ewen-lbh/ffcss Late-Unicode normalization vulnerability

### Summary The function `lookupPreprocess()` is meant to apply some transformations to a string by disabling characters in the regex `[-_ .]`. However, due to the use of late Unicode normalization of type NFKD, it is possible to bypass that validation and re-introduce all the characters in the regex `[-_ .]`. ```go // lookupPreprocess applies transformations to s so that it can be compared // to search for something. // For example, it is used by (ThemeStore).Lookup func lookupPreprocess(s string) string { return strings.ToLower(norm.NFKD.String(regexp.MustCompile(`[-_ .]`).ReplaceAllString(s, ""))) } ``` Take the following equivalent Unicode character U+2024 (․). Initially, the `lookupPreprocess()` function would compile the regex and replace the regular dot (.). However, the U+2024 (․) would bypass the `ReplaceAllString()`. When the normalization operation is applied to U+2024 (․), the resulting character will be U+002E (.). Thus, the dot was reintroduced back. ### Impact The `lookupPreprocess()` can be easily bypassed with equivalent Unicode characters like U+FE4D (﹍), which would result in the omitted U+005F (_), for instance. It should be noted here that the variable `s` is user-controlled data coming from [/cmd/ffcss/commands.go#L22-L28](https://github.com/ewen-lbh/ffcss/blob/master/cmd/ffcss/commands.go#L22-L28) the command args. The `lookupPreprocess()` function is only ever used to search for themes loosely (case insensitively, while ignoring dashes, underscores and dots), so the actual security impact is classified as low. ### Remediation A simple fix would be to initially perform the Unicode normalization and then the rest of validations. ### References - https://sim4n6.beehiiv.com/p/unicode-characters-bypass-security-checks

Quoted source text, attributed separately from HOL analysis.