Research
Every technique has a story.
These are the ideas inside RipTide, told the way they happened: the intruder we were up against, what we tried first, what broke, and what finally worked. Real numbers where we have them, measured on the machines you'd actually use.
Answer like the real thing
A decoy is only as good as its answers.
Scanners and AI agents ask for things that aren't there. These are the ways RipTide answers anyway, convincingly, on any size of box.
Free, local decoy replies
A 23 MB local model picks the right pre-written reply in milliseconds. No GPU, no API key, no cloud bill.
Read the story Decoy replies · 11 min readDynamic replies, written by an LLM
A large language model writes a fresh, in-character reply to any request in about a third of a second. Every reply is checked before it ships.
Read the story Stealth · 9 min readWire-accurate server personalities
Decoys that speak Apache, nginx, IIS, Go or uvicorn down to header order, error pages and when they hang up.
Read the storyKnow what touched it
Separate what it said from what it did.
Every visitor introduces itself. RipTide checks the introduction against the evidence, and never confuses the two.
Crawler verification
Anyone can print a "Googlebot" name tag. RipTide checks every crawler's claim against the addresses its operator publishes.
Read the story Provenance · 8 min readObserved vs. self-reported
An agent can tell a decoy anything. RipTide labels every fact by how it knows it, so a claim never turns into a conclusion.
Read the story Grouping · 9 min readIntrusion groups
A new IP address is free. A new playbook isn't. RipTide links sessions by behavior, and only when two kinds of evidence agree.
Read the storyDecoys for the new attack surface
Where intruders go looking now.
Model servers, agent tools and login pages: the places an intruder, or its AI agent, reaches for first.
MCP server decoys
An internal tool server that speaks both versions of MCP, asks every caller to introduce itself, and has nothing real behind it.
Read the story AI infrastructure · 8 min readOllama and vLLM decoys
Model servers that answer like Ollama and vLLM down to the error bytes, run no model, and record every prompt.
Read the story Credential capture · 8 min readCanary credentials
Decoys that ask for credentials like real servers, fingerprint every one offered, and know their own planted canaries on sight.
Read the storyHow we write these
Stories, with the numbers attached.
People remember stories. Engineers trust numbers. Each piece has both.
Told as a story
Each one opens on a scene, names the problem, admits what didn't work, and ends with what changed. Scenes written to illustrate a technique are marked as illustrative.
Measured, and labeled as ours
Benchmarks come from our own test runs, with the hardware and the date. Outside statistics link to their source. If a feature isn't shipping, it isn't in here.
Make contact.
Read something you want to see live? Tell us which one. We'll show you.
Book a briefing
Thirty minutes with the people who built it. Bring your hardest question.
- Watch a live agent set off a detection
- Map decoys to your crown jewels
- Plan a first deployment in one sitting
Caught. In the good way.
Thanks. We'll reply to , usually within one business day.