Team-2-Pre-qualification-Microsoft-Hack/naija-shield-backend

# NaijaShield Backend **Real-time fraud detection and intelligence infrastructure for Nigerian telecommunications.** NaijaShield intercepts phone calls and SMS messages at the network level, classifies them with AI, and delivers instant warnings — without requiring an app on the user's device. The backend comprises two cooperating services: a .NET Core API that orchestrates the full threat pipeline and enterprise intelligence layer, and a Python FastAPI sidecar that handles voice transcription and deepfake analysis. --- ## Overview | Service | Runtime | Role | |---|---|---| | **Core API** | ASP.NET Core (.NET 9, C#) | Threat pipeline, persistence, fraud ring intelligence, authentication, SignalR push | | **AI Voice Sidecar** | Python 3.10+ (FastAPI) | Speech-to-text (Azure Speech), deepfake probability scoring | **Live deployment:** `api-naijashield-dev-a5ggd0e…` --- ## Key Features **Threat Detection** - Real-time classification of SMS and voice calls using Azure OpenAI (GPT-5.4 mini) via Microsoft Semantic Kernel - Supports five Nigerian languages: English, Pidgin, Yoruba, Hausa, and Igbo — including code-switching - Covers the major attack typologies: OTP phishing, bank impersonation, USSD tricks, vishing, and family social engineering - Risk scoring (0–100) with deterministic status thresholds: Blocked ≥ 85, Monitoring ≥ 50, Allowed .azurecomm.net", "AT-Username": "sandbox", "AT-API-Key": "...", "AzureOpenAI": { "Endpoint": "https:// .openai.azure.com/", "DeploymentName": "gpt-4o" }, "Jwt": { "Secret": "at-least-32-character-local-dev-secret", "Issuer": "NaijaShield", "Audience": "NaijaShieldClient", "AccessTokenExpiryMinutes": 60, "RefreshTokenExpiryDays": 7 }, "Jambonz": { "AccountId": "...", "ApiBaseUrl": "your-jambonz-instance", "ApiToken": "...", "ServerUrl": "your-backend-url" } } ``` ### 3. Start the Core API ```bash dotnet restore dotnet run ``` The API starts on `localhost …