The Privacy Stack for People Who Actually Have Something to Hide

Everyone says they have nothing to hide until their threat model includes an ex, an employer, a border agent, a data broker, or a state that changed its mind about what is legal last Tuesday.

Privacy advice on the internet is mostly theater. Use a VPN, they say. Use Signal, they say. As if downloading two apps will undo the fact that your laptop has a permanent advertising ID, your phone logs your location every four minutes, and your entire personality is sitting in plaintext in someone's S3 bucket.

If you actually have something to hide, and by that I mean real work, real sources, real research, real risk, you need a stack, not a tip.

Here is the stack I actually run.

Layer 0: Your Threat Model, But Be Honest

Programmers love to over architect before they understand requirements. Privacy people do the same thing. They buy a Faraday bag before they figure out who they are hiding from.

Write it down. Literally. Open a text file that will never touch the cloud and write three lines:

1. Who is my adversary and what do they want..? 2. What are they capable of technically, legally, financially..? 3. What happens if I lose…?

If your adversary is Instagram ads, your stack is different than if your adversary is a private investigator who charges $300 an hour. Most of us are somewhere in between. We are not trying to disappear from the NSA. We are trying to make surveillance expensive, noisy, and inconclusive.

This is where most people fail, because they try to build for movie level adversaries with convenience level discipline. A good stack is boring. It is consistent. It is reproducible. Think infrastructure as code, but for your life.

Layer 1: Identity Compartmentalization

One identity is a single point of failure. You would never run prod and staging on the same database. Why are you running your activism, your shitposting, your family group chat, and your client work from the same Apple ID?

The stack is:

One human, three personas minimum. Personal, public, private. Each gets its own email, its own number, its own payment method, its own browser profile. No cross contamination. No logging into personal Gmail from your research laptop because you were lazy.

I use SimpleLogin or Addy for email aliasing, MySudo or JMP.chat for phone numbers, and privacy.com or cash loaded virtual cards for payments. Each persona lives in its own Firefox profile with Multi-Account Containers and a hardened user.js. Or better, its own OS user account.

The mistake is thinking you can be anonymous. You cannot. You can be compartmented, which is more useful. You want an investigator to find three different boring people instead of one interesting person.

This is tedious. That is the point. OpSec that feels effortless is not working.

I automated most of the annoying parts. Log cleaning, metadata stripping, alias rotation, inbox triage. The scripts I actually use daily are in the Python OpSec Suite:

m Python OpSec Suite

It is not magic, it is just the boring Python glue that makes you actually stick to the system.

Layer 2: Network Transport Is Not Just a VPN

Your ISP sees everything. Your VPN provider sees everything after that. You just moved the problem.

The real network stack is layered:

Base layer: A router you own running OpenWrt or pfSense. Not the plastic box your ISP gave you. This is where you force DNS over HTTPS to a provider you chose, kill telemetry at the DNS level, and isolate your IoT junk onto a VLAN that cannot talk to anything.

Transit layer: Mullvad or Proton VPN on WireGuard, paid in cash or Monero, always on at the router. Then Tor for specific tasks inside that. VPN over Tor or Tor over VPN debates are mostly religious wars. My rule is simple. If you need to hide that you are using Tor from your ISP, use VPN then Tor. If you need to hide that you are using a VPN from the site you are visiting, use Tor then VPN. Most of the time, just use Tor Browser for research tasks and keep it separate.

Radio layer: This is the one people skip. Your phone is a tracking beacon that also makes calls. When you are doing sensitive work, it should not be near you, not on silent, not in airplane mode. Off or in a real Faraday sleeve. And yes, you should test your Faraday sleeve. Most Amazon ones are lies. Put your phone inside, call it. If it rings, you bought a fashion accessory.

Layer 3: The Endpoint, Where Privacy Actually Dies

You can have perfect network hygiene and still get owned because your OS is a keylogger with a weather widget.

My daily driver is not Qubes, because I like getting work done, but my sensitive work happens on a separate machine. Think cheap ThinkPad, Fedora or Debian, full disk encryption, no account sync, no biometrics. For maximum paranoia, Tails on a USB stick that you physically destroy when you are done.

On your main machine, harden what you have. Disable ad personalization, disable Siri, disable all the helpful AI features that are just euphemisms for we are reading your screen. Use LittleSnitch or OpenSnitch to see what is actually phoning home. You will be horrified.

And please strip your files. Every photo you take has GPS coordinates, every PDF has your name in it. I run a simple watcher that nukes EXIF before anything leaves my machine:


from PIL import Image
piexif = {}
# open, strip, save copy
# never share originals, ever


from PIL import Image
piexif = {}
# open, strip, save copy
# never share originals, ever

It is a one liner that saves you from the most embarrassing OpSec fail possible. There are ten more scripts like that, for PDFs, Office docs, audio files, in the suite I linked above.

Layer 4: Communications That Do Not Betray You

Signal is great. Signal is also tied to a phone number, which is tied to an identity, which is tied to a billing address.

Use it, but understand its limits.

For high risk comms: Signal with disappearing messages, registration with a non personal number, and you both verify safety numbers in person. For ultra high risk: SimpleX Chat or Cwtch, which have no user IDs at all. No phone number, no central directory to subpoena.

Email is a postcard. If you must use it, use Proton to Proton with PGP, and assume the metadata is still public forever. The subject line is not encrypted. The fact that you emailed a journalist at 2am is not encrypted.

And stop using your voice assistant to take notes about your sensitive project. I know it is convenient. So is leaving your house key under the mat.

This is why I went all in on local AI. If your AI assistant lives in someone else's cloud, you do not have an assistant, you have a witness. I built a completely offline assistant that runs on a Raspberry Pi 5 with a Hailo-8 accelerator. No internet, no logs, no terms of service. It does transcription, summarization, RAG over my own docs. It is my second brain that cannot be subpoenaed because it has never talked to the internet. The full build guide, parts list, and model setup is here.

Build an Offline AI Assistant on Raspberry Pi 5 + Hailo-8 Paid field manual in the offline systems / cyberdeck route. Hardware you can carry, software you can control, and…

It is the single biggest privacy upgrade I have made in two years.

Layer 5: The Physical and RF Layer Nobody Talks About

Your laptop leaks. Not metaphorically. Literally, via radio.

Your WiFi probes every network you have ever joined. Your Bluetooth is shouting your headphone's name. Your AirPods case is a tracker. Your car knows where you were.

Get a $30 NRF52840 dongle and run a BLE scanner sometime. Walk around your apartment. Count how many devices are screaming their presence. It is sobering.

This is where custom hardware stops being a hobby and becomes part of the stack. The Flipper Zero was a gateway drug. It showed everyone that RF is not magic. But the real power is in the custom pets you build yourself.

I am talking about ESP32-S3 boards with a tiny OLED, a CC1101 or NRF24 module, a PN532 for NFC, and a 500mAh lipo, all tucked into a 3D printed case that looks like a Tamagotchi. It does not hack anything. It listens. It tells you when there is a new BLE tracker following you. It tells you when your hotel room has a hidden device beaconing. It sniffs for deauth frames that indicate someone is trying to kick you off a network. It is a canary, not a weapon.

You can make it wearable. You can make it a cyberdeck card that slots into your backpack. You can give it a personality with a local LLM that narrates threats in plain English, like your own paranoid AI pet. Once you build one, you realize how much of the physical world is just badly secured wireless.

If you want to understand the other side of this, how RF mapping works at scale from the air, you need to study aerial reconnaissance.

[]](https://numbpilled.gumroad.com/l/dronereconblackboook) DRONE RECON BLACK BOOK $200 Drone + SDR = Autonomous RF Mapping, GPS Spoofing & AI-Directed Aerial…

Read it defensively. Once you see how trivial it is to map every WiFi and Bluetooth device in a neighborhood from 100 feet up, you will never trust your smart doorbell again.

Layer 6: The Pet Cyberdeck, Your Offline Brain

The final layer is the one I love most. The custom cyberdeck, the handheld, the wearable that is yours and only yours.

Forget the commercial trackers. Build a device that serves you. My current iteration is an ESP32-S3 plus a Raspberry Pi Zero 2 W, e-ink display, LoRa for off grid messaging, and a little keyboard. It runs no Google services. It holds my passwords via a hardware secure element, it generates TOTP codes, it holds my dead man switch notes encrypted. It is my digital soul in a box.

You can build this for under $100. The schematic is not the hard part. The hard part is deciding what it should not do. A good privacy pet does three things well and refuses to do anything else. Mine does: 1, detect trackers, 2, hold secrets, 3, provide offline AI assistance without ever touching WiFi.

These little devices are like cats. They are useless to anyone but you, and they make you smile when they boot up. That emotional attachment is a feature. You take care of things you like.

If you want a master list of hardware, firmware, custom PCBs, and the exact tooling I use for these builds, from custom Flipper firmware to ESP32 Marauder forks to Meshtastic privacy configs, I keep it all updated in the Rogue Loadout v3.7.

ROGUE LOADOUT v3.7- 2026 EDITION Underground Tooling Dossier for Security Operators Paid field manual in the privacy / opsec route. Less paranoia cosplay, more repeatable systems you can actually use…

It is the underground tooling dossier I wish I had when I started. It is not a shopping list, it is a system for choosing tools that do not betray you.

The Stack Is a Practice

Look, there is no final form. Privacy is not a product you buy, it is a practice you maintain, like writing tests. Some days you will be perfect. Some days you will log into your anon account from your personal IP because you are tired. The goal is not purity. The goal is to raise the cost of targeting you from trivial to expensive.

Start with one layer this week. Compartment your identities. Build one ESP32 scanner that tells you when a Tile tracker is following you home. Flash Tails to a USB stick and actually use it for one research task.

You do not need to become a ghost. You just need to become boring, compartmented, and slightly annoying to track. In 2026, that is the closest thing to invisible that still lets you have a life.

Build the stack. Keep it offline. Keep it weird.

Contents