Skip to content

Suliman8/SkyShield

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

3 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

SkyShield - RF-Based Drone Detection & Threat Classification System

Real-time drone detection, payload analysis, and threat classification using RF signatures, micro-Doppler radar, and machine learning - at 1/100th the cost of commercial solutions.


The Problem

Drones are everywhere. Some are harmless (photography, delivery, agriculture). Some are not (surveillance, smuggling, weaponized attacks). Current counter-drone systems cost $100K - $5M+ and are only available to military and government agencies.

There is no affordable system that can:

  1. Detect drones in the sky via RF signals
  2. Classify them as friendly or hostile in real-time
  3. Detect if a "known safe" drone has been modified to carry a payload
  4. Neutralize hostile drones

SkyShield fills this gap.


What Makes SkyShield Different

Feature Commercial Systems SkyShield
Cost $100K - $5M+ $500 - $3,000
RF Drone Detection Yes Yes
Payload Detection Limited (military only) Yes (micro-Doppler + EMI)
"Familiar Drone" Attack Defense No Yes
Motor EMI Weight Estimation No Yes
Proprietary & Affordable No (expensive only) Yes
Neutralization Options Some 4 methods

5 Novel Research Contributions (Nobody Has These)

  1. "Familiar Drone" Spoofing Defense - Detects when a known consumer drone (e.g., DJI Mini) has been modified with a hostile payload, even though its RF signature says "safe"
  2. Motor EMI Passive Payload Weight Estimation - Uses the same SDR to pick up motor electromagnetic interference and estimate payload weight. Zero additional hardware cost
  3. Multi-Layer Fusion at Under $3,000 - Combines RF + Radar + EMI + Behavior AI at a fraction of commercial cost (proprietary implementation)
  4. Acoustic Gyroscope Disruption for C-UAS - Applying MEMS gyroscope resonance attacks as a drone neutralization method
  5. Combined Detect + Classify + Neutralize - Full pipeline in one affordable system

System Architecture

DETECTION (What's in the sky?)
+-- Layer 1: RF Signature Matching          [$25-200]  - Identify drone model via SDR
+-- Layer 2: Motor EMI Weight Estimation     [FREE]    - Detect payload via motor noise
+-- Layer 3: Micro-Doppler Radar             [$50-300] - Payload shape/weight via FMCW radar
+-- Layer 4: Flight Behavior AI              [FREE]    - Anomaly detection from flight patterns
+-- Layer 5: "Familiar Drone" Defense        [FREE]    - Override safe RF if payload detected

CLASSIFICATION (Is it dangerous?)
+-- Known drone + No payload                = SAFE
+-- Known drone + Payload detected          = HOSTILE (novel)
+-- Unknown drone + No payload              = SUSPICIOUS
+-- Unknown drone + Payload detected        = CRITICAL THREAT

NEUTRALIZATION (What to do about it?)
+-- Option A: GPS Spoofing                   [$100-500] - Deviate to safe landing zone
+-- Option B: Selective RF Jamming           [$200-2K]  - Trigger Return-to-Home
+-- Option C: Protocol Cyber Takeover        [$100-500] - Full control (MAVLink drones)
+-- Option D: Acoustic Gyro Disruption       [$50-300]  - Ultrasonic destabilization

How It Works

Detection Pipeline

[RF Antenna + SDR] --> [Signal Processing] --> [Feature Extraction] --> [ML Classifier] --> [Alert]
     (Hardware)           (FFT / DSP)          (Spectrogram)           (CNN Model)        (Dashboard)
                                                    |
                                                    v
                                          [Micro-Doppler Radar] --> [Payload Detection]
                                               (FMCW 24/77GHz)       (Weight/Shape)
  1. RF Capture - SDR receiver captures drone-to-controller RF signals (2.4 / 5.8 GHz)
  2. Signal Processing - FFT converts raw RF into spectrograms
  3. Feature Extraction - Frequency, modulation, bandwidth, hopping patterns
  4. ML Classification - CNN compares RF fingerprint against known drone database
  5. Payload Detection - Micro-Doppler radar + motor EMI analysis detects added weight
  6. Multi-Layer Fusion - All layers combined for final threat verdict

The "Familiar Drone" Attack (And How We Defeat It)

ATTACK SCENARIO:
  Attacker buys DJI Mini --> Attaches explosive --> Flies toward target
                                                         |
                                                         v
                                              RF says "DJI Mini = SAFE"
                                                         |
                                                         v
                                              Normal systems: NO ALERT (breach!)

SKYSHIELD RESPONSE:
  RF says "DJI Mini"    --> BUT radar says "+650g payload"
                         --> AND motor EMI says "RPM +18% above normal"
                         --> AND behavior AI says "sluggish flight"
                                                         |
                                                         v
                                              OVERRIDE: "PAYLOAD DETECTED - HOSTILE"

Hardware Requirements

Minimum Setup (~$200-500)

Component Example Cost
SDR Receiver RTL-SDR v4 $25-35
Wideband Antenna 2.4/5.8 GHz omnidirectional $20-50
Processing Unit Raspberry Pi 4 (4GB) $50-75
MicroSD + Power 64GB + 5V supply $20-30

Full Setup (~$500-3,000)

Component Example Cost
SDR Receiver HackRF One / USRP B210 $200-2,000
Wideband Antenna Directional + omnidirectional $50-200
FMCW Radar Module TI IWR1443 / IWR6843 $50-300
Processing Unit NVIDIA Jetson Orin Nano $200-500
Antenna Array (optional) For direction finding $100-500

Software Stack

Layer Technology Purpose
RF Capture GNU Radio, SoapySDR Interface with SDR hardware
Signal Processing Python (NumPy, SciPy) FFT, spectrogram, feature extraction
ML Classification TensorFlow / PyTorch Drone RF signature classification
Radar Processing Python + TI mmWave SDK Micro-Doppler payload detection
Database SQLite / PostgreSQL Known drone RF fingerprint storage
Real-time Pipeline Python / C++ Low-latency signal processing
Dashboard Flask / React Real-time alerts and visualization

Project Structure

SkyShield/
+-- index.html                  # Research document (feasibility, market, technical details)
+-- README.md                   # This file
+-- CLAUDE.md                   # AI assistant context file
+-- src/                        # Source code (coming soon)
|   +-- capture/                # RF signal capture modules
|   +-- processing/             # Signal processing pipeline (FFT, spectrogram)
|   +-- classification/         # ML model training and inference
|   +-- radar/                  # Micro-Doppler radar processing
|   +-- fusion/                 # Multi-layer detection fusion engine
|   +-- neutralization/         # Drone neutralization modules (research only)
|   +-- dashboard/              # Web-based real-time dashboard
|   +-- database/               # Drone RF signature database
+-- models/                     # Trained ML models
+-- data/                       # RF signature datasets
|   +-- raw/                    # Raw SDR captures
|   +-- spectrograms/           # Processed spectrogram images
|   +-- labeled/                # Labeled training data
+-- docs/                       # Additional documentation
|   +-- papers/                 # Reference academic papers
|   +-- hardware/               # Hardware setup guides
+-- tests/                      # Test suite

Development Roadmap

Phase Timeline Description Status
Phase 1 Week 1-2 RF capture setup (SDR + GNU Radio + antenna) Planned
Phase 2 Week 3-4 Signal processing pipeline (FFT, spectrograms) Planned
Phase 3 Week 5-8 Dataset collection (fly drones, record RF signatures) Planned
Phase 4 Week 9-12 ML model training (CNN on spectrograms) Planned
Phase 5 Week 13-16 Radar integration (TI IWR1443 + micro-Doppler) Planned
Phase 6 Week 17-20 Real-time system + dashboard Planned
Phase 7 Week 21-24 Neutralization research (acoustic + GPS in shielded lab) Planned
Phase 8 Week 25-30 Field testing + optimization Planned

Payload Detection Methods

Method Accuracy Hardware Cost How It Works
Micro-Doppler Radar 90-97% $50-300 Radar detects rotor RPM changes from payload weight
Motor EMI Profiling 80-90% FREE (uses SDR) Motor EM noise shifts frequency with payload weight
Radar Cross-Section 85-95% $100-500 Drone reflects radar differently with payload attached
Flight Behavior AI 75-88% FREE (software) ML detects abnormal flight patterns from payload

Detectable Payload Types

Payload Weight Detectable? Method
Camera / Surveillance +100-300g Partial RPM shift + slight RCS change
Package / Box +500g-2kg Yes Major RPM + RCS + flight anomaly
Explosive / IED +200g-1kg Yes RPM + RCS + metallic radar reflection
Chemical / Spray +300g-1.5kg Yes Liquid sloshing vibration + weight
Signal Jammer +100-400g Yes Active RF emission + RPM shift

Neutralization Methods (Research Only)

LEGAL WARNING: Drone jamming, GPS spoofing, and active neutralization is ILLEGAL in most countries without government authorization. These methods are documented for authorized research, controlled lab environments, and academic study ONLY.

Method Effect Range Cost Legal Risk Novelty
RF Jamming RTH / Land 100m-2km $200-2K HIGH Low
GPS Spoofing Redirect drone 100m-1km $100-500 VERY HIGH Medium
Protocol Exploit Full control 100m-1km $100-500 HIGH Medium-High
RTH Trigger Return to launch 100m-2km $100-500 HIGH Low
Acoustic Gyro Attack Destabilize 10-50m $50-300 LOW VERY HIGH
Net / Physical Capture 10-100m $50-500 LOW Low

Market Context

The global Counter-UAS market is valued at $3.88 billion (2026), projected to reach $16.45 billion by 2034.

Commercial Competitors

Company Product Price Country
Dedrone (Axon) RF-360 ~$16K/sensor USA
DroneShield DroneSentry $50K-500K+ Australia
D-Fend Solutions EnforceAir Military pricing Israel
Rohde & Schwarz ARDRONIS $100K+ Germany
Anduril Sentry Tower Military contracts USA
Rafael Drone Dome Military contracts Israel
BlueHalo Titan C-UAS $1M+ USA

SkyShield's position: Affordable proprietary alternative ($500-3K) targeting defense agencies, security firms, and research institutions. No commercial product offers payload detection + neutralization at this price point.


Key Academic References

  • Payload Detection: "Multi-Frequency Radar Micro-Doppler Based Classification of Micro-Drone Payload Weight" - Frontiers in Signal Processing, 2021
  • Motor EMI: "Characteristic mode analysis of electromagnetic interference susceptibility in motor driven unmanned drones" - Nature Scientific Reports, 2025
  • Acoustic Attack: "Rocking Drones with Intentional Sound Noise on Gyroscopic Sensors" - USENIX Security, 2015
  • Neuromorphic Detection: "Neuromorphic Drone Detection: an Event-RGB Multimodal Approach" - arXiv, 2024
  • Passive Radar: "UAV Detection with Passive Radar: Algorithms, Applications, and Challenges" - MDPI Sensors, 2025
  • Sensor Fusion: "Radar and Acoustic Sensor Fusion using a Transformer Encoder" - arXiv, 2025
  • GPS Spoofing: "An Experimental Study of GPS Spoofing and Takeover Attacks on UAVs" - USENIX Security, 2022
  • Comprehensive Survey: "Securing the Skies: A Comprehensive Survey on Anti-UAV Methods" - CVPR, 2025

Legal Disclaimer

This project is for defensive security research and educational purposes only.

  • RF monitoring laws vary by country. Check local telecommunications regulations before operating SDR equipment.
  • RF jamming is illegal in virtually all countries without government authorization.
  • GPS spoofing is a federal crime in most jurisdictions.
  • Neutralization techniques should only be tested in RF-shielded environments (Faraday cages) or with proper government authorization.
  • The authors are not responsible for any misuse of this research.
  • Always comply with local aviation, telecommunications, and electronic warfare regulations.

Proprietary Notice

This is a private, proprietary research project. All rights reserved.

  • This repository and all its contents are confidential and owned by Suliman Khan
  • Unauthorized copying, distribution, or use is strictly prohibited
  • No part of this project may be reproduced, modified, or shared without explicit written permission
  • For collaboration or licensing inquiries, contact the repository owner

License

All Rights Reserved. See LICENSE for details.


SkyShield - Proprietary Counter-UAS Research
Suliman Khan | 2026 | All Rights Reserved

Releases

No releases published

Packages

 
 
 

Contributors

Languages