Repository navigation
Security and Privacy
Struis ICT edited this page Sep 2, 2026
·
2 revisions
InSearch reads files you point it at — often logs and documents from other people or systems — so it is built to treat every input as untrusted.
- Offline by default. InSearch never sends your queries, results, files, or any telemetry — those always stay on your machine.
-
The one optional network feature is the opt-in update check. It's off until you say yes to a one-time prompt (and you can turn it off any time in ℹ About). When on, it occasionally asks
github.comover HTTPS whether a newer release version exists — it downloads and installs nothing, and sends nothing about you or your searches. It uses the operating system's own HTTPS stack (WinHTTP/Schannel), so no third-party HTTP/TLS library is bundled. See the FAQ. - No indexing service. There's no background daemon; searches run on demand over the folders you choose.
-
Parser panics are isolated. The PDF/Office parsers run inside a guard (
catch_unwind); a malformed or malicious document that would crash a parser is simply skipped, not fatal to the scan. - Size caps. Documents and files are size-limited before decoding, to bound memory against huge or decompression-amplified ("zip bomb") inputs.
-
No XML external entities. The
docxreader is a non-validating pull parser that does not resolve DTDs or external entities, so XXE / entity-expansion attacks don't apply. - Streaming search. Plain-text search streams files rather than loading them whole, with a bounded results channel so a huge match set can't exhaust memory.
The optional "Search with InSearch" context menu is registered per-user under HKCU (no administrator rights), and only when you enable it in Settings. It launches InSearch with the selected path — nothing more.
Found a security issue? Please open an issue at https://github.com/StruisICT/InSearch/issues (or contact Struis ICT for anything sensitive).
© Struis ICT — https://struisict.com · Buy me a coffee ☕