SentinelXEnterprise network securityby Uzumaru

The packet, the log,
and the request.

Three programs, not one. Misaka decides what crosses your network. Kuroko tells you what it means. Uiharu answers the door. Nothing to route around.

System readoutlive

$

Trusted by teams at

AnthropicCloudflareGitHubGoogleLeasewebMicrosoftOracleVercelAnthropicCloudflareGitHubGoogleLeasewebMicrosoftOracleVercel

Component 01

Misaka

Data collection & network control

The part of SentinelX that actually stands between your network and the internet. Everything goes through it, and it decides what happens next.

Plane
Enforcement
Runs in
Kernel space
Licence
Closed source
Version
v1.0.0
01

Blocking that holds

Change the DNS, use a public resolver, type the IP by hand — none of it gets around the policy you set.

  • Deep packet inspection
  • TLS SNI identification
  • Certificate fingerprints
  • IPv4 / IPv6
02

Traffic goes where you say

Give any destination its own way out. Development traffic and business traffic never share a path.

  • Policy-based routing
  • Multi-egress balancing
  • VLAN / VRF isolation
  • Device authentication
03

Rules take effect now

Edit a rule in the console. Every device on the network is following it seconds later. Nothing restarts.

  • eBPF kernel driver
  • Hot update < 10ms
  • Central policy store
  • Version rollback
04

Nothing goes unrecorded

Every connection is written down and cannot be edited after the fact. This is what Kuroko reads.

  • Tamper-proof chain
  • ISO 27001 / SOC 2
  • S3 / Elasticsearch
  • 90-day hot retention

What happens to a packet

  1. 01

    Ingress

    The packet is captured on the way out, before anything can intervene.

  2. 02

    Identity

    Who sent it, from which device, and where it is headed.

  3. 03

    Match

    Checked against your policy. Block, allow, steer, or mirror.

  4. 04

    Enforce

    The decision is applied and the connection is logged.

Packet flow

Anything leaving your network is sorted on the way out and sent to the exit you chose for it.

USEROUTBOUND TRAFFICSentinelXMISAKA ENGINEeBPF · DPI · PolicyNORMAL OUTclean trafficAUDIT OUTrouted & loggedBLACKLISTLOGGED & AUDITED● LOG STOREsinkholed + recordedALLOWINSPECTDENYPACKET TYPEClean trafficSuspiciousThreat / Blocked

NORMAL OUT

Clean traffic exits via primary WAN gateway

AUDIT OUT

Financial & social traffic routed to secondary interface and logged

BLACKLIST

Explicit / blocked content — sinkholed and logged

Component 02

Kuroko

Log analysis algorithm

Millions of log lines go in. A short list of things worth your attention comes out.

Plane
Analysis
Reads
Misaka logs
Licence
CC BY-NC 4.0
Version
v1.0.0
Project-Uzumaru/SentinelX-Algorithm ↗
01

It learns what normal is

For every machine and every person on your network, separately. Nobody writes the rules.

  • Per-host models
  • Rolling windows
  • Seasonality aware
02

It judges each one on its own terms

A quiet build server and a busy workstation are not held to the same number.

  • Relative bands
  • 0–100 anomaly score
  • No global thresholds
03

It tells you the moment it matters

To your console, a webhook, or the SIEM you already use.

  • Webhook
  • SIEM export
  • Severity routing
04

It shows its work

Every finding comes with the records behind it, so you can see exactly what it saw.

  • Evidence bundles
  • Replayable windows
  • Immutable references

What it looks like

Baseline deviationhost 192.168.1.55 · 87/100
Egress volumecritical
Destination entropyhigh
Off-hours sessionselevated
Request cadencenormal
Policy hit ratenormal

Finding— sustained egress to three previously unseen destinations, 4.1× baseline volume, outside the host's normal activity window.

Component 03

Uiharu

Request risk control & bot defense

Every request is judged before your application ever sees it. Person or bot. Customer or attacker. Real click or replay.

Plane
Judgment
Verdict
0–100, four levels
Latency
< 20ms inline
Licence
Closed source
Version
v1.0.0
01

It looks at intent

Blocklists know what was bad yesterday. Uiharu decides about the request in front of it, which is how it catches things nobody has named yet.

  • Behavioural scoring
  • No rule authoring
  • Adapts per route
02

It knows a bot from a person

Automated clients pretending to be browsers do not pass as people here.

  • TLS fingerprinting
  • Headless detection
  • Pointer & timing signals
03

It knows where they really are

Even when the request is dressed up to look like it came from the same city as your customer.

  • Residential proxies
  • TOR exits
  • Datacenter ASNs
04

It sees the campaign, not the request

Credential stuffing, scraping, carding, enumeration — these are patterns. Uiharu reads them as patterns.

  • Cross-request correlation
  • Velocity windows
  • Campaign clustering
05

It rejects what is being reused

Replayed signatures, duplicated payloads, and links lifted onto somebody else’s site.

  • Signature replay
  • Origin binding
  • Payload dedupe
06

It runs how you want it

Off, watching quietly, or blocking on its own. Your call, and you can make a different one per route.

  • Off / monitor / block
  • Per-route thresholds
  • Zero integration

Live screening

Inbound screening0 scored · 0 stopped

waiting for inbound requests…

94%
of bad traffic never reaches you
<20ms
to decide, while the request waits
0
lines of code to turn it on

By the numbers.

Measured in production deployments

Misaka

0.00%

Block accuracy

Misaka

0μs

Policy latency

Uiharu

0ms

Risk verdict

Misaka

0k+

Concurrent sessions

Compared with DNS-level blockingTraditionalSentinelX
Bypassable via DNS changeYesImpossible
Bypassable via DoH / DoTYesImpossible
Bypassable via manual IPYesImpossible
Traffic steering controlNoneGranular routing
Behavioral anomaly detectionNoneKuroko, per-host baselines
Inbound request screeningNoneUiharu, inline
Audit log retentionNoneTamper-proof chain

One system, one bill.

Every plan covers all three components plus the managed platform around them. All plans include a 30-day trial.

PlanSmall teamsStarter$299/moLarge organizationsEnterprise$999/moAir-gapped & sovereignGov / MilitaryCustom
Misaka — network control
EndpointsUp to 200UnlimitedUnlimited
Precision domain blockingYesYesYes
Traffic steeringBasicGranularGranular
Connection log retention7 days90 days, tamper-proofCustom audit chain
DeploymentManagedManaged or on-premiseAir-gapped
Kuroko — log analysis
Baseline modelingSelf-hostedManagedManaged
Anomaly scoringSelf-hostedReal-timeReal-time
SIEM / webhook export—YesYes
Uiharu — request risk control
Request screening—Inline, < 20msInline, < 20ms
Bot & automation defense—YesYes
Custom rules & thresholds—Per routeFully custom
Programme
Compliance reporting—ISO 27001 / SOC 2Dedicated assessment
SupportCommunity24/7 dedicatedOn-site guarantee
CryptographyStandardStandardFIPS 140-2
Contact salesContact salesContact sales