fix: Correctly sanitize input everywhere (#5697)

Best reviewed commit-by-commit; the commit messages explain what is
done.
This commit is contained in:
Hocuri
2024-06-28 14:36:09 +02:00
committed by GitHub
parent 4ed9c04e9b
commit ac198b17bf
11 changed files with 226 additions and 124 deletions

View File

@@ -22,7 +22,8 @@
clippy::bool_assert_comparison,
clippy::manual_split_once,
clippy::format_push_string,
clippy::bool_to_int_with_if
clippy::bool_to_int_with_if,
clippy::manual_range_contains
)]
use std::fmt;
@@ -36,7 +37,6 @@ use once_cell::sync::Lazy;
use regex::Regex;
// TODOs to clean up:
// - Check if sanitizing is done correctly everywhere
// - Apply lints everywhere (https://doc.rust-lang.org/cargo/reference/workspaces.html#the-lints-table)
#[derive(Debug)]
@@ -263,27 +263,27 @@ impl rusqlite::types::ToSql for ContactAddress {
}
}
/// Make the name and address
/// Takes a name and an address and sanitizes them:
/// - Extracts a name from the addr if the addr is in form "Alice <alice@example.org>"
/// - Removes special characters from the name, see [`sanitize_name()`]
/// - Removes the name if it is equal to the address by setting it to ""
pub fn sanitize_name_and_addr(name: &str, addr: &str) -> (String, String) {
static ADDR_WITH_NAME_REGEX: Lazy<Regex> = Lazy::new(|| Regex::new("(.*)<(.*)>").unwrap());
let (name, addr) = if let Some(captures) = ADDR_WITH_NAME_REGEX.captures(addr.as_ref()) {
(
if name.is_empty() {
strip_rtlo_characters(captures.get(1).map_or("", |m| m.as_str()))
captures.get(1).map_or("", |m| m.as_str())
} else {
strip_rtlo_characters(name)
name
},
captures
.get(2)
.map_or("".to_string(), |m| m.as_str().to_string()),
)
} else {
(
strip_rtlo_characters(&normalize_name(name)),
addr.to_string(),
)
(name, addr.to_string())
};
let mut name = normalize_name(&name);
let mut name = sanitize_name(name);
// If the 'display name' is just the address, remove it:
// Otherwise, the contact would sometimes be shown as "alice@example.com (alice@example.com)" (see `get_name_n_addr()`).
@@ -295,31 +295,77 @@ pub fn sanitize_name_and_addr(name: &str, addr: &str) -> (String, String) {
(name, addr)
}
/// Normalize a name.
/// Sanitizes a name.
///
/// - Remove quotes (come from some bad MUA implementations)
/// - Trims the resulting string
///
/// Typically, this function is not needed as it is called implicitly by `Contact::add_address_book`.
pub fn normalize_name(full_name: &str) -> String {
let full_name = full_name.trim();
if full_name.is_empty() {
return full_name.into();
}
/// - Removes newlines and trims the string
/// - Removes quotes (come from some bad MUA implementations)
/// - Removes potentially-malicious bidi characters
pub fn sanitize_name(name: &str) -> String {
let name = sanitize_single_line(name);
match full_name.as_bytes() {
[b'\'', .., b'\''] | [b'\"', .., b'\"'] | [b'<', .., b'>'] => full_name
.get(1..full_name.len() - 1)
match name.as_bytes() {
[b'\'', .., b'\''] | [b'\"', .., b'\"'] | [b'<', .., b'>'] => name
.get(1..name.len() - 1)
.map_or("".to_string(), |s| s.trim().to_string()),
_ => full_name.to_string(),
_ => name.to_string(),
}
}
/// Sanitizes user input
///
/// - Removes newlines and trims the string
/// - Removes potentially-malicious bidi characters
pub fn sanitize_single_line(input: &str) -> String {
sanitize_bidi_characters(input.replace(['\n', '\r'], " ").trim())
}
const RTLO_CHARACTERS: [char; 5] = ['\u{202A}', '\u{202B}', '\u{202C}', '\u{202D}', '\u{202E}'];
/// This method strips all occurrences of the RTLO Unicode character.
/// [Why is this needed](https://github.com/deltachat/deltachat-core-rust/issues/3479)?
pub fn strip_rtlo_characters(input_str: &str) -> String {
input_str.replace(|char| RTLO_CHARACTERS.contains(&char), "")
const ISOLATE_CHARACTERS: [char; 3] = ['\u{2066}', '\u{2067}', '\u{2068}'];
const POP_ISOLATE_CHARACTER: char = '\u{2069}';
/// Some control unicode characters can influence whether adjacent text is shown from
/// left to right or from right to left.
///
/// Since user input is not supposed to influence how adjacent text looks,
/// this function removes some of these characters.
///
/// Also see https://github.com/deltachat/deltachat-core-rust/issues/3479.
pub fn sanitize_bidi_characters(input_str: &str) -> String {
// RTLO_CHARACTERS are apparently rarely used in practice.
// They can impact all following text, so, better remove them all:
let input_str = input_str.replace(|char| RTLO_CHARACTERS.contains(&char), "");
// If the ISOLATE characters are not ended with a POP DIRECTIONAL ISOLATE character,
// we regard the input as potentially malicious and simply remove all ISOLATE characters.
// See https://en.wikipedia.org/wiki/Bidirectional_text#Unicode_bidi_support
// and https://www.w3.org/International/questions/qa-bidi-unicode-controls.en
// for an explanation about ISOLATE characters.
fn isolate_characters_are_valid(input_str: &str) -> bool {
let mut isolate_character_nesting: i32 = 0;
for char in input_str.chars() {
if ISOLATE_CHARACTERS.contains(&char) {
isolate_character_nesting += 1;
} else if char == POP_ISOLATE_CHARACTER {
isolate_character_nesting -= 1;
}
// According to Wikipedia, 125 levels are allowed:
// https://en.wikipedia.org/wiki/Unicode_control_characters
// (although, in practice, we could also significantly lower this number)
if isolate_character_nesting < 0 || isolate_character_nesting > 125 {
return false;
}
}
isolate_character_nesting == 0
}
if isolate_characters_are_valid(&input_str) {
input_str
} else {
input_str.replace(
|char| ISOLATE_CHARACTERS.contains(&char) || POP_ISOLATE_CHARACTER == char,
"",
)
}
}
/// Returns false if addr is an invalid address, otherwise true.
@@ -668,4 +714,64 @@ END:VCARD
assert_eq!(contacts[0].profile_image.as_deref().unwrap(), "/9j/4AAQSkZJRgABAQAAAQABAAD/4gIoSUNDX1BST0ZJTEUAAQEAAAIYAAAAAAQwAABtbnRyUkdCIFhZWiAAAAAAAAAAAAAAAABhY3NwAAAAAAAAAAAAAAAAL8bRuAJYoZUYrI4ZY3VWwxw4Ay28AAGBISScmf/2Q==");
}
}
#[test]
fn test_sanitize_name() {
assert_eq!(&sanitize_name(" hello world "), "hello world");
assert_eq!(&sanitize_name("<"), "<");
assert_eq!(&sanitize_name(">"), ">");
assert_eq!(&sanitize_name("'"), "'");
assert_eq!(&sanitize_name("\""), "\"");
}
#[test]
fn test_sanitize_single_line() {
assert_eq!(sanitize_single_line("Hi\naiae "), "Hi aiae");
assert_eq!(sanitize_single_line("\r\nahte\n\r"), "ahte");
}
#[test]
fn test_sanitize_bidi_characters() {
// Legit inputs:
assert_eq!(
&sanitize_bidi_characters("Tes\u{2067}ting Delta Chat\u{2069}"),
"Tes\u{2067}ting Delta Chat\u{2069}"
);
assert_eq!(
&sanitize_bidi_characters("Tes\u{2067}ting \u{2068} Delta Chat\u{2069}\u{2069}"),
"Tes\u{2067}ting \u{2068} Delta Chat\u{2069}\u{2069}"
);
assert_eq!(
&sanitize_bidi_characters("Tes\u{2067}ting\u{2069} Delta Chat\u{2067}\u{2069}"),
"Tes\u{2067}ting\u{2069} Delta Chat\u{2067}\u{2069}"
);
// Potentially-malicious inputs:
assert_eq!(
&sanitize_bidi_characters("Tes\u{202C}ting Delta Chat"),
"Testing Delta Chat"
);
assert_eq!(
&sanitize_bidi_characters("Testing Delta Chat\u{2069}"),
"Testing Delta Chat"
);
assert_eq!(
&sanitize_bidi_characters("Tes\u{2067}ting Delta Chat"),
"Testing Delta Chat"
);
assert_eq!(
&sanitize_bidi_characters("Tes\u{2069}ting Delta Chat\u{2067}"),
"Testing Delta Chat"
);
assert_eq!(
&sanitize_bidi_characters("Tes\u{2068}ting Delta Chat"),
"Testing Delta Chat"
);
}
}