Android SDK · v1.0 · Open Source

Device risk intelligence,
silent at the point of decision.

One Android SDK. Nine fraud signals. Runs at KYC and loan disbursal — no user friction, no data stored on device. DPDP compliant.

9
Fraud Signals
3
Risk Bands
<200ms
API Response Target
0
PII Collected

Loan apps and KYC flows are blind to device-level fraud.

Malware, rooted devices, emulators, and remote control apps compromise the integrity of every decision made on top of them. Most fintechs only detect this after disbursement — if at all.

01
🦠

Malware on device

Loan app malware intercepts OTPs, captures credentials, and manipulates on-screen content at the moment of signing.

02
🖥️

Emulator farms

Fraud rings run hundreds of synthetic applications through emulators. No device check means no detection.

03
📱

Remote control abuse

AnyDesk, TeamViewer, and similar apps allow third parties to operate a victim's device during an active session.

04
🔓

Root & hook exploits

Rooted devices and Frida/Xposed hooks bypass app security controls entirely — including your fraud rules.

Three lines of code. One API call. Instant risk score.

The SDK runs a silent device scan at your chosen checkpoint — KYC, login, or loan disbursal — and returns a structured risk assessment — pilot target under 200ms.

1
📦

Add the SDK

Drop the .aar into your Android project. No Play Services dependency. Minimum SDK 21.

2
⚡

Trigger at checkpoint

Call DeviceRiskSDK.assess(context) at KYC start or before disbursal confirmation. Silent. No UI.

3
📊

Act on the score

Receive a risk band (LOW / MEDIUM / HIGH) and triggered signal IDs. Block, step up, or flag for review — your logic, your call.

// Kotlin — trigger at KYC checkpoint
val result = DeviceRiskSDK.assess(context, SDKConfig(
apiKey = "your_api_key",
clientId = "nbfc_xyz",
sessionId = UUID.randomUUID().toString()
))

when (result.riskBand) {
RiskBand.HIGH → blockApplication()
RiskBand.MEDIUM → triggerStepUp()
RiskBand.LOW → proceedNormally()
}

9 deterministic signals. No ML black box.

Every signal maps to a specific fraud vector. Weights are documented and auditable. You see exactly why a device scored HIGH.

ROOT_01
Rooted Device
System-level access bypasses all app security controls
+40
EMU_01
Emulator Detected
Device is a virtual machine — common in fraud farm operations
+50
MAL_01
Malware / Blacklisted App
Known malicious or blacklisted package detected on device
+50 (cap 60)
HOOK_01
Runtime Hooking
Frida or Xposed framework active — can manipulate app behaviour
+20
PERM_01
Loan Permission Combo
Suspicious combination of SMS, contacts, and device admin permissions
+30
RC_01
Remote Control App
AnyDesk, TeamViewer, or equivalent present and potentially active
+25
OVER_01
Overlay + Accessibility Combo
Draw-over permission paired with accessibility service — screen capture risk
+25
PERM_02
SMS + Device Admin Combo
OTP interception vector combined with device admin control
+20
NOTIF_01
Notification Listener
App registered to read all device notifications including OTPs
+15

Clear thresholds. Actionable outputs.

Score is capped at 100. Every assessment returns a band, a score, and the specific signal IDs that triggered — so your team knows exactly what fired.

Low Risk
0 – 30
Proceed normally. No anomalies detected.
Medium Risk
31 – 60
Step-up verification or flag for manual review.
High Risk
61 – 100
Hold for manual review or step-up verification. Investigate triggered signals.

Built for Indian regulatory requirements.

Zero PII collected. No data stored on device. API responses are ephemeral. Designed for DPDP Act 2023 compliance from day one.

🛡️

DPDP Act 2023

No personal data collected or transmitted. Assessment is signal-only.

🔒

Zero PII

Device signals are behavioural and technical — no name, phone, Aadhaar, or biometric data.

📋

Audit Trail

Every assessment returns a unique assessment_id. Full signal log available for compliance review.

🔑

API Key Auth

Per-client API keys. Rotate or revoke at any time. No shared credentials across tenants.

Part of a Connected Investigation Stack

Device intelligence is the first layer.
Investigation and regulatory defence are what follow.

The Device Risk SDK captures evidence at the point of decision. A decision engine reconstructs the attack chain. An investigation standard packages it for regulatory defence. Three layers — one connected investigation stack.

Layer 01
Device Risk SDK
Evidence at point of capture
9 deterministic signals. Fires at KYC, login, or disbursal. Returns risk band, triggered signals, and evidence payload — in under 200ms.
Current page ↗
→
feeds into
Layer 02
ZIC Engine
Attack chain reconstruction
Investigator canvas. Maps device signals into entity relationship graphs. Surfaces fraud rings, mule networks, and attack sequences — on one screen.
Learn more ↗
→
structured by
Layer 03
ZAIS™
Regulatory defence package
ATO Investigation Standard. 42 deterministic signals, 3 playbooks, SHA-256 sealed evidence. RBI-implemented. Every decision defensible and replayable.
View ZAIS™ ↗
⚠
Device intelligence alone is not enough
A HIGH risk band tells you something is wrong. ZIC and ZAIS tell you what happened, why, and how to defend the decision.
✓
Evidence, not just scoring
Every signal the SDK triggers is preserved as a timestamped, SHA-256 sealed evidence artifact — ready for RBI audit, customer dispute, or regulator review.
→
Start with the SDK, extend to the full stack
Pilot the Device Risk SDK first. When you need investigation-grade analysis, ZIC and ZAIS are already designed to receive SDK output directly.