FREQUENTLY ASKED QUESTIONS
About CyPhER
Last reviewed August 19, 2026 · Mitchell Sipus, PhD, Founder & CEO
01In short
CyPhER automates the security operations center. It pulls from your existing SIEM, analyzes every log and all network activity, finds the problems distributed across your network, clusters the alerts, and pushes back a single consolidated threat event ready for mitigation. It compresses observed data by up to 75,000x, which is what lets it run on low compute and low storage, and it requires no new dashboard and can execute courses of action automatically or hand them to an analyst.
02Questions
What is the single biggest thing CyPhER does for a security team?
CyPhER automates the security operations center. It integrates with the security information and event management system you already run, ingests your logs and your live network activity, and analyzes all of it rather than a sampled portion. From that complete picture it locates the problems distributed across your network, clusters the related alerts into the events they actually belong to, and pushes back a single consolidated threat event to your existing console for direct mitigation. The work that consumes most of a security team's day, meaning correlating alerts across tools, reconstructing what belongs to what, and separating real activity from noise, is the work CyPhER performs before an analyst ever opens a ticket. It does this at speed and at low cost, compressing observed data by up to 75,000x, which is what allows complete analysis to run on ordinary hardware with a low compute and low storage footprint rather than requiring a dedicated appliance or a graphics processing unit.
What is automated network discovery?
CyPhER maps your entire network from observed activity rather than from documentation. Most organizations running large or long-lived infrastructure have network diagrams that stopped matching reality years ago, asset inventories assembled from several partial sources, and segments nobody currently owns. Because CyPhER observes all communications rather than sampling them, the map it produces reflects what actually communicates with what, including the connections nobody knew existed. Understanding the network is the prerequisite for everything downstream, since you cannot segment, restrict, or defend what you cannot see, and discovery is a byproduct of the sensing rather than a separate product you have to buy and run.
What are automated courses of action?
A course of action is the response step that follows a finding, and CyPhER can generate and execute them automatically. Our AI blueprint interprets friendly activity from threatening activity across all observed traffic, which means the system distinguishes the normal operation of your environment from the activity that warrants a response, and it does so from a factual record rather than from a probability score. From there the response can be automated end to end, including cordoning off affected segments, quarantining implicated systems, and applying mitigations through your existing enforcement tooling. Full automation is available where an organization wants it, and it is available specifically because the finding underneath it is deterministic, which is what makes acting on it without a human safe enough to consider.
Can a human stay in the loop?
Human in the loop is optional and fully supported, and organizations choose different points on that spectrum for good reasons. What makes either choice workable is the nature of the finding. Our approach is deterministic and mathematical, operating on OSI layer 2 data, which produces sensing and mitigation that are concise and precise rather than probabilistic. An analyst reviewing a CyPhER finding is reviewing observed facts and can confirm or dismiss it by inspection, which is faster and more trustworthy than adjudicating a confidence score produced by a model. That same property is what makes full automation defensible for organizations that want it. Systems that act autonomously on probabilistic findings transfer risk rather than reducing it, since an automated action against a false positive is itself an operational incident.
How fast is CyPhER?
CyPhER analyzes complete network and log data at speeds that change what a security team can attempt. In an operational evaluation with a government customer, CyPhER distilled 2.5 million alerts down to 5 threat events. The relevant comparison is not to another tool's benchmark but to the workflow it replaces, since the same reduction performed by analysts working through alerts individually is measured in months or years and in practice never completes. Speed at this scale is what makes complete analysis practical rather than theoretical, and it is the reason sampling is unnecessary in our architecture when it remains a hard requirement in most others.
Does CyPhER require my team to adopt a new tool?
CyPhER has a zero adoption curve. It deploys as a backend solution with no dashboard for your team to learn, monitor, or staff. Findings are pushed into the security information and event management system your analysts already use, so the queue they work is the same queue, in the same interface, populated with consolidated threat events instead of raw alert volume. Courses of action can run automatically, which means the workflow extends from finding to mitigation without introducing another console into it. Products in this market frequently arrive with an investigation interface that has to be learned, tuned, and staffed before it produces value, and the most common reason a capable security product underperforms in production is that the console it requires never gets adopted.
What does CyPhER actually do?
CyPhER passively observes all communications crossing its observation points and turns them into a complete, queryable record of network activity at OSI layers 2 through 4. From that record it identifies adversarial behavior such as reconnaissance, lateral movement, beaconing, and staging, and it does so without signatures, without agents on endpoints, and without sending any traffic into the network. Four configurations address different parts of the problem, with CyPhER Fusion conducting massive log analysis over existing alert volume, CyPhER Rx conducting blazing fast real time network traffic analysis, CyPhER Edge deploying at the switch, and CyPhER OT deploying onto operational technology equipment including programmable logic controllers. The output defenders receive is a small set of specific, evidence-backed findings rather than a stream of raw alerts. In an operational United States Air Force evaluation, CyPhER collapsed 2.5 million security alerts into 5 findings in under 20 minutes.
Does CyPhER use signatures?
CyPhER is signature-independent. Signature-based tools can only detect attacks that have been seen before, catalogued, and distributed as detection content, which means the newest and most targeted attacks are precisely the ones they miss, and it means detection quality depends on how quickly a vendor ships updates. CyPhER instead works from the complete record of network behavior, so an adversary technique does not need to be previously known to be visible. Novel tooling still has to communicate across the network to accomplish anything, and communication is what we observe. Signature-based detection remains useful as a complementary layer for known commodity threats, and CyPhER coexists with it, but we do not depend on it.
Does CyPhER use machine learning or AI?
CyPhER's analytical methods are built on algebra rather than on trained neural network architectures, and that choice drives most of what differentiates us. Tools in the behavioral AI category place a learned model, often a convolutional or deep network, against network data and ask it to decide what looks anomalous, which means the output is a probability, the model requires training data that may already contain the adversary, and the reasoning inside the model is not directly inspectable. Algebraic methods operate differently. The same input produces the same result every time, so findings are reproducible rather than probabilistic. The operations involved map efficiently onto modern computing hardware, which makes analysis at full network scale computationally practical where neural approaches would require far more compute for less certain answers. Results arrive faster, because the work is direct computation rather than inference through a trained model. And the path from a finding back to the observed activity that produced it remains traceable, so an analyst confirms a finding by inspecting evidence rather than by trusting a score. Layering additional models does not resolve this, which is worth stating plainly because the market's most common answer to the limits of behavioral AI is more AI arranged hierarchically. Every layer inherits the epistemic status of the layer beneath it, so a stack built on a learned baseline produces a more sophisticated estimate rather than a fact. We do apply modern AI, and we apply it where it belongs, which is helping analysts interrogate and act on a factual record rather than manufacturing the record itself.
What network layers does CyPhER observe?
CyPhER observes OSI layers 2 through 4, which cover the addressing, connection, and transport behavior of every communication on the network. This is a deliberate design choice with two consequences worth understanding. First, layers 2 through 4 are where the structure of adversary movement lives, since reconnaissance, lateral movement, and exfiltration are all visible as patterns of who connected to whom, when, and how. Second, this is the layer at which complete observation stays computationally tractable at full network scale, which is what allows CyPhER Rx to conduct blazing fast traffic analysis across environments where other approaches are forced to sample. Behavior at these layers is also unaffected by encryption, which is a growing structural problem for platforms whose detection depends on content-level inspection.
How does CyPhER handle encrypted traffic?
Encrypted traffic is fully visible to CyPhER because our detection works from communication behavior at OSI layers 2 through 4. CyPhER Rx analyzes network traffic in real time and resolves who communicated with whom, when, in what volume, and in what rhythm, and none of that is obscured by encryption. Platforms built on content inspection lose fidelity as encryption spreads, and the industry's workarounds, including decrypting traffic in transit and inferring from certificate metadata, add cost, operational risk, and privacy complications. An adversary can encrypt what they say, and they cannot encrypt the fact that they said it, to which system, and how often, and those observable facts are what our analysis reasons over.
How large a network can CyPhER observe?
CyPhER was designed for large-scale networks from the beginning, because it emerged from MIT research aimed at national-scale network defense and was validated on high performance research computing infrastructure. Scale is where the deterministic approach is most differentiated, since sampling exists in the first place because most architectures cannot handle complete observation at size. Our deployments and evaluations with United States Air Force organizations involve environments of significant operational scale. For a specific environment, sizing is an engineering conversation about observation points and traffic profile.
04All topics