PhishGuard AI v0.4.0 - Explainable IDN Detection and Reproducible Evaluation

PhishGuard AI v0.4.0 strengthens explainable phishing detection, reproducible evaluation, contributor safety, and command-line portability.

Highlights

• Conservative punycode and Unicode hostname indicators that remain contextual rather than treating all internationalized domains as malicious.
• A documented 14-sample public-safe URL regression benchmark reporting TP=7, TN=7, FP=0, and FN=0. These are fixture metrics, not population-level accuracy claims.
• Plain ASCII CLI output contributed by @BeauDevCode for legacy terminals and CI logs.
• Repository policy checks that reject unexpected binaries, executable content, symlinks, risky workflow patterns, and unreviewed runtime dependencies.
• Standards-based Python packaging with tests across Python 3.10-3.13, CodeQL, checksums, and signed build-provenance attestation.

Verification

• 28 unit tests
• Python 3.10, 3.11, 3.12, and 3.13
• Package build and isolated wheel installation
• CodeQL and repository security policy

Contributors

Thank you to @BeauDevCode for the ASCII output contribution and to everyone reviewing and testing the project.

Full changelog: v0.3.0...v0.4.0