Security Dashboard
Real-time LLM threat detection overview — DeBERTa-v3 + EfficientNet-B3
🔍
0
Total Scanned
🛡️
0
Attacks Blocked
⚠️
0
Flagged
✅
0
Allowed Safe
📈
Scan Activity
live📋
Recent Scans
| Time | Input | Attack Type | Confidence | Status |
|---|---|---|---|---|
| No scans yet | ||||
🎯
Attack Distribution
📊
Detection Stats
—
detection accuracy
⚡
Live Feed
LIVE
Waiting for activity...
🌐
API Health
/gateway/chatchecking
/gateway/chat/multimodalchecking
Server—
Threat Scanner
Real-time multimodal attack detection — DeBERTa hierarchical pipeline
0 chars
⚔️ Attack prompts — click to test
Prompt Injection
Jailbreak
Data Poisoning
Model Inversion
Bypass Filter
✅ Safe medical prompts
Diabetes
Blood Pressure
Medications
Metformin
Aspirin dosage
📊
Session Stats
Scanned0
Blocked0
Flagged0
Allowed0
🕐
Last 5 Scans
🤖
Detection Pipeline
1
DeBERTa Binary
SAFE vs ATTACK
2
DeBERTa Subtype
Attack classification
F
Keyword Fallback
Only if conf < 40%
Scan History
Complete log of all threat detection scans
Time
Input
Attack Type
Conf.
Mode
Status
Settings
Configure API connection and detection parameters
🌐 API Configuration
Backend connection
🎚️ Detection Thresholds
Tune sensitivity (visual only)
60%
65%
👤 Account
Your profile
A
Admin User
admin@healthguard.io
ℹ️ System Info
Model and framework details
Version1.0.0
Text ModelDeBERTa-v3-base
Image ModelEfficientNet-B3
LLM BackendGroq / Llama3.1-8b
API FrameworkFastAPI
PipelineBinary → Subtype → LLM