder that Unicode normalization — something most developers treat as a solved problem — has sharp edges when the input isn't clean. The norm.Iter infinite loop is subtle: it doesn't crash, doesn't corrupt memory, and doesn't log an error. It just stops making progress, silently consuming resources until the process is killed or the service times out.

The fix is straightforward: upgrade golang.org/x/text to v0.39.0. But the broader lesson is about defense in depth — validate UTF-8 at your input boundaries, scan your dependency tree regularly, and treat // indirect dependencies with the same security scrutiny as your own code. A one-line version bump in go.mod is all it took to close this door.