Blueural Edge
Your network, your data, your AI.
A compact NVIDIA Jetson-based appliance that sits on your LAN and analyses traffic locally. Enterprise-grade detection that installs in an afternoon and never ships your raw traffic offsite.
At a glance
- Form Factor
- Fanless, compact chassis
- AI Compute
- 40–275 INT8 TOPS
- Connectivity
- 1–4 GbE, optional Wi-Fi/4G
- Camera Input
- Multi-stream via DeepStream
- Traffic Ingest
- SPAN/mirror port or passive TAP
- Deployment
- Out-of-band, non-inline
Capabilities
- Hardened Linux base built on the NVIDIA JetPack stack
- Signature-based intrusion detection using Suricata and Snort
- Anomaly-detection models for traffic no signature covers
- DeepStream pipeline for optional camera analytics
- Local web interface for setup and health monitoring
- Model updates delivered from the Blueural service
- Encrypted local log storage with configurable retention
Why it works this way
Non-disruptive by design
The appliance runs out-of-band. If it fails, traffic keeps flowing because it was never in the forwarding path.
Local inference
GPU acceleration is what makes on-premise analysis of both flows and video practical on a modest budget.
No data exhaust
Raw traffic and video stay on the device. Only alerts and metadata leave your premises.
Built for small teams
Designed for organizations with an IT generalist rather than a security operations centre.
An NVIDIA Jetson module in a fanless enclosure
The appliance is a Jetson compute module on a carrier board, sized so the whole unit runs passively cooled in a comms cabinet. The GPU is what makes analysing live traffic on-premise practical rather than theoretical.
- Inference runtime
- CUDA + TensorRT
- Video pipeline
- DeepStream
- Base image
- Hardened JetPack Linux
- Traffic path
- Out-of-band only
How it taps your network
Blueural Edge sits out-of-band. You either mirror traffic from a managed switch via a SPAN port, or install a passive network TAP. The appliance sees a copy of the traffic and never sits between your users and the internet.
What it analyses
- Network flow records and connection metadata
- DNS queries and responses, including tunneling patterns
- DHCP leases, to fingerprint new and unknown devices
- Firewall and syslog events forwarded from your gateway
- Optional camera feeds for physical-security correlation
Configurations
Three classes sized by network load and camera count.
Compact
Entry-class module
- Compute class
- ~40 INT8 TOPS
- Memory
- 8 GB
- Network
- 2 × 1GbE
- Camera streams
- Up to 2
- Power envelope
- 7–15 W
Standard
Most commonMid-class module
- Compute class
- ~100–157 INT8 TOPS
- Memory
- 16 GB
- Network
- 2 × 2.5GbE
- Camera streams
- Up to 4
- Power envelope
- 10–25 W
Extended
High-class module
- Compute class
- ~200–275 INT8 TOPS
- Memory
- 32–64 GB
- Network
- 2 × 10GbE + SFP
- Camera streams
- Up to 8
- Power envelope
- 30–60 W
Configurations are sized by network load. Talk to us about which class fits your environment.
Explore the platform
Blueural Sentinel
Analytics & Management Platform
One picture across every site.
Blueural Deception
Adaptive Decoys
Let intruders identify themselves.
Blueural Hunter
Threat Hunting Workbench
Go looking, instead of waiting.
Blueural Recon
External Exposure Scanner
See what an outsider sees.
Want to see Blueural Edge on your network?
Tell us about your environment and we will walk you through how it would be deployed.
Get in Touch