SpamGuard: AI Spam Filterv1.1

What changed

• New featureBrand-impersonation detector for 35+ brands
• Localization33-language keyword packs with on-device detection
• New featureMechanisms to defeat evasion techniques like leetspeak and punycode
• New featureSender burst-rate tracker to flag campaign-frequency messaging
• New featureUser feedback loop to adjust future spam verdicts
• UXGranular controls for resetting filter memory
• UXOpt-in activity previews to enhance privacy
• PerformanceSwift actor-based data layer for safer concurrency

Original release notes

Major filter rebuild — SpamGuard now defends against the latest SMS scam tricks in 33 languages.

NEW
• Brand-impersonation detector catches "Apple ID alert at apple-id-secure.xyz" patterns across 35+ brands.
• 33-language keyword packs with on-device language detection.
• Defeats evasion: leetspeak ("fr33b3t"), zero-width characters, punycode lookalikes ("paypa1.com"), RTL-override URLs, percent-encoded TLDs, fragmented keywords ("f.r.e.e.b.e.t").
• Sender burst-rate tracker flags unknown senders messaging at campaign frequency.
• Mark as spam / not spam — feedback adjusts future verdicts automatically.

FASTER & SAFER
• Atomic counter file for cross-process accuracy.
• Reset Filter Memory: granular controls for activity, feedback, rate, and statistics.
• Activity preview is now opt-in — only senders and verdicts are kept by default.
• Hashed sender identifiers never leave your device.

UNDER THE HOOD
• Swift actor-based data layer for safer concurrency.
• 230 unit and integration tests across 19 locales.
• Privacy manifest updated to declare every data type accurately.