Bring the signals together.
The planned collection pipeline brings topology, device health and link conditions through Security Onion. The benefit: a common view of network behavior built from telemetry your team already collects.
A data foundation for network AI
Turn network data
into foresight.
Cascern brings devices, links and their changing conditions into one dynamic graph—giving AI models a connected view of your network to learn from. The first application: forecasting which node may fail next.
A Five Insights prototype. Seeking technical early adopters to test the data foundation and the models built on it.
DYNAMIC GRAPH / CONNECTED OVER TIMEILLUSTRATIVE SCENARIO
Different devices. One connected view.
Device health, link conditions and topology become a shared representation.
Concept illustration: multiple device types, their relationships and changing conditions. Not a live product interface or model output.
Make telemetry useful togetherConnect topology, device health and link behavior.
Create data AI can learn fromPreserve relationships and change over time.
Build toward earlier decisionsStart with predicting the next node failure.
01 / The value
Your network has a history.
Give AI a way to read it.
A device metric describes one part of a network. A dynamic graph connects that measurement to the surrounding devices, links and recent changes. Cascern is being built to make that connected history available as a data source for network AI.
Keep the relationships.
Radios, SATCOM terminals, routers, edge devices and hosts become nodes connected by links. Repeated snapshots capture how their conditions change together, giving models context beyond individual measurements.
Build models on that foundation.
The graph history becomes input for training AI models. Cascern’s first models explore next-node failure prediction; the broader opportunity is a shared data foundation for future network AI research and applications.
PLANNED COLLECTION STACK / Security Onion · Prometheus · NetBox · Zeek · Suricata · Chrony
02 / A new data foundation
Pixels give image AI a picture.
Graphs give network AI context.
Pixels encode an image in a form a model can learn from. Cascern aims to do something similar for network behavior: represent the devices, their connections and their changes over time in a structured form that AI can use.
INPUTS
Signals from your network
Topology · node health · traffic · throughput · packet loss · latency
THE FOUNDATION
A sequence of dynamic graphs
Different device types, connected links and changing conditions in one representation.
THE APPLICATION
Models that learn the patterns
First focus: which node is most likely to fail next—or whether no failure is expected.
The value is in both layers: the mechanism that collects and structures network behavior, and the AI models trained on it. Broader model applications remain a direction for development, not validated capabilities.
03 / First application
See the risk.
Evaluate the alternatives.
When a node fails, traffic shifts and other devices can come under pressure. Cascern’s first application uses recent graph history to explore which node may fail next, giving operators another signal when considering a response.
More time to consider a response.
The goal is earlier warning to help teams evaluate rerouting, backup links or relay repositioning before connectivity is lost. Operators remain responsible for decisions and actions.
Build on your monitoring stack.
The planned integration adds a graph-based data and modeling layer alongside existing collection and monitoring. On-premises Docker deployment is planned next to Security Onion.
A foundation for model experiments.
Explore whether a connected history of network conditions can support useful predictions. Early feedback will help assess the representation, collection requirements and model behavior.
Current evidence comes from simulated networks. Earlier warning and operational value on real networks still need to be evaluated with technical users.
04 / Who we want to hear from
Bring your network expertise.
Shape both layers.
Help us test whether the data captures what matters, whether the collection approach fits your environment, and whether the models produce useful signals. We need candid technical feedback as the prototype becomes ready.
Network engineers
Assess whether the graph captures the devices, links and changes that matter in practice.
SOC & NOC practitioners
Evaluate how the planned collection pipeline fits existing telemetry and monitoring workflows.
Tactical communications teams
Help define representative tests for connectivity under jamming, disruption or attack.
Integrators & researchers
Explore the data representation and model behavior in labs and research settings.
05 / Where we are today
A clear direction.
A prototype to evaluate.
The dynamic-graph approach and first predictive models are being developed toward a pilot. Early adopters can help evaluate the foundation before broader real-world claims are made.
Data representationGraph snapshots of nodes, links and changing conditions
Model testingTrained and tested on simulated networks
Live telemetryIntegration pipeline in development
DeploymentPlanned on-premises Docker container alongside Security Onion
Early-adopter roleAssess collection fit, graph representation and model usefulness
A few things to know
Before you get involved.
Is Cascern ready for production use?
No. Cascern is a prototype. Its models have been trained and tested only on simulated data. The live-telemetry pipeline is still being built, and performance on real networks has not yet been validated.
Does it replace our monitoring tools?
Cascern is designed to build on existing telemetry collection and monitoring. Its planned pipeline reads through Security Onion and structures the data as a sequence of graphs for AI models to learn from.
Is Cascern just a failure prediction model?
The value spans two layers: collecting and structuring network behavior as dynamic graph data, and training AI models on that representation. Next-node failure prediction is the first application. Other applications are a future research direction, not capabilities available today.
What does a dynamic graph capture?
Nodes represent devices such as radios, SATCOM terminals, routers and hosts; links represent their connections. Snapshots record node availability, traffic and connectivity alongside link availability, throughput, loss and latency. Their recent history lets models examine changing conditions across a connected network.
Will it reroute traffic or fix failures automatically?
No. Cascern is intended to provide predictive decision support. Network operators remain responsible for evaluating a warning and choosing whether to take action.
What happens after I express interest?
The Five Insights team will contact you to learn about your environment and discuss a suitable evaluation as the prototype becomes ready. Submitting the form does not automatically grant software access or commit you to a purchase.
Your application information
Five Insights uses the name, email, organization, role and testing interests you submit to evaluate participation and contact you about Cascern. Campaign source information may be saved with your application to understand how you found this page. This form does not request network telemetry. You can ask the team to remove your application when they contact you.