Fraud-APK & Stalkerware Forensic Workbench for cyber cells and law enforcement.
Analyze → Detect → Report · 0% false positives · 85.7% detection rate · Court-ready evidence
One command: suspect APK → verdict + evidence package in seconds
Turn a suspicious Android app (APK/XAPK) into a court-ready evidence package:
- 🔒 Evidence integrity — SHA-256 hashing at intake, append-only hash-linked chain-of-custody, read-only source handling
- 🔍 Deep static analysis — manifest, permissions, 40+ secret patterns, C2/endpoint extraction, Firebase probing, packer detection, native
.soanalysis - 🛡️ Detection engine — extensible YARA rules (11 families), behavioral heuristics (stalkerware/droppers/notification-capture), Quark-style crime sequences
- ⚖️ Calibrated verdict — weighted risk score; 0% false positives, 0% missed malware on 43-sample validation
- 📄 Court-ready reporting — PDF report, STIX 2.1 / CSV / JSON / YARA IOC exports, interactive evidence graph, MISP/OpenCTI push
- 🖥️ Three interfaces — PyQt6 desktop GUI, FastAPI web dashboard + REST API, terminal CLI + batch mode
Validated against 28 real malware families (Dendroid, Exodus, BrazilianRAT, GhostPush, xHelper, xbot, KrepBank, JSSMSers, MalBus, FakeAV, FakeBanker, and more) and 15 benign apps (F-Droid, K-9 Mail, OpenVPN, OpenCamera, Aurora Store, KDE Connect, NewPipe, etc.).
| Metric | Result |
|---|---|
| Malware detected (MALICIOUS) | 85.7% (24/28) |
| Malware flagged for review (SUSPICIOUS) | 14.3% (4/28) |
| Malware missed entirely | 0% |
| Benign false positives (MALICIOUS) | 0% |
| Benign flagged for review (SUSPICIOUS) | 80% |
Honest: This is a forensic triage tool. SUSPICIOUS means "analyst must review" — no malware is silently missed, and no innocent app is falsely accused. Known-good open-source apps are allowlisted. Full report:
docs/validation/EXPANDED_REPORT.md.
PyQt6 Desktop GUI — Findings table (left) and Evidence Viewer with Chain of Custody (right)
Web Dashboard (left) and Interactive Evidence Topology Graph (right)
Terminal CLI — clean findings output with severity-colored verdict
Court-oriented PDF report — executive summary with evidence hash manifest and chain of custody
Pipeline: Acquisition → Static Analysis → Detection → Verdict → Reporting → User Interfaces
# Install dependencies
pip install -r requirements.txt
# GUI (recommended)
python -m apk_forensic_kit
# CLI — analyze an APK
python -m apk_forensic_kit --cli analyze suspect.apk --case-dir cases/Case_001 --analyst "ASI-001"
# Batch analyze a folder of APKs
python -m apk_forensic_kit --cli batch -d ./seized/ --cases-root cases/ --analyst "ASI-001"
# Start the web dashboard + REST API
python -m apk_forensic_kit --cli serve --host 0.0.0.0 --port 8000
# Verify chain of custody
python -m apk_forensic_kit --cli verify-custody --case-dir cases/Case_001# Upload and analyze
curl -F "file=@suspect.apk" -F "analyst=ASI-001" http://localhost:8000/api/analyze
# List cases, download PDF report, view evidence graph
curl http://localhost:8000/api/cases
curl http://localhost:8000/api/cases/Case_001/report.pdfcases/<CaseID>/
├── case.sqlite # evidence, findings, chain-of-custody (append-only)
├── report.pdf # court-oriented executive + technical report
├── result.json # full analysis result
├── graph.html # interactive evidence topology graph
└── iocs/
├── iocs.json # structured indicators
├── iocs.csv # spreadsheet indicators
├── iocs.stix2.json # STIX 2.1 bundle (share with other agencies)
└── generated.yar # YARA rules from extracted indicators
- Every artifact SHA-256/SHA-1/MD5 hashed at intake, verified before use
- Original APK is never modified (read-only analysis)
- Chain-of-custody is append-only and hash-linked; tampering detectable via
verify-custody - Tool versions + config fingerprinted per run for reproducibility
python tests/test_golden.py # 12 golden testsdocs/validation/ACCURACY_REPORT.md— Initial 10+8 validationdocs/validation/EXPANDED_REPORT.md— Expanded 43-sample validationdocs/validation/accuracy_expanded.json— Full per-sample results
This tool is intended for authorized digital-forensics and law-enforcement work. Only analyze applications you are legally permitted to examine. The authors are not liable for misuse. Malware samples referenced in validation were handled in an isolated environment, analyzed statically (never executed), and deleted from disk after testing.
GPL-3.0 — Vivek Goswami