Existing open-source tools rely heavily on signatures, which are easily bypassed by packing or obfuscation. I built SemanticsAV to detect the structural logic of malware rather than just matching hashes.

Key features:

1. AI-Native: Detects evasive threats based on architectural patterns. We are starting with PE and ELF formats, with plans to expand coverage to all malicious execution vectors.

2. Privacy-First: The engine has zero network capabilities. It runs entirely locally on your CPU.

3. Constant-Time Scanning: Unlike traditional AV, performance does not degrade as the number of known threats increases.

4. Free: Free for both personal and commercial use.

Note on Open Source: The CLI is open source (MIT), but the core detection engine is provided as a closed-source binary. I made this trade-off to protect the IP, but I ensured privacy by removing all networking features from the binary. You can verify this via the CLI code.

I wrote a detailed write-up on the architecture and benchmarks against ClamAV here: https://semanticsav.hashnode.dev/its-time-to-end-the-era-of-...

Feedback is welcome!