CWE-167 describes input that includes an additional, unexpected special element, which the product does not handle correctly. The extra character is meant to confuse a filter while a browser or parser still reads the content the way the attacker wants.
It matters most for anyone writing filters that look for known bad patterns in web input.
How it happens
A filter looks for a literal pattern, such as the start of a script tag. Adding one more special character in front or inside changes what the filter sees, but a forgiving downstream parser skips the extra character and still finds the tag.
import re
# Vulnerable: removes a literal pattern and trusts the result
def sanitize(html: str) -> str:
return re.sub(r"(?i)script", "", html)
Variants of this filter fall to inputs where a leftover or extra character makes the pattern disappear and reappear.
The observed examples are all of this kind. CVE-2000-0116 is an extra less than sign in front of a script tag that bypasses cross-site scripting prevention. CVE-2001-1157 is another extra less than sign in front of a script tag. CVE-2002-2086 involves the string script with an opening bracket and is probably a cleansing error, according to MITRE.
What an attacker gains
The record lists integrity impact as unexpected state. In these cases it means running script in a victim's browser through a filter that was supposed to stop it.
How to find it
No detection methods are listed. Use these checks.
- Send inputs with an extra opening bracket, doubled brackets and stray characters around tags to every field that is later displayed.
- Review sanitizers that remove or replace patterns by text matching.
- Pass filter output through the filter again. If it changes, the filter is not stable.
How to fix it
- Do not try to clean markup by removing strings. Encode on output for the context where the value lands, so special characters render as text.
from html import escape
safe = escape(user_value) # extra brackets become harmless text
- Anticipate that extra special elements will be injected, and use an allowlist for what is acceptable.
- If you must accept some markup, use a maintained sanitizer built on a real parser and keep it current.
- Reject input that does not conform instead of repairing it.
Quick check
Why does an extra bracket bypass a filter?
The filter looks for an exact pattern. The extra character breaks that match, but a lenient parser downstream ignores it and still sees the tag.
