Staff Monitored
0
Across all banks
AML Screenings
0
This month
Partner Banks
0
Connected
🚨 Recent Threat Alerts
No active alerts. System is monitoring.
⚙️ SENTINEL Health
SHIELD Engine
● Checking...
Compliance Service
● Checking...
Identity Service
● Checking...
Passport Service
● Checking...
📊 Threat Intelligence Summary
HIGH RISK SIGNALS
0
Vouch cartel suspects, SIM farms, synthetic MoMo activity
MEDIUM RISK SIGNALS
0
Velocity anomalies, GPS collisions, suspicious vouching patterns
MEMBERS CLEARED
0
Passed all SHIELD checks, TIU verified, clean record
Live Threat Alerts
Real-time SHIELD detection queue
Loading alerts...
Staff Profiles
All monitored bank staff members
Transaction Monitor
Real-time transaction surveillance — flag anomalies before theft occurs
🔍 Screen Transaction
⚠️ Suspicious Patterns
Pattern: Round-Amount Clustering
Multiple round-amount withdrawals (GHS 1,000; 5,000; 10,000) by same staff in short period — classic pre-embezzlement signal.
Pattern: After-Hours Activity
System access or transactions outside normal working hours (before 7am, after 7pm, weekends) without supervisor authorisation.
Pattern: Override Frequency
Staff using system override function more than 3 times per month — each override should be supervisor-approved and audited.
Behaviour Pattern Analysis
SENTINEL tracks 14 behavioural indicators that precede insider theft
🔴 High-Risk Indicators
💰
Lifestyle Inflation
Personal spending exceeds declared salary — vehicle, property, or luxury purchases
🔓
Credential Sharing
Login from multiple devices or locations simultaneously
📱
External Communication During Transactions
Phone calls or WhatsApp activity during large transaction processing
🔄
Account Cycling
Moving funds through multiple accounts to obscure source
🟡 Medium-Risk Indicators
😰
Financial Distress Signals
Personal loan defaults, debt collection calls at workplace, gambling
🌙
After-Hours System Access
Accessing core banking system outside authorised hours without log entry
🔍
Excessive Data Queries
Searching customer accounts not related to assigned portfolio
📋
Approval Bypass Attempts
Attempting transactions above personal authorisation limit
AML Screening
Screen members against AML watchlists and PEP databases under Ghana AML Act 2020 (Act 1044)
🔍 Screen Member
📊 Recent Screenings
No screenings yet this session.
Suspicious Activity Reports (SAR)
Generate and submit SARs to Ghana Financial Intelligence Centre (FIC) under Act 1044
📋 Generate SAR
📜 SAR Log
No SARs generated this session.
Legal Note
All SARs are submitted to Ghana FIC under Act 1044 S.35. Tipping off the subject of a SAR is a criminal offence under Act 1044 S.41.
Flag Management
Escalate, manage, and resolve member flags
🚩 Escalate Flag
📊 Flag Statistics
⚫ BLACK flags0
🔴 RED flags0
🟠 AMBER flags0
🟡 YELLOW flags0
SHIELD Scores
Behavioural Credibility Index and Dynamic Risk Scores from the SHIELD engine
📊 Get SHIELD Score
⚠️ Early Warnings
Vouch Cartel Detection
Graph cycle detection — identify closed vouching rings per anti-manipulation architecture v1.0
🕸️ Cartel Detection Rules
Graph Cycle Detection
Every vouch is stored in a graph DB. Directed cycles of length 2-8 are detected in real-time. Any cycle → all members flagged.
Temporal Spread Requirement
Vouches between members who joined within 30 days carry 40% weight reduction. Genuine communities have temporal depth.
Hard Vouching Cap
Max 12 unique members vouched per 365-day window. Beyond 12, all further vouches carry ZERO weight. No exceptions.
Default Accountability
When a member defaults, all vouchers lose Trust Score points proportional to vouch weight. Selling vouches is financially catastrophic.
🔍 Run Cartel Check
Vouch Integrity History
Enter a member ID and run check to see history.
Vouch Integrity
Inspect individual vouch relationships and integrity scores
Partner Banks
Banks connected to SENTINEL staff monitoring
No partner banks onboarded yet. Click above to onboard your first bank.
Onboard Partner Bank
Connect a bank to SENTINEL staff monitoring and UBL trust infrastructure
What SENTINEL gives your bank:
✓ Real-time staff behaviour monitoring
✓ Insider theft early warning system
✓ AML / PEP screening for all customers
✓ SAR generation and FIC submission
✓ UBL Trust Score integration
✓ Bank of Ghana regulatory reporting
Trust Integrity Unit (TIU)
Permanent adversarial testing and fraud pattern response — the immune system of UBL
🧠 TIU Operating Procedures
Daily
Review all automated fraud flags from past 24 hours. Triage: investigate / monitor / clear.
Weekly
Red team briefing — present findings from current exploitation attempt cycle.
Monthly
Fraud pattern report to TOC — aggregate stats on flagged accounts, confirmed fraud, emerging attack vectors.
Quarterly
Full adversarial simulation — red team attempts to extract a loan using the most sophisticated current attack vector.
Annually
Full external security audit by independent firm with West African financial fraud expertise.
⚡ Fraud Response Ladder
L1
Automated Flag
SHIELD detects anomaly. Account enters enhanced monitoring. No member notification.
L2
TIU Review
TIU analyst manually reviews flag. Access to loans restricted pending review.
L3
Account Freeze
Full account freeze. SAR generated. Network broadcast to all partner banks.
L4
FIC Referral
Permanent UIH blacklist. FIC criminal referral. Biometric anchor prevents re-registration.
Audit Chain
Immutable audit trail — all SENTINEL actions logged and cryptographically chained
AI Intelligence Centre
SHIELD AI — IsolationForest + GradientBoosting + 90-day predictive scoring
AI Assessments Run
0
This session
Anomalies Detected
0
IsolationForest flags
High Default Risk
0
>35% predicted default
Model Version
4.0.0-ai
GB + IF + Rules
🔍 Run AI Assessment
Full 3-layer AI risk assessment for any member
🧠 Model Architecture
Layer 1 — GradientBoosting (50%)
12-feature default probability classifier. Trained on synthetic member profiles. Returns probability of default + top 4 risk factor importances.
Layer 2 — IsolationForest (35%)
Anomaly detection model. Flags behavioral outliers that don't match legitimate member profiles. Contamination: 25%.
Layer 3 — Rules Engine (15%)
Trust score × 10. Baseline from verified community data. Overrides ML when trust band is verified/graduated.
Layer 4 — 90-Day Prediction
Forward-looking risk: base ML + trajectory adjustment + community trust decay. Updated every 72 hours.
📊 Batch AI Scoring
Run AI assessment across all staff members at once
Click "Score All Staff" to run AI assessments across all registered staff members.
🎙️ MFCC Voice Print Verification
Check voice print similarity across all registered members (85% threshold)