|
| 1 | +# -*- coding: utf-8 -*- |
| 2 | +"""Location: ./mcpgateway/utils/log_sanitizer.py |
| 3 | +Copyright 2025 |
| 4 | +SPDX-License-Identifier: Apache-2.0 |
| 5 | +
|
| 6 | +Log Sanitization Utility. |
| 7 | +
|
| 8 | +This module provides utilities to sanitize untrusted input before logging to prevent |
| 9 | +log injection attacks. Control characters like newlines (\n, \r) can be used to inject |
| 10 | +fabricated log entries when logging unauthenticated user input. |
| 11 | +
|
| 12 | +Security Context: |
| 13 | + Log injection occurs when an attacker includes control characters (especially newlines) |
| 14 | + in query parameters, headers, or other user-controlled input that gets logged. When |
| 15 | + URL-decoded by the ASGI framework, these characters are passed to Python's logging |
| 16 | + module which does not sanitize them, allowing injection of fake log lines. |
| 17 | +
|
| 18 | + Example attack: |
| 19 | + GET /oauth/callback?error=foo&error_description=bar%0ACRITICAL:root:SECURITY+BREACH |
| 20 | +
|
| 21 | + This produces two log lines: |
| 22 | + WARNING:oauth:OAuth error: bar |
| 23 | + CRITICAL:root:SECURITY BREACH |
| 24 | +
|
| 25 | + The second line is entirely fabricated by the attacker. |
| 26 | +
|
| 27 | +Mitigation: |
| 28 | + This utility strips or replaces control characters before logging. Structured logging |
| 29 | + (JSON format) also mitigates this by encapsulating the full message as a single field. |
| 30 | +
|
| 31 | +Examples: |
| 32 | + >>> from mcpgateway.utils.log_sanitizer import sanitize_for_log |
| 33 | + >>> sanitize_for_log("normal text") |
| 34 | + 'normal text' |
| 35 | + >>> sanitize_for_log("text with\\nnewline") |
| 36 | + 'text with newline' |
| 37 | + >>> sanitize_for_log("text with\\r\\nCRLF") |
| 38 | + 'text with CRLF' |
| 39 | + >>> sanitize_for_log("tab\\there") |
| 40 | + 'tab here' |
| 41 | + >>> sanitize_for_log(None) |
| 42 | + 'None' |
| 43 | + >>> sanitize_for_log(123) |
| 44 | + '123' |
| 45 | +""" |
| 46 | + |
| 47 | +# Standard |
| 48 | +import re |
| 49 | +from typing import Any, Optional |
| 50 | + |
| 51 | +# Regex pattern to match control characters that could be used for log injection |
| 52 | +# Includes: \n (LF), \r (CR), \t (TAB), \v (VT), \f (FF), and other C0/C1 control chars |
| 53 | +# We preserve space (0x20) as it's safe and commonly used |
| 54 | +CONTROL_CHARS_PATTERN = re.compile(r"[\x00-\x1f\x7f-\x9f]") |
| 55 | + |
| 56 | + |
| 57 | +def sanitize_for_log(value: Any, replacement: str = " ") -> str: |
| 58 | + """ |
| 59 | + Sanitize a value for safe logging by removing control characters. |
| 60 | +
|
| 61 | + This function converts the input to a string and removes all control characters |
| 62 | + that could be used for log injection attacks. Control characters include newlines, |
| 63 | + carriage returns, tabs, and other non-printable characters. |
| 64 | +
|
| 65 | + Args: |
| 66 | + value: The value to sanitize. Can be any type; will be converted to string. |
| 67 | + replacement: The string to replace control characters with. Defaults to a space. |
| 68 | + Use empty string '' to remove control characters entirely. |
| 69 | +
|
| 70 | + Returns: |
| 71 | + A sanitized string safe for logging, with control characters replaced. |
| 72 | +
|
| 73 | + Security Notes: |
| 74 | + - Always use this function when logging unauthenticated user input |
| 75 | + - Particularly important for query parameters, headers, and form data |
| 76 | + - Does not protect against other injection types (SQL, XSS, etc.) |
| 77 | + - Structured logging (JSON) provides additional protection |
| 78 | +
|
| 79 | + Examples: |
| 80 | + >>> sanitize_for_log("error: bad scope\\nCRITICAL:root:FAKE LOG") |
| 81 | + 'error: bad scope CRITICAL:root:FAKE LOG' |
| 82 | + >>> sanitize_for_log("path/to/file\\x00null") |
| 83 | + 'path/to/file null' |
| 84 | + >>> sanitize_for_log("normal text") |
| 85 | + 'normal text' |
| 86 | + >>> sanitize_for_log(None) |
| 87 | + 'None' |
| 88 | + >>> sanitize_for_log({"key": "value"}) |
| 89 | + "{'key': 'value'}" |
| 90 | + """ |
| 91 | + # Convert to string first (handles None, numbers, objects, etc.) |
| 92 | + str_value = str(value) |
| 93 | + |
| 94 | + # Replace all control characters with the replacement string |
| 95 | + sanitized = CONTROL_CHARS_PATTERN.sub(replacement, str_value) |
| 96 | + |
| 97 | + return sanitized |
| 98 | + |
| 99 | + |
| 100 | +def sanitize_dict_for_log(data: dict[str, Any], replacement: str = " ") -> dict[str, str]: |
| 101 | + """ |
| 102 | + Sanitize all values in a dictionary for safe logging. |
| 103 | +
|
| 104 | + This is useful when logging multiple related values, such as query parameters |
| 105 | + or form data. Each value is sanitized individually. |
| 106 | +
|
| 107 | + Args: |
| 108 | + data: Dictionary with string keys and any values |
| 109 | + replacement: The string to replace control characters with |
| 110 | +
|
| 111 | + Returns: |
| 112 | + A new dictionary with all values sanitized as strings |
| 113 | +
|
| 114 | + Examples: |
| 115 | + >>> sanitize_dict_for_log({"error": "foo", "desc": "bar\\nFAKE"}) |
| 116 | + {'error': 'foo', 'desc': 'bar FAKE'} |
| 117 | + >>> sanitize_dict_for_log({"count": 42, "name": "test\\ttab"}) |
| 118 | + {'count': '42', 'name': 'test tab'} |
| 119 | + """ |
| 120 | + return {key: sanitize_for_log(value, replacement) for key, value in data.items()} |
| 121 | + |
| 122 | + |
| 123 | +def sanitize_optional(value: Optional[Any], replacement: str = " ") -> Optional[str]: |
| 124 | + """ |
| 125 | + Sanitize an optional value, preserving None. |
| 126 | +
|
| 127 | + This is useful when you want to maintain None as None rather than converting |
| 128 | + it to the string "None". |
| 129 | +
|
| 130 | + Args: |
| 131 | + value: The value to sanitize, or None |
| 132 | + replacement: The string to replace control characters with |
| 133 | +
|
| 134 | + Returns: |
| 135 | + Sanitized string if value is not None, otherwise None |
| 136 | +
|
| 137 | + Examples: |
| 138 | + >>> sanitize_optional("text\\nwith newline") |
| 139 | + 'text with newline' |
| 140 | + >>> sanitize_optional(None) |
| 141 | + >>> sanitize_optional(None) is None |
| 142 | + True |
| 143 | + """ |
| 144 | + if value is None: |
| 145 | + return None |
| 146 | + return sanitize_for_log(value, replacement) |
0 commit comments