Demo Dashboard

Professional-grade Streamlit dashboard for demonstrating and testing the fraud detection system.

Overview

The demo dashboard provides a visual interface for:

Real-time transaction testing with attack simulation presets
Score visualization with interactive gauges
Analytics and decision history
Policy inspection

Dashboard Sections

The sidebar displays real-time system health:

Indicator	Description
System Status	Overall health (Healthy/Degraded/Down)
Redis	Velocity counter store status
PostgreSQL	Evidence storage status
Policy	Policy engine loaded status
Policy Version	Current policy version number

Also shows API endpoint information and last decision summary.

Transaction Simulator

Interactive form for submitting test transactions with pre-configured attack scenarios.

Attack Presets

Preset	Description	Expected Decision
Normal Transaction	Typical legitimate $50 transaction	ALLOW
Card Testing Attack	Small $1 amount with datacenter IP	REVIEW/BLOCK
Velocity Attack	High-value $1500 from new account	FRICTION/REVIEW
Geographic Anomaly	US card transacting from Nigeria	REVIEW
Bot/Emulator Attack	Emulated device via Tor exit node	BLOCK
Friendly Fraud Risk	High-value from dispute-prone profile	FRICTION

Customizable Fields

Amount (dollars)
Service ID (telco service identifier)
Event Subtype (sim_activation, topup, device_upgrade, sim_swap)
Card Token
User ID / Subscriber ID
Phone Number and IMEI (telco-specific)
Device flags (Emulator, Rooted)
Network flags (Tor, Datacenter, VPN)

Decision Result Display

When a transaction is submitted, the dashboard shows:

Decision Badge

Large colored badge indicating the decision:

ALLOW (green) - Transaction approved
FRICTION (yellow) - Additional verification required (3DS, OTP)
REVIEW (orange) - Manual review queue
BLOCK (red) - Transaction declined

Score Gauges

Four interactive gauges showing:

Gauge	Description	Color Scale
Overall Risk	Combined risk score (0-100%)	Green to Red
Criminal	Criminal fraud probability	Green to Red
Friendly Fraud	Friendly fraud probability	Green to Red
Bot Score	Automation/bot probability	Green to Red

Color thresholds:

0-30%: Green (low risk)
30-60%: Yellow (medium risk)
60-80%: Orange (elevated risk)
80-100%: Red (high risk)

Detailed Score Breakdown

Expandable section showing individual detector scores:

Card Testing Score
Velocity Score
Geo Anomaly Score
Confidence Level

Triggered Reasons

Each risk factor that contributed to the decision is displayed as a card:

[CRITICAL] BOT_EMULATOR
Device appears to be an emulator

[HIGH] BOT_TOR_EXIT
Transaction from Tor exit node

Severity levels: CRITICAL, HIGH, MEDIUM, LOW

Friction/Review Information

If friction is required, shows:

Friction type (3DS, OTP, STEP_UP, CAPTCHA)
User-facing message

If review is required, shows:

Review priority (HIGH, MEDIUM, LOW)
Review notes summarizing concerns

Analytics Dashboard

Visual analytics for the last 24 hours of transactions.

Key Metrics Row

Metric	Description
Total Transactions	Count of decisions made
Avg Risk Score	Mean risk score
P95 Risk Score	95th percentile risk score
Avg Latency	Mean processing time
P99 Latency	99th percentile processing time

Decision Distribution Chart

Donut chart showing breakdown of decisions:

ALLOW (green)
FRICTION (yellow)
REVIEW (orange)
BLOCK (red)

Hourly Transaction Volume

Bar chart showing transaction count per hour.

Latency Over Time

Line chart showing average latency per hour with 200ms target line.

Decision History

Query historical decisions from PostgreSQL evidence table.

Filters

Limit: 25, 50, 100, or 250 records
Decision type: Multi-select ALLOW/FRICTION/REVIEW/BLOCK

Table Columns

Column	Description
Transaction ID	Unique transaction identifier
Decision	ALLOW/FRICTION/REVIEW/BLOCK
Amount	Transaction amount in dollars
Risk Score	Overall risk percentage
Latency	Processing time in milliseconds
Time	Timestamp of decision

Summary Statistics

Approval Rate (% ALLOW decisions)
Block Rate (% BLOCK decisions)
Avg Risk Score
Avg Latency

Policy Inspector

View the current policy configuration without editing.

Policy Overview

Version number
Default action
Description

Score Thresholds

Display of threshold configuration for each score type:

Score Type	Block	Review	Friction
Risk	90%	70%	50%
Criminal	85%	65%	45%
Friendly	95%	60%	40%

Active Rules

Expandable list of all enabled rules showing:

Rule name and priority
Description
Action (BLOCK, REVIEW, FRICTION)
Friction type (if applicable)
Conditions (JSON format)

Lists Summary

Count of entries in:

Blocklists (Cards, Devices, IPs, Users)
Allowlists (Cards, Users, Services)

Raw Policy

Expandable section showing full YAML configuration.

Policy Settings (v1.1)

Edit policy configuration with full version control and audit trail.

Current Version Info

Displays at the top:

Current Version (semantic: MAJOR.MINOR.PATCH)
Last Changed By
Last Updated timestamp
Policy Hash (for integrity)

Thresholds Tab

Interactive sliders for each score type (Risk, Criminal, Friendly):

Friction Threshold (green zone)
Review Threshold (yellow zone)
Block Threshold (red zone)

Validation: Ensures friction < review < block. Invalid configurations are blocked.

Rules Tab

Existing Rules Display:

Sorted by priority (lower = higher priority)
Status indicators (enabled/disabled)
Expandable details with conditions JSON
Delete button for each rule

Add New Rule Form:

Rule ID and Name
Description
Priority (1-1000)
Enabled toggle
Action (ALLOW, FRICTION, REVIEW, BLOCK)
Friction Type (3DS, OTP, STEP_UP, CAPTCHA)
Review Priority (LOW, MEDIUM, HIGH, URGENT)
Conditions (JSON editor)

Blocklists Tab

Manage blocked entities:

List Type	Description
Card Tokens	Permanently block specific cards
Device IDs	Block compromised devices
IP Addresses	Block malicious IPs
User IDs	Block fraudulent users

Each item can be added or removed. Changes create new policy versions.

Allowlists Tab

Manage trusted entities that skip fraud checks:

List Type	Description
Card Tokens	VIP or corporate cards
User IDs	Verified trusted subscribers
Service IDs	Whitelisted telco services

Version History Tab

View complete policy change history:

Version number with active indicator
Change type (threshold, rule_add, rule_delete, rollback, etc.)
Change summary
Changed by (user)
Created at timestamp
Rollback button for non-active versions

Rollback creates a new version from old content (immutable history).

Running the Dashboard

Prerequisites

Fraud API running on port 8000
Redis running on port 6379
PostgreSQL running on port 5432

Start Command

cd /Users/omega/Projects/FraudDetection
streamlit run dashboard.py --server.port 8501

Environment Variables

Variable	Default	Description
FRAUD_API_URL	http://localhost:8000	Fraud API endpoint
POSTGRES_URL	postgresql://fraud_user:fraud_pass@localhost:5432/fraud_detection	Database URL

UI/UX Design

Color Scheme

Element	Color	Usage
Primary	#3b82f6	Buttons, links, active states
Success	#10b981	ALLOW decisions, healthy status
Warning	#f59e0b	FRICTION decisions, degraded status
Danger	#ef4444	BLOCK decisions, errors
Review	#f97316	REVIEW decisions

Typography

Headers: System font, bold
Body: System font, regular
Metrics: Monospace for numbers
Code: Monospace with background

Layout

Responsive two-column layout for simulator
Full-width charts for analytics
Collapsible sections for detailed information
Sticky sidebar for system status

Demo Workflow

Recommended Demo Sequence

Start with Normal Transaction
- Show ALLOW decision
- Point out low risk scores
- Highlight fast latency (<10ms)
Card Testing Attack
- Show detection of small amount + datacenter IP
- Explain velocity counter concept
Geographic Anomaly
- Show country mismatch detection
- Explain geo-risk scoring
Bot/Emulator Attack
- Show BLOCK decision
- Walk through multiple triggered reasons
- Point out CRITICAL severity flags
Analytics Dashboard
- Show decision distribution
- Highlight latency performance
- Explain 200ms target
Decision History
- Filter by BLOCK decisions
- Show evidence capture for disputes
Policy Inspector
- Explain threshold configuration
- Show rule priority system
- Demonstrate hot-reload capability
Policy Settings (v1.1)
- Adjust a threshold slider (show validation)
- Add a test rule (show version increment)
- View version history
- Demonstrate rollback capability

Troubleshooting

Dashboard Not Loading

Check that Streamlit is installed:

pip install streamlit plotly pandas

API Connection Failed

Verify API is running:

curl http://localhost:8000/health

Database Connection Failed

Check PostgreSQL connection:

docker-compose ps

Charts Not Displaying

Ensure Plotly is installed:

pip install plotly

Dashboard Version: 1.1 Last Updated: January 3, 2026 - Added Policy Settings tab with version control

Overview​

Dashboard Sections​

System Status (Sidebar)​

Transaction Simulator​

Attack Presets​

Customizable Fields​

Decision Result Display​

Decision Badge​

Score Gauges​

Detailed Score Breakdown​

Triggered Reasons​

Friction/Review Information​

Analytics Dashboard​

Key Metrics Row​

Decision Distribution Chart​

Hourly Transaction Volume​

Latency Over Time​

Decision History​

Filters​

Table Columns​

Summary Statistics​

Policy Inspector​

Policy Overview​

Score Thresholds​

Active Rules​

Lists Summary​

Raw Policy​

Policy Settings (v1.1)​

Current Version Info​

Thresholds Tab​

Rules Tab​

Blocklists Tab​

Allowlists Tab​

Version History Tab​

Running the Dashboard​

Prerequisites​

Start Command​

Environment Variables​

UI/UX Design​

Color Scheme​

Typography​

Layout​

Demo Workflow​

Recommended Demo Sequence​

Troubleshooting​

Dashboard Not Loading​

API Connection Failed​

Database Connection Failed​

Charts Not Displaying​