Simple Replace for Project Name and other PII

[hawkerm]

I was just about to go write a tool like this myself; though found this via NuGet - is there a way to install via WinGet too? (Could be a nice add-on option in https://msbuildlog.com/ too.)

Was trying to understand what needs to be implemented to add-on to this tool though? The structure of this repo seems fairly complex, though looks to be extensible?

It could also be handy to have an option to ensure all replacements are of equal length, as path length is important at times for diagnosing certain issues.

Anyway, I was hoping to find options for:

• Replacing project names (or parts of them at least)
• Replacing Username in Home Directory paths (think that's part of the add-on, but unclear how that works from readme)
• Replacing GitHub urls from repo connections

Thanks!

[JanKrivanek]

Thank you for your interest and for the thoughts about features!

First few quick notes for short term:

• Binlog Viewer and binlogtool both already exposes redacting exposed from DotUtils.MsBuild.SensitiveDataDetector :-) (it was added as part of: Sensitive data redacting PoC KirillOsenkov/MSBuildStructuredLog#711). I definitely owe some documentation update for that.
• Username detecting and replacing works out of the box - it tries to detect username based on curent environemnt and based on regexes detecting Home dir in the binlog paths.
• The other mentioned replacement would need to be explicitly passed as 'secret' to the above mentioned toolings - those will then replace it either with fixed length asterixes mask or a fixed length pattern making it possible to distinguish the different sensitive data (e.f. 'REDACTED_PWD01', 'REDACTED_USERNAME' and similar) - the choice between those two is IIRC possible only in binlog viewer (and the CLI tool exposed by this repository)
• It could also be handy to have an option to ensure all replacements are of equal length - this should already be a case today. Or do you mean that the replacement should be of a same length as the original substring being replaced? I didn't want to make this possible - as revealing length of secret might expose a risk (for short credentials). But if this is understood and still needed - it can be added.

I'm wondering if above can unblock you?

For the proper unblocking I should better user document the CLI tool in this repo, make it easier to plug-in custom redacting and possibly add more configurability for replacements - I hope to make some improvements early next year, as I was planning to add more auto-redaction patterns as well.

[hawkerm]

• Thanks @JanKrivanek, I totally missed it in the binlog viewer tool. I guess I was looking for a more 'Find/Replace' style search where I could individually look for information I wanted to replace, didn't expect to look in the File menu for that. Will try it out next time.
• Quick note: Username could be different depending on where the log comes from, may not have been my machine, but want to redact it before passing it on. So having a general substitution option would be great, and that'd unblock the project name change.
• Yes, I was referring to the replaced string being the same length, so if my username was 'hawkerm' then the replaced string would be the same length like 'user123'. We've had issues hit by MAX_PATH in logs before, so having the actual file path be the correct length could be important in edge-case scenarios.

Yeah, I think just some notes on the tools architecture, and an example of implementing a custom pattern could be a big help. It looks like patterns can be added outside of recompilation, or did I misunderstand the extra nuget package? Though just being exposed in binlog viewer is great to know. Certainly, easier than using the hex editor, though did find that extension pretty cool for VS Code. 😋

[JanKrivanek]

The username detection works on binlogs from other envs as well (it recognizes the home dir pattern https://github.com/dotutils/MSBuild.BinlogRedactor/blob/main/src/DotUtils.MsBuild.SensitiveDataDetector/UsernameDetector.cs#L19-L20 and extracts the username from it).

The length of replacement and MAX_PATH - good point - noted

Adding custom patterns is currently regretfully a recompile thing. I planned pluggability - but didn't implement that yet (but I want to add it soon). So as of now, just custom literal strings can be specified at run time.

I'll update here once the improvements are added