Windows Defender flags v0.35.0 autoshop-gui.exe as Trojan:Win32/Wacatac.B!ml

[gerialkot]

Hi,

Microsoft Defender on Windows 11 immediately quarantines the prebuilt autoshop-gui.exe from the v0.35.0 release as:

Trojan:Win32/Wacatac.B!ml

I downloaded the GUI executable from the GitHub release. Defender quarantined it before I could verify the local file hash. The README lists the expected SHA-256 as:

b58b7be4b5e83b3f70e33b8023aba3bcfb985a1762d6eed7e39ef1deafb9a492

I realize !ml can be a heuristic/false-positive detection, so I'm not assuming the binary is malicious. However, since the Windows release binary appears to be built locally rather than by a public Windows GitHub Actions workflow, it would be useful to know:

Can you reproduce the same Defender detection on the published binary or on a clean local build from the v0.35.0 tag?
Has the release binary been checked with Microsoft Defender / VirusTotal or submitted to Microsoft as a false positive?
Would you consider building the Windows release artifacts in GitHub Actions so the binary provenance is independently visible/reproducible?

Thanks — the project looks very interesting, but Defender's Trojan alert makes me hesitant to whitelist/run the prebuilt executable without clarification.

[skymanbp]

Hi,

Microsoft Defender on Windows 11 immediately quarantines the prebuilt autoshop-gui.exe from the v0.35.0 release as:

Trojan:Win32/Wacatac.B!ml

I downloaded the GUI executable from the GitHub release. Defender quarantined it before I could verify the local file hash. The README lists the expected SHA-256 as:

b58b7be4b5e83b3f70e33b8023aba3bcfb985a1762d6eed7e39ef1deafb9a492

I realize !ml can be a heuristic/false-positive detection, so I'm not assuming the binary is malicious. However, since the Windows release binary appears to be built locally rather than by a public Windows GitHub Actions workflow, it would be useful to know:

Can you reproduce the same Defender detection on the published binary or on a clean local build from the v0.35.0 tag? Has the release binary been checked with Microsoft Defender / VirusTotal or submitted to Microsoft as a false positive? Would you consider building the Windows release artifacts in GitHub Actions so the binary provenance is independently visible/reproducible?

Thanks — the project looks very interesting, but Defender's Trojan alert makes me hesitant to whitelist/run the prebuilt executable without clarification.

Got it. I'm looking into it right now. Sorry for the late reply.

[skymanbp]

@gerialkot Thanks for the report, and for checking the hash angle instead of assuming the worst.

What I've done:

• Re-downloaded autoshop-gui.exe from the v0.35.0 release: SHA-256 matches the README (b58b7be4b5e83b3f70e33b8023aba3bcfb985a1762d6eed7e39ef1deafb9a492), so it's the exact binary built from the tag, not a tampered asset.
• Ran that exact file through VirusTotal: 0/69, Microsoft's own engine included: https://www.virustotal.com/gui/file/b58b7be4b5e83b3f70e33b8023aba3bcfb985a1762d6eed7e39ef1deafb9a492
• Submitted it to Microsoft as a false positive (submission ID cd681984-ca36-43f8-a7dd-5ad064c994b8). I'll post their verdict here when it comes back.

Can't reproduce the quarantine on my dev box (real-time protection is off there for build performance), but I don't doubt it that Wacatac.B!ml is Defender's ML heuristic, and unsigned Rust executables trip it all the time.

On provenance: fair point. Releases are built locally today; I want to move the Windows release builds to GitHub Actions for the next release so the artifacts are reproducible from the tag.

Meanwhile, v1.0.0 (2026-08-24) supersedes v0.35.0, with SHA-256 for all four assets in the release notes. Whatever you run, verify the hash first.

Leaving this open until Microsoft's verdict lands.

[skymanbp]

@gerialkot Windows Defender Security Intelligence has completed the case and claimed no mw detected.

[skymanbp]

v1.1.0 is now built entirely by GitHub Actions on public runners: every asset on the
release page comes from the
tagged workflow run (33353741532),
with the exact toolchain in the run log and each file's SHA-256 in checksums.txt.

That closes the commitment made in this thread after the Defender false positive
(Microsoft submission cd681984, later cleared 0/69 on VirusTotal): a binary anyone
can rebuild from a public tag is a claim a reporter can verify, and a binary built on
one laptop is not. If Defender ever flags an asset again, the Actions run for its tag
is the reference to submit.