Chefvolkan/aegisai-android-malware-ml

AegisAI Android Malware ML

Docker Space for the AegisAI backend Android malware model endpoint.

Runtime endpoints:

• —GET /health
• —GET /models/android-malware/status
• —POST /models/android-malware/analyze
• —GET /models/advanced/status
• —POST /models/advanced/network/analyze
• —POST /models/advanced/apk-graph/analyze

Set this Space secret before exposing it to the Android app:

• —AEGISAI_AUTH_TOKEN

AEGISAI_REQUIRE_AUTH defaults to true; protected endpoints fail closed if the token is missing.

The Docker image downloads Hachirou18/NyerAndroidMalware from Hugging Face and verifies the expected SHA-256 before serving inference. The advanced APK graph endpoint uses bundled backend model artifacts under backend_models/ and statically extracts DEX invoke-reference graphs; it does not execute APKs.