The vulnerability allows homograph attacks due to improper handling of Unicode characters with char codes greater than 255 during the decoding process. The provided commit 831716afd2d6e9f1c1e28cdf84ccb1eb17692e43 patches this issue. The patch modifies the decode method (which is part of the object returned by the base factory function) in three different module formats (cjs, esm, ts). In each case, the core logic change is the addition of a check to ensure that the character code obtained from the input string is not greater than 255 before using it to access the BASE_MAP. If charCode > 255, the function now returns, effectively treating such characters as invalid for the given base encoding. Before this change, character codes greater than 255 could be used to index BASE_MAP, leading to undefined behavior or incorrect decoding, which could be exploited by using Unicode characters that look similar to valid base characters but have different char codes.