nano-id reduced entropy due to inadequate character set usage

Description

Affected versions of the nano-id crate incorrectly generated IDs using a reduced character set in the nano_id::base62 and nano_id::base58 functions. Specifically, the base62 function used a character set of 32 symbols instead of the intended 62 symbols, and the base58 function used a character set of 16 symbols instead of the intended 58 symbols. Additionally, the nano_id::gen macro is also affected when a custom character set that is not a power of 2 in size is specified.

It should be noted that nano_id::base64 is not affected by this vulnerability.

Impact

This can result in a significant reduction in entropy, making the generated IDs predictable and vulnerable to brute-force attacks when the IDs are used in security-sensitive contexts such as session tokens or unique identifiers.

Patches

The flaws were corrected in commit a9022772b2f1ce38929b5b81eccc670ac9d3ab23 by updating the the nano_id::gen macro to use all specified characters correctly.

PoC

use std::collections::BTreeSet;

fn main() {
    test_base58();
    test_base62();
}

fn test_base58() {
    let mut produced_symbols = BTreeSet::new();

    for _ in 0..100_000 {
id = "RUSTSEC-2024-0343"
        for c in id.chars() {
            produced_symbols.insert(c);
        }
    }

    println!(
        "{} symbols generated from nano_id::base58",
        produced_symbols.len()
    );
}

fn test_base62() {
    let mut produced_symbols = BTreeSet::new();

    for _ in 0..100_000 {
id = "RUSTSEC-2024-0343"
        for c in id.chars() {
            produced_symbols.insert(c);
        }
    }

    println!(
        "{} symbols generated from nano_id::base62",
        produced_symbols.len()
    );
}