Project Overview
MobiGuard AI is a decentralized AI-powered personal and community security app built for AfroEdge Africa. It protects against burglary, kidnapping, and property invasion threats using advanced edge machine learning running directly on smartphones — with no internet, no CCTV infrastructure, and no expensive monitoring subscriptions required.
The Engineering Challenge
- Communities in Africa face serious security threats — burglary, kidnapping, property invasion — with no reliable internet infrastructure or affordable CCTV monitoring solutions available.
- Standard cloud-based security apps fail entirely when internet is unavailable, leaving users unprotected exactly when they need it most.
- Multi-sensor fusion was required — accelerometer, gyroscope, microphone MFCC, ambient light, barometric pressure — all processed on-device in real-time without draining battery or overheating.
- False alarm prevention was critical: the system had to distinguish genuine intrusions from normal movement patterns like phones being picked up, walked around with, or placed on surfaces.
- Emergency communication needed to work even when both internet and cell networks were unavailable, requiring a mesh communication system between neighbouring devices.
Our Solution
- Built a lightweight TFLite Edge AI model (<20MB) that runs entirely on-device — fusing data from accelerometer, gyroscope, microphone MFCC, ambient light, and barometric pressure for comprehensive intrusion detection.
- Implemented Adaptive Anomaly Scoring with dynamic thresholds (T = μ + kσ) that auto-adapt to each environment, preventing false alarms with posture mode recognition when phones are moved normally.
- Developed 3-hop encrypted Bluetooth Mesh communication that propagates alerts across neighbouring houses even with zero internet — with offline store-forward retaining unconfirmed alerts for up to 7 days.
- Built an Event Classification Engine that categorises detected activity as Minor / Suspicious / Danger / Critical, triggering appropriate escalation responses automatically.
- Implemented Silent Rescue Mode (anti-kidnapping): covert triple-tap or secret gesture activation with zero vibration, silent alert broadcast, encrypted audio snapshots for evidence, and automatic timed triggers.
- Added Multi-Channel Emergency Suite: one-tap PANIC button, hidden long-press gesture trigger, loud deterrent siren alarm, SMS fallback when mesh fails, and multi-path propagation (mesh + SMS + escalation).
Results & Outcomes
- Full intrusion detection with <20MB on-device AI — no cloud dependency whatsoever
- Bluetooth Mesh covering 14+ neighbours per node with 3-hop encrypted flood delivery
- Silent rescue mode operational — covert anti-kidnapping protection with evidence capture
- Adaptive anomaly scoring eliminates false alarms from normal phone handling
- Alerts retained offline for up to 7 days and synced when connectivity restored
- Deployed on iOS (App Store) and Android (Google Play)
App Screenshots
Why AfroEdge Africa Chose FoogleTech Software
FoogleTech Software is a specialist engineering company with over a decade of expertise in AI, embedded systems, IoT, and full-stack software development — serving product teams and enterprises across AI & Machine Learning and beyond. Our engineers don't just write code — they understand the domain, the constraints, and the real-world pressures that ship deadlines create. For AfroEdge Africa, that meant deploying a pre-vetted team with direct experience in Mobile App, Edge AI, TFLite, reducing ramp-up time from months to days and delivering production-quality work from the first sprint.
Every FoogleTech engagement starts with a structured discovery phase, follows a disciplined agile delivery model with daily engineering syncs, and ends with complete documentation handover — so your in-house team owns the outcome. No black boxes, no lock-in.