PYSEC-2026-1057

### Impact _What kind of vulnerability is it? Who is impacted?_

Anyone parsing JSON from an untrusted source is vulnerable.

JSON strings that contain escaped surrogate characters not part of a proper surrogate pair were decoded incorrectly. Besides corrupting strings, this allowed for potential key confusion and value overwriting in dictionaries.

Examples:

```python # An unpaired high surrogate character is ignored. >>> ujson.loads(r'"\uD800"') '' >>> ujson.loads(r'"\uD800hello"') 'hello'

# An unpaired low surrogate character is preserved. >>> ujson.loads(r'"\uDC00"') '\udc00'

# A pair of surrogates with additional non surrogate characters pair up in spite of being invalid. >>> ujson.loads(r'"\uD800foo bar\uDC00"') 'foo bar𐀀' ```

### Patches _Has the problem been patched? What versions should users upgrade to?_

Users should upgrade to UltraJSON 5.4.0.

From version 5.4.0, UltraJSON decodes lone surrogates in the same way as the standard library's `json` module does, preserving them in the parsed output:

```python3 >>> ujson.loads(r'"\uD800"') '\ud800' >>> ujson.loads(r'"\uD800hello"') '\ud800hello' >>> ujson.loads(r'"\uDC00"') '\udc00' >>> ujson.loads(r'"\uD800foo bar\uDC00"') '\ud800foo bar\udc00' ```

### Workarounds _Is there a way for users to fix or remediate the vulnerability without upgrading?_

Short of switching to an entirely different JSON library, there are no safe alternatives to upgrading.

### For more information If you have any questions or comments about this advisory: * Open an issue in [UltraJSON](http://github.com/ultrajson/ultrajson/issues)