Incorrect Parsing of Numbers with Different Radices

The product parses numeric input assuming base 10 (decimal) values, but it does not account for inputs that use a different base number (radix).


Description

Frequently, a numeric input that begins with "0" is treated as octal, or "0x" causes it to be treated as hexadecimal, e.g. by the inet_addr() function. For example, "023" (octal) is 35 decimal, or "0x31" is 49 decimal. Other bases may be used as well. If the developer assumes decimal-only inputs, the code could produce incorrect numbers when the inputs are parsed using a different base. This can result in unexpected and/or dangerous behavior. For example, a "0127.0.0.1" IP address is parsed as octal due to the leading "0", whose numeric value would be the same as 87.0.0.1 (decimal), where the developer likely expected to use 127.0.0.1.

The consequences vary depending on the surrounding code in which this weakness occurs, but they can include bypassing network-based access control using unexpected IP addresses or netmasks, or causing apparently-symbolic identifiers to be processed as if they are numbers. In web applications, this can enable bypassing of SSRF restrictions.

Demonstrations

The following examples help to illustrate the nature of this weakness and describe methods or techniques which can be used to mitigate the risk.

Note that the examples here are by no means exhaustive and any given weakness may have many subtle varieties, each of which may require different detection methods or runtime controls.

Example One

The below demonstrative example uses an IP validator that splits up an IP address by octet, tests to ensure each octet can be casted into an integer, and then returns the original IP address if no exceptions are raised. This validated IP address is then tested using the "ping" command.

import subprocess

def validate_ip(ip: str):

  split_ip = ip.split('.')
  if len(split_ip) > 4 or len(split_ip) == 0:

    raise ValueError("Invalid IP length")


  for octet in split_ip:

    try:

      int(octet, 10)

    except ValueError as e:

      raise ValueError(f"Cannot convert IP octet to int - {e}")



  # Returns original IP after ensuring no exceptions are raised
  return ip


def run_ping(ip: str):

  validated = validate_ip(ip)
  # The ping command treats zero-prepended IP addresses as octal
  result = subprocess.call(["ping", validated])
  print(result)

If run_ping() were to be called with one or more zero-prepended octets, validate_ip() will succeed as zero-prepended numerical strings can be interpreted as decimal by a cast ("012" would cast to 12). However, as the original IP with the prepended zeroes is returned rather than the casted IP, it will be used in the call to the ping command. Ping DOES check and support octal-based IP octets, so the IP reached via ping may be different than the IP assumed by the validator. For example, ping would considered "0127.0.0.1" the same as "87.0.0.1".

Example Two

This code uses a regular expression to validate an IP string prior to using it in a call to the "ping" command.

import subprocess
import re

def validate_ip_regex(ip: str):

  ip_validator = re.compile(r"((25[0-5]|(2[0-4]|1\d|[1-9]|)\d)\.?\b){4}")
  if ip_validator.match(ip):

    return ip

  else:

    raise ValueError("IP address does not match valid pattern.")



def run_ping_regex(ip: str):

  validated = validate_ip_regex(ip)
  # The ping command treats zero-prepended IP addresses as octal
  result = subprocess.call(["ping", validated])
  print(result)

Since the regular expression does not have anchors (CWE-777), i.e. is unbounded without ^ or $ characters, then prepending a 0 or 0x to the beginning of the IP address will still result in a matched regex pattern. Since the ping command supports octal and hex prepended IP addresses, it will use the unexpectedly valid IP address (CWE-1389). For example, "0x63.63.63.63" would be considered equivalent to "99.63.63.63". As a result, the attacker could potentially ping systems that the attacker cannot reach directly.

Example Three

Consider the following scenario, inspired by CWE team member Kelly Todd.

Kelly wants to set up monitoring systems for his two cats, who pose very different threats. One cat, Night, tweets embarrassing or critical comments about his owner in ways that could cause reputational damage, so Night's blog needs to be monitored regularly. The other cat, Taki, likes to distract Kelly and his coworkers during business meetings with cute meows, so Kelly monitors Taki's location using a different web site.

Suppose /etc/hosts provides the site info as follows:

taki.example.com 10.1.0.7
night.example.com 010.1.0.8

The entry for night.example.com has a typo "010" in the first octet. When using ping to ensure the servers are up, the leading 0 causes the IP address to be converted using octal. So when Kelly's script attempts to access night.example.com, it inadvertently scans 8.1.0.8 instead of 10.1.0.8 (since "010" in octal is 8 in decimal), and Night is free to send new Tweets without being immediately detected.

See Also

Comprehensive Categorization: Resource Lifecycle Management

Weaknesses in this category are related to resource lifecycle management.

Numeric Errors

Weaknesses in this category are related to improper calculation or conversion of numbers.

Comprehensive CWE Dictionary

This view (slice) covers all the elements in CWE.

Weaknesses Introduced During Implementation

This view (slice) lists weaknesses that can be introduced during implementation.

Weakness Base Elements

This view (slice) displays only weakness base elements.


Common Weakness Enumeration content on this website is copyright of The MITRE Corporation unless otherwise specified. Use of the Common Weakness Enumeration and the associated references on this website are subject to the Terms of Use as specified by The MITRE Corporation.