Stop Building Plumbing.
Start Building Detection.

Drop a production-ready Windows kernel telemetry layer into your security product and start receiving real-time endpoint events today — without building drivers, debugging pool leaks, or losing engineers to kernel plumbing.

Get the Dev Kit →
x64 + ARM64 14 event families Kernel-to-userspace in microseconds Ships with a working reference agent 15+ years Windows kernel experience
Built on proven experience

Developed by a principal engineer with 15+ years of hands-on Windows kernel development. Not a research project — production driver work, shipped in commercial security products.

Production quality, verifiably

Built to pass Driver Verifier. No pool leaks, no IRQL violations, no race conditions under load. Tested against Patch Tuesday regressions, not just a lab VM.

No hidden behavior

No outbound connections. No telemetry phoning home. No kill switches in kernel code. The SDK does exactly what the documentation says — nothing else.

Kernel Telemetry Is Brutally Hard

Before you can write your first detection rule, train your first model, or ship your first demo, someone has to solve Windows kernel telemetry. That means:

  • A minifilter driver for file system events — creates, writes, renames, deletes
  • A WFP callout driver for network flows, DNS, and TLS metadata
  • Kernel notification callbacks for process start/exit, image load, thread creation, and registry writes
  • Handle acquisition monitoring via ObRegisterCallbacks — the only way to see OpenProcess calls
  • A high-throughput, crash-safe channel from kernel space to your user-mode agent
  • ETW consumer integration for authentication, WMI, AMSI, and scheduled task events
  • An IRQL-aware, pool-leak-free implementation that survives Patch Tuesday

This is specialized work. It takes a team of senior kernel engineers three to five years to do it correctly. Most security startups spend their entire seed round here — and by the time the plumbing works, the market has moved on.

The most dangerous assumption in security product development is that kernel telemetry is a six-month project.

What Heathen Gives You

Heathen replaces 3–5 years of kernel engineering with a single integration.

Link against the SDK, implement a single C++ interface, and your agent starts receiving normalized, structured security events from the Windows kernel — in real time, on both x64 and ARM64.

You Handle

  • Detection logic and threat intelligence
  • Your customer experience and dashboards
  • Your response capabilities and integrations
  • The market differentiation that earns the deal

We Handle

  • Every line of kernel code
  • Driver signing and compatibility across Windows 10 1903 through Windows 11
  • The ring buffer, rollup, and deduplication layer
  • ETW consumer integration across five Microsoft providers
  • MITRE-mapped starter detection rules as a baseline

Your engineers work on the problems that grow your business. We work on the problems that would have stopped you before you started.

Three Steps to Your First Event

Integration takes under a day for a working prototype. Most of that time is driver installation, not code.

Step 01

Load the drivers.

A single DriverOrchestrator call starts HeathenCNC (ring buffer broker), HeathenEDR (filesystem), and HeathenNetEDR (network). Load order, timing, and lifecycle are managed for you.

Step 02

Register your callbacks.

Fill in a HeathenEventCallbacks struct with the handlers your product needs. Every callback receives a typed C++ struct — no parsing, no offsets, no raw buffers. Unused families cost nothing.

Step 03

Receive events.

HeathenCNC shares a lock-free, memory-mapped ring buffer between kernel and your user-mode agent. Latency from kernel event to callback: microseconds. A slow consumer doesn’t drop the kernel.

// Full integration example — this is not pseudocode.
HeathenEventCallbacks cbs{};

cbs.onProcess = [](const EventHeader* hdr, void* ctx) {
    if (hdr->EventType == EVENT_PROCESS_START) {
        auto* e = static_cast<const ProcessStartEvent*>(hdr);
        // image path, command line, parent PID, user SID — all here
    }
};
cbs.onFile    = [](const EventHeader* hdr, void* ctx) { /* ... */ };
cbs.onHandle  = [](const EventHeader* hdr, void* ctx) { /* ... */ };
cbs.onNetwork = [](const EventHeader* hdr, void* ctx) { /* ... */ };

HEATHEN_RING* ring = HeathenRingCreate(16 << 20);
HeathenRingConsumeAndDispatch(ring, cbs);  // blocks; returns on each kernel signal

Everything That Matters on a Windows Endpoint

Event Family What You See
File System Create (new vs. existing), write+close with last-writer PID, rename with old/new path, delete, attribute changes
Process Start (image path, command line, user SID, PPID), exit, real-time allow/deny decisions before the process runs
Network TCP/UDP flow create/close, DNS queries with response IPs, TLS SNI, inbound and outbound with port and protocol
Registry Key/value create, open, set, delete — including writes into service key paths
Image Loads Every DLL and driver mapped into any process — path, base address, size, kernel vs. user-mode flag; real-time allow/deny before the image maps
Thread Creation Source and target PID — cross-process threads (remote thread injection) flagged automatically
Handle Acquisition Every OpenProcess / OpenThread / DuplicateHandle — original requested access, actually granted access, NTSTATUS
Named Pipes Server create and client connect, pipe name, role, async flag — C2 pipe name patterns detected out of the box
Authentication Logon/logoff (success, failure, type), special logon, network vs. interactive vs. service
WMI Activity Queries, filter/consumer/binding create — WMI persistence mechanism coverage
AMSI / Scripts Every script block evaluation, scan result, interpreter identification
Scheduled Tasks Create, delete, enable, disable — task name, path, and action
Token Privileges Privilege enable/disable events — SeDebugPrivilege, SeImpersonatePrivilege, and others
DNS Per-query, per-process — FQDN, record type, response addresses, error codes

Seven MITRE ATT&CK Rules, Ready to Use

Heathen ships an optional static detection library with seven baseline rules. Pure library — no threads, no I/O, no global state — callable directly from any event callback.

Rule Technique Severity
VSS / shadow copy deletion T1490 Critical
Cross-process memory access T1055 Suspicious
Remote thread injection T1055.003 Suspicious
Process suspension T1055 Suspicious
C2 named pipe patterns T1570 Suspicious
Service key modification T1543.003 Notable
Cross-process termination T1562 Notable

These are starting points, not a finished product. Your detection capability is your moat — Heathen gives you the foundation, not the ceiling.

The Cost of Building It Yourself

We understand the instinct. Kernel code that ships in your product feels better when your team wrote it. Here is what that actually costs:

Build It Yourself Heathen SDK
Time to first reliable file event 12–18 months Same day
Time to full telemetry coverage 3–5 years Included
Engineers required 3–5 senior kernel developers 0
Estimated total cost $3M–$8M in engineering alone Licensing fee
Kernel verifier / pool leak debugging Hundreds of hours Done
Windows version compatibility Your problem on every Patch Tuesday Our problem
Driver signing EV certificate + WHQL process Included in SDK
ARM64 support Separate project Included

The engineering talent that can build a correct, verifier-clean, production-grade minifilter stack is genuinely rare and expensive. The same talent can build detection logic, response orchestration, and customer-facing capabilities that are visible, valuable, and differentiated.

Heathen exists because the kernel work is hard, slow, thankless — and identical for every company that has to do it.

How This Works

Commercial License

Ready to ship to customers? Commercial licenses cover redistribution rights, priced by endpoint count at deployment.

  • Distribution rights — ship the drivers in your product
  • Startup: up to 1,000 endpoints — $1,800 / year
  • Growth: up to 10,000 endpoints — $9,600 / year
  • Scale / Enterprise — see full pricing
  • Priority support and compatibility guarantees

Endpoints are counted per device running your product. A 5,000-endpoint deployment requires the 10,000-endpoint tier — you are never billed retroactively for growth within a tier.

No retroactive licensing. No audits. Units deployed during an active subscription term are licensed in perpetuity for that version. Non-renewal stops new deployments and updates — existing installs keep running. No kill switches.

Get the Dev Kit

Fill out the form below. You’ll get the dev kit download and a no-distribution license within one business day — reviewed personally, not by a bot. Prefer to ask questions first? Email the developer directly →

Request received.

We’ll review your submission and send you the dev kit link within one business day.

Please open the agreement above before accepting.

We don’t spam. We don’t share your information. We respond to every submission personally. See our privacy policy.