CONTENTS
OPSEC MASTERCLASS
Operational security, from first steps to expert
OPSEC is the habit of working out what an attacker could learn about you, and closing those gaps before they're used. This course starts with a password manager and ends with Tor, hardened operating systems and running OPSEC for a whole company. Each chapter explains every term the first time it appears, so you can start from zero. Everything here is defensive: it's about protecting yourself, your family and your business.
- PART 1The basics. Accounts, devices and scams. This alone stops most attacks on most people.Start here
- PART 2Your footprint. What you leak without noticing, and how to split your life into separate boxes.Go to part 2
- PART 3Advanced. Anonymity networks, hardened systems, key management, real cases.Go to part 3
- PART 4Teams and companies. Policy, detection and what to do when it fails.Go to part 4
PART 1
THE BASICS
If you only read one part, read this one. Stolen passwords, missing updates and convincing messages are behind most break-ins, for individuals and for companies.
01
What OPSEC is
Security tools protect a thing. OPSEC protects information about you: what you own, where you are, who you talk to, what you're about to do. It's a way of thinking before it's a set of tools.
The term comes from the US military. In 1966 and 1967, during the Vietnam War, a team code-named Purple Dragon was asked why enemy forces seemed to know about bombing missions in advance. They found no spy and no broken cipher. The warnings came from small, unclassified details: flight plans filed with civil air traffic control, predictable radio patterns, supply movements. Each detail was harmless. Together they gave the mission away. The team called the discipline of finding and closing those leaks operations security, and in 1988 it became a formal US government program.
The same thing happens to ordinary people and small companies. A burglar doesn't need your alarm code if your holiday photos say the house is empty. A scammer doesn't need to hack your bank if your LinkedIn page says who approves payments and your out-of-office reply says that person is away until Monday.
The five steps
The original process has five steps, and it still works. You'll see it again in every chapter.
- Critical informationList what would hurt you if the wrong person had it.
- ThreatsWork out who would want it and what they're able to do.
- VulnerabilitiesFind where that information leaks today.
- RiskWeigh how likely each leak is against how bad it would be.
- CountermeasuresFix the worst ones first, then start again.
Security, privacy and anonymity
These three words get mixed up, and each one needs different tools.
| WORD | WHAT IT MEANS | EXAMPLE |
|---|---|---|
| Security | Nobody gets into your accounts, devices or data without permission. | A hardware key on your email account. |
| Privacy | People know who you are, but they can't see what you do or say. | An encrypted message to a friend. The service knows you two talk. It can't read the text. |
| Anonymity | People can see what's done, but they can't tell it was you. | Reading a site through Tor. The site sees a visit. It can't tell whose. |
A VPN gives you a little privacy from your internet provider and almost no anonymity. A strong password gives you security and no privacy at all. Knowing which of the three you need saves a lot of wasted effort, and that's what the next chapter is for.
02
Threat modeling
A threat model is a short, honest answer to “who would come after me, and how?” Without one you either do too little, or you burn out doing things that don't matter for your situation.
Five questions
The Electronic Frontier Foundation, a digital-rights group, uses five questions. Write the answers down. It takes ten minutes.
- What do I want to protect? These are your assets: your inbox, your savings, your home address, your customer list, a source's name, your children's school.
- Who do I want to protect it from? These are your adversaries. Be specific. “Hackers” isn't an answer. “Someone running a mass phishing campaign” or “my ex” is.
- How bad is it if I fail? Embarrassing, expensive, or dangerous?
- How likely is it? Mass scams reach everyone. A government-grade phone exploit reaches very few people.
- How much trouble am I willing to go through? A measure you stop using after two weeks protects nothing.
Who the adversaries are
| ADVERSARY | WHAT THEY CAN DO | WHAT THEY WANT | WHAT USUALLY STOPS THEM |
|---|---|---|---|
| Mass criminals | Send millions of phishing messages. Try leaked passwords on every site. Automated, not aimed at you. | Money, accounts to resell. | Part 1 of this course. |
| Targeted criminals | Research you first. Phone your staff. Swap your SIM. Send a fake invoice from a look-alike domain. | A large payment, ransom, crypto. | Parts 1 and 2, plus payment checks (chapter 17). |
| People who know you | Know your birthday, pet and old address. May have held your phone. May share an account with you. | Control, access, revenge. | Recovery settings, device checks, chapters 3, 10 and 18. |
| Online mobs | Dig through years of posts. Cross-reference usernames and photos. Publish your address. | To scare or punish you. | Chapters 6, 9 and 11. |
| Companies and data brokers | Collect location, purchases and browsing from apps and trackers, then sell profiles. | Advertising revenue. | Chapters 6, 7 and 15. |
| States and spyware vendors | Legal orders to providers. Network monitoring. Phone exploits that need no click. | Intelligence, prosecution, pressure. | Part 3, and outside help. No single tool is enough. |
Pick the profile closest to yours
Choose one to see who you're most likely up against and which chapters to read first.
Most people
- LIKELY ADVERSARIES
- Mass phishing, people trying leaked passwords, fake delivery and bank texts, data brokers.
- HOW IT USUALLY HAPPENS
- A password reused from a breached site, or a login page that looked right.
- READ FIRST
- Accounts, devices, phishing, then your public footprint.
A small business
- LIKELY ADVERSARIES
- Invoice fraudsters, ransomware crews, a former employee whose login still works.
- HOW IT USUALLY HAPPENS
- An email asks accounts to change a supplier's bank details. Or one mailbox is taken over and used to phish your customers.
- READ FIRST
- Accounts, phishing, OPSEC for a company, when it goes wrong.
Visible online: creators, streamers, journalists, founders with an audience
- LIKELY ADVERSARIES
- Harassment campaigns, obsessive followers, impersonators, SIM swappers who want your handle.
- HOW IT USUALLY HAPPENS
- Someone links your handle to an old username, then to a home address through a data broker or a photo taken from your window.
- READ FIRST
- Your footprint, compartmentalization, phones, home and travel, knowing you're targeted.
Someone who knows you: an ex-partner, a relative, a stalker
- WHAT'S DIFFERENT
- They know your answers to security questions. They may have had your unlocked phone, share a family cloud account with you, or pay your phone bill.
- HOW IT USUALLY HAPPENS
- A shared account, a location-sharing setting that was never turned off, a tracker in a bag or car, or a monitoring app installed while they had the phone.
- READ FIRST
- Account recovery, phones, trackers, signs of monitoring.
- BEFORE YOU CHANGE ANYTHING
- If you may be in danger, removing a monitoring app or changing passwords can tell the person you've noticed. Talk to a domestic-abuse support service first, from a device they've never touched, and plan the order of changes with them.
A well-funded attacker: states, mercenary spyware
- WHO THIS IS
- Investigative journalists, activists, lawyers, dissidents, and people close to them.
- HOW IT USUALLY HAPPENS
- Legal demands to your providers, monitoring at the network level, or spyware delivered through a message you never have to open.
- READ FIRST
- Phones, Tor, hardened systems, fingerprinting.
- GET HELP
- Don't do this alone. Access Now runs a free Digital Security Helpline for civil society, and Amnesty International's Security Lab checks phones for spyware.
Ranking the risks
Put each risk in a grid: how likely, and how bad. Work from the bottom right corner outwards. Here is the grid filled in for a typical small-business owner.
| MINOR DAMAGE | SERIOUS DAMAGE | SEVERE DAMAGE | |
|---|---|---|---|
| UNLIKELY | Someone guesses the guest Wi-Fi password. | A laptop is stolen on a train. | A phone exploit aimed at you personally. |
| POSSIBLE | An old forum account is taken over. | A former employee still has a working login. | Ransomware encrypts the shared drive, and the backups with it. |
| LIKELY | Spam and scam calls after a data-broker listing. | A reused password opens your email. | A fake “new bank details” invoice gets paid. |
The three red cells in this example are all handled in part 1 and chapter 17, with free tools. That's typical. The exotic threats get the attention, and the ordinary ones do the damage.
03
Accounts and logins
Most account takeovers come from a password that was reused, guessed or typed into a fake page. Three things fix nearly all of it: a password manager, a second login step that can't be phished, and recovery settings you've checked.
Passwords
When a website is breached, attackers get a list of emails and passwords and try each pair on hundreds of other sites. This is called credential stuffing, and it's why one reused password is enough to lose your email, your shop and your bank login in an afternoon.
A password manager is an app that creates a different random password for every site, stores them encrypted, and fills them in for you. You remember one long password and it remembers the rest. Bitwarden is free and open source. KeePassXC is free and keeps everything in a file on your own computer. 1Password is paid and polished. The ones built into Apple and Google devices are fine too.
For the few passwords you must type from memory (the manager itself, your computer login), use a passphrase: five or six random words picked by dice or by the manager's generator. Words you choose yourself are much easier to guess than they feel.
Current guidance from NIST, the US standards body, matches this: length matters more than odd symbols, there's no reason to change a password on a schedule if it hasn't leaked, and every new password should be checked against lists of breached ones. Password managers do that check for you.
The second step
Multi-factor authentication (MFA, also called two-step login or 2FA) asks for something besides the password. The methods are very far from equal.
| METHOD | STOPS A LEAKED PASSWORD | STOPS A FAKE LOGIN PAGE | STOPS A SIM SWAP | NOTES |
|---|---|---|---|---|
| Code by text message | yes | no | no | Better than nothing. Use it only where nothing else is offered. |
| Code from an app (TOTP) | yes | no | yes | Free. Aegis on Android, Ente Auth or 2FAS on any phone. |
| Approve a prompt on your phone | yes | partly | yes | Attackers spam prompts until you tap yes. Turn on number matching. |
| Passkey | yes | yes | yes | Built into phones, computers and password managers. |
| Hardware security key | yes | yes | yes | A small USB or NFC device, about 30 to 60 euros. Buy two and keep one at home. |
Passkeys and hardware keys are phishing-resistant. They check the real address of the website before they answer, so a fake page gets nothing even if you're fully fooled. Chapter 5 shows why app codes can't do this.
Your email is the master key
Nearly every account you have can be reset by a link sent to your inbox. And your inbox can often be reset through a phone number or an old second address. So the real strength of your bank login is the strength of the weakest thing above it in this picture.
Go through your main email account's security page and check each of these:
- Recovery phone and recovery email. Remove numbers you no longer own and addresses you don't protect as well as the main one.
- Security questions. A mother's maiden name is in public records. Treat the answer as a second password: put a random string in it and store it in your manager.
- Backup codes. Print them or store them in the manager. They're how you get back in after losing your phone.
- Devices and sessions. Sign out anything you don't recognize.
- Connected apps. Remove third-party apps that can read your mail and that you've stopped using.
- Forwarding and filters. Attackers who get in once often add a rule that quietly forwards your mail. There shouldn't be one you didn't create.
04
Devices
A device that's up to date, encrypted, locked and backed up survives most of what happens to devices: malware, theft, loss and coffee.
Updates
Most attacks on devices use flaws that already have a fix. The gap between a fix being published and criminals using the flaw is now often days. Turn on automatic updates for the operating system, the browser and your apps, and restart when asked. A device that no longer receives security updates (an old phone, an old router) shouldn't hold anything you care about.
Disk encryption
Full-disk encryption scrambles everything stored on the device so it's unreadable without your password. Without it, anyone who steals the laptop can pull the drive out and read every file, and your login screen doesn't get in the way.
- iPhone and recent Android phones: on by default once you set a passcode.
- Mac: FileVault, in System Settings under Privacy & Security.
- Windows: BitLocker, or “Device encryption” on Home editions. Check where the recovery key went. By default it's uploaded to your Microsoft account.
- Linux: LUKS, chosen during installation.
Encryption protects a device that's switched off or freshly restarted. A laptop that's only asleep keeps its keys in memory. If you're leaving it somewhere risky, shut it down.
Lock screen
Use at least a six-digit code on a phone. A longer code with letters is better if your threat model includes someone with forensic tools. Set the screen to lock after a minute or less. Hide message contents on the lock screen: login codes that show on a locked phone are readable by whoever picks it up.
Backups
The usual rule is 3-2-1: three copies of anything important, on two different kinds of storage, with one copy somewhere else. For most people that means the device itself, an external drive at home, and an encrypted cloud backup. Ransomware encrypts every drive it can reach, so a backup drive that stays plugged in all the time gets encrypted along with the laptop. Unplug it between backups, or use a cloud service that keeps old versions.
Restore one file from each backup twice a year. Plenty of people discover their backup stopped working months ago on the day they need it.
Browser
The browser is where most attacks arrive. Keep it updated. Install as few extensions as you can, because an extension can read every page you open, and popular ones do get sold to people who add tracking or worse. An ad and tracker blocker such as uBlock Origin is worth having: malicious adverts are a real route for fake download pages. Download software from the maker's own site or the official app store, never from a search advert.
Everything else on your network
Change the router's default admin password, update its firmware, and replace it if the maker has stopped updating it. Put cameras, TVs and other gadgets on the guest network so that a weak doorbell can't reach your laptop.
05
Phishing and social engineering
Attackers find it easier to ask for access than to break in. Social engineering means tricking a person into doing something. Phishing is the same thing done by message.
The forms it takes
- Email phishing. A message that looks like your bank, Microsoft, a delivery firm or a colleague, with a link to a fake login page.
- Smishing and vishing. The same by text message and by phone call. Caller ID can be faked, so the number on the screen proves nothing.
- Spear phishing. Written for you in particular, using details from your website and social media.
- Business email compromise. A message that seems to come from your boss or a supplier asks for a payment or for bank details to be changed. There's no malware and often no link, only a request.
- Prompt bombing. The attacker already has your password and sends login approvals to your phone again and again, sometimes with a call “from IT” asking you to accept one.
- Help-desk attacks. The attacker calls your phone carrier or your company's IT support, pretends to be you, and asks for a reset.
- Voice and video cloning. A few seconds of public audio is enough to copy a voice.
Why a login code doesn't stop a fake page
Modern phishing kits don't just copy a login page. They sit between you and the real site and pass everything along in both directions. This is called an adversary-in-the-middle attack.
- You open the linkThe page looks right because it's the real page, relayed through the attacker's server.
- You type your passwordThe attacker's server forwards it to the real site.
- You type the six-digit codeThat's forwarded too, within seconds, while it's still valid.
- The real site logs “you” inThe attacker keeps the session cookie and is now signed in as you, with no need for the password again.
A passkey or hardware key breaks step 3. It signs a response that includes the address the browser is really on. On the fake address, the signature is useless to the real site. This is the single biggest upgrade available for an account that matters.
Habits that work
- Go there yourself. When a message says there's a problem with an account, don't use its link. Open the site from your bookmark or the app.
- Let the password manager judge. It fills in passwords only on the real address. If it doesn't offer to fill, stop and look at the address bar.
- Check money requests on a second channel. Any new or changed bank details get confirmed by calling a number you already had, never a number from the message.
- Notice pressure. Urgency, secrecy and authority are the three levers. “Do it now, don't tell anyone, the director asked” is the pattern in nearly every case.
- Agree a family or team code word. If a call from a relative or the boss asks for money, ask for the word. Voice cloning can't answer that.
- Never read a code to anyone. No real bank, carrier or IT team asks you to read out a login code or approve a prompt you didn't start.
If you clicked
It happens to security professionals too. Speed matters more than embarrassment. Change that password from another device, sign out all sessions, check for new forwarding rules, and tell whoever needs to know. Chapter 19 has the full order.
PART 2
YOUR FOOTPRINT
Part 1 kept people out of your accounts. This part is about what can be learned without getting in at all: what's already public, what your devices give away, and how separate parts of your life get tied together.
06
What the internet already knows about you
Before anyone targets you, they look you up. OSINT, open-source intelligence, means collecting information from public sources. Do it to yourself first, so you see what they'd see.
Where it comes from
- Data brokers and people-search sites. Companies that assemble names, addresses, phone numbers, relatives and past addresses from public records and app data, and sell the result.
- Breach dumps. Old leaks contain your email, phone number, old passwords and sometimes your address and date of birth.
- Social media. Your posts, and your friends' posts, tags, comments and follower lists. Other people publish more about you than you do.
- Usernames. The same handle on a gaming forum in 2012 and on a professional account today links the two in one search.
- Public registers. Company registers, land records, court records, electoral rolls, domain registration (WHOIS), depending on the country.
- Photos. Hidden location data in the file, and visible clues in the picture.
An hour of self-audit
- Search your full name in quotes, with your town, employer and each phone number and email address you've used. Use two search engines. Look at image results too.
- Search each username you've ever used.
- Enter your email addresses at Have I Been Pwned, a free service that tells you which breaches they appear in. Change any password that was in one.
- Open your social profiles in a private browser window, signed out, and see what a stranger sees.
- Look up your own domain's WHOIS record and check that it doesn't show your home address.
- Write down everything you'd rather wasn't there. That list is your work for the month.
Removing things
- Data brokers. In the EU and UK you can demand deletion under the GDPR's right to erasure. In the US each broker has an opt-out form, and California has a single deletion request system that covers every registered broker. Listings come back, so repeat it a couple of times a year.
- Search results. Google's “Results about you” tool removes results that show your home address, phone number or email.
- Old accounts. Delete them rather than abandoning them. An abandoned account with an old password is a future breach.
- Old posts. Tighten who can see past posts, and remove the ones that show your home, your routine, your children's school or your car's plate.
Photos
A photo file carries EXIF data: the camera model, the time, and on a phone usually the exact GPS position. Large social networks strip it when you upload. Email, cloud-share links, “send as file” in chat apps and your own website keep it.
The picture itself gives things away too: a street sign, a shop front, the skyline from your window, a parcel label on the table, a badge on a lanyard, a reflection in a pair of glasses. People who do this as a hobby can place a photo from the shape of a ridge line. Before posting from home, look at what's behind you.
Timing
Post about a trip after you're home. A live holiday story says the house is empty, and a live conference selfie tells a scammer that your team can be told “she's travelling and needs this paid today.”
07
Metadata
Metadata is information about a communication, as opposed to its content: who, when, where, how long, how much. Encryption hides the content. It rarely hides the metadata, and the metadata is often enough.
A phone record showing a call to an oncology clinic, then to a sister, then to an insurance firm tells a story without one word of the calls. Michael Hayden, a former director of both the NSA and the CIA, said in 2014: “We kill people based on metadata.”
What each party sees when you browse
HTTPS is the encryption behind the padlock in the address bar. It hides what you send and read on a site. It doesn't hide which site you're on.
| WHO'S WATCHING | PLAIN HTTPS | VPN + HTTPS | TOR BROWSER |
|---|---|---|---|
| Others on the Wi-Fi, and its owner | Which sites, when, how much data. Not the pages. | That you use a VPN, and how much data. | That you use Tor, and how much data. |
| Your internet provider | Which sites, when, how much data. Your name and address. | That you use a VPN, which one, when. | That you use Tor, unless you use a bridge. |
| The VPN company | Not involved. | Your real IP address, which sites, when. Often your payment details. | Not involved. |
| The website | Your IP address, your browser's fingerprint, and anything you log in with. | The VPN's IP address, the same fingerprint, the same login. | A Tor exit address and a fingerprint shared with every Tor user. |
| Trackers on the page | Cookies and fingerprint follow you between sites. | The same. A VPN doesn't change this. | Blocked or isolated per site. |
Two details behind that table. Your device asks a DNS server to turn a site's name into a numeric address, and that question is often sent unencrypted to your provider. And when the encrypted connection starts, the site's name is sent in the clear in a field called SNI. Turning on “secure DNS” in the browser closes the first leak. The second is being fixed by a newer standard, Encrypted Client Hello, which isn't yet in use everywhere.
Every email carries headers: sender, recipient, time, subject, and the chain of servers it passed through. Some mail programs include the sender's own IP address. Headers are never end-to-end encrypted, even with PGP. If the fact that you wrote to someone is the secret, email is the wrong tool.
Documents
Word, Excel and PDF files store the author's name, the organization, the software version, edit times and sometimes earlier text. Before publishing a file, use “Inspect Document” in Office or run exiftool or mat2 on it. Chapter 15 covers the ways redaction goes wrong.
Size and timing
Even when everything else is hidden, an observer can see how much data moves and when. If a message of a certain size leaves your connection and one of the same size arrives somewhere else a moment later, the two can be matched. This is called traffic correlation, and it's the main weakness of every anonymity tool. Chapters 12 and 16 come back to it.
08
Messaging and email
End-to-end encryption (E2EE) means only the people in the conversation hold the keys. The company running the service carries the messages and can't read them. What differs between apps is what else they know and keep.
| SERVICE | CONTENT E2EE BY DEFAULT | WHAT THE COMPANY CAN HAND OVER | WATCH OUT FOR |
|---|---|---|---|
| Signal | yes | The date the account was created and the last time it connected. If the request names an active username, the phone number behind it too. | Needs a phone number to sign up. Set a username so you don't have to share the number. |
| yes | Who you talk to, when, your contact list, groups, IP addresses. | Cloud chat backups aren't end-to-end encrypted until you turn that on. | |
| iMessage | yes | Lookup logs. Message content too, if iCloud backup is on without Advanced Data Protection. | Falls back to plain SMS when the other person isn't on Apple. |
| Telegram | no | Ordinary chats and all groups are stored on Telegram's servers. | Only “secret chats” are E2EE, and only one to one. |
| SMS | no | Everything, from the carrier. | Can be intercepted on the phone network itself. |
| Ordinary email | no | Everything in the mailbox. | Stored for years on both sides. |
| Email with PGP, or Proton to Proton | body only | Headers: who, when, subject line. | Easy to get wrong. One reply in clear text quotes the whole thread. |
Where encrypted chats still leak
- The other person's phone. Your message is exactly as safe as the least careful device in the conversation.
- Backups. An encrypted chat copied to an unencrypted cloud backup can be read from the backup. Check this on both sides.
- Notifications. Previews on the lock screen, and synced copies on a laptop or tablet you forgot about.
- Linked devices. Attackers trick people into scanning a QR code that links the attacker's computer to their account. Review the linked-devices list in each app.
- Screenshots and forwarding. No encryption protects against a recipient who shares.
Settings worth changing
- Disappearing messages. Set a default timer. Old conversations are a liability on a lost or seized phone. A week or four weeks suits most people.
- Registration lock or PIN. Stops someone who takes over your phone number from registering your account on their device.
- Safety numbers. For sensitive contacts, compare the code shown in the app in person or over a call. It confirms nobody is sitting between you. If the app says the number changed, ask why before sending anything delicate.
- Encrypted backups. On WhatsApp, turn on end-to-end encrypted backup. On iPhone, turn on Advanced Data Protection where Apple offers it.
Email, realistically
Treat email as a postcard that's kept forever. Secure the account very well (chapter 3), use it for what needs a record, and move anything sensitive to an encrypted messenger. If you need to send a file by email, encrypt the file itself and send the password by another route.
09
Compartmentalization
Compartmentalization means splitting your life into boxes so that a problem in one doesn't spread to the others. It's the most powerful idea in this course, and it scales from two email addresses to entirely separate machines.
Start with three boxes
- Critical. Bank, tax, password manager, domain registrar, main cloud account. One email address used for nothing else and given to nobody. It never appears in a breach because no shop or newsletter ever had it.
- Personal. Friends, family, social accounts under your own name.
- Disposable. Shopping, newsletters, loyalty cards, sign-ups, anything that will be sold or leaked sooner or later.
Email aliases make the third box painless. An alias service gives you a different forwarding address for every site, all landing in one inbox. When one starts receiving spam, you know who leaked it and you switch that alias off. SimpleLogin and addy.io have free plans. Apple's Hide My Email and Firefox Relay do the same.
Do the same with phone numbers where you can. Keep your real mobile number for people and for the bank, and use a second number for sign-ups and classified ads. Your phone number is the best single identifier anyone has for you: it rarely changes and it's in every contact list and every broker file.
How separate identities get linked
Say you keep a personal identity and a separate one for a project, a side business or a pseudonymous account. The two stay apart only as long as nothing connects them. One shared item is enough.
Other links people forget: a contact list uploaded by an app from the phone that holds both identities, a photo reused as an avatar, a username pattern, following the same twenty niche accounts from both profiles, and logging in to both from the same browser window “just once.”
Levels of separation
| LEVEL | SETUP | KEEPS APART | STILL SHARED |
|---|---|---|---|
| 1. Aliases | A different email address per purpose. | Breach lists, spam, easy lookups. | Everything on the device and network. |
| 2. Browser profiles | A separate profile, or a separate browser, per identity. | Cookies, logins, history, saved passwords. | IP address, device fingerprint. |
| 3. OS user accounts | A second user account on the computer. | Files, apps, most accidents. | IP address, hardware, admin-level malware. |
| 4. Virtual machines | A computer inside your computer, per identity. | Nearly everything in software. | The host machine, and the network unless routed apart. |
| 5. Separate devices | A second laptop or phone, bought and used apart. | Hardware, accounts, contact lists. | You: your habits, your writing, your location. |
Most people need levels 1 and 2. A business owner should also keep company admin accounts away from everyday browsing: a separate browser profile used only for the bank, the registrar and the admin consoles removes a large share of risk for no money.
10
Phones
A phone holds your messages, your second login step, your photos and your location history, and it reports where it is all day. It deserves its own chapter.
What a phone gives away by being on
- To the mobile network. The SIM has a permanent number called the IMSI and the handset has one called the IMEI. To receive calls, the phone tells nearby cell towers where it is. The carrier logs this, and it's available to police with the right order. A new SIM in the same handset is linked to the old one through the IMEI.
- To apps. Location permission, the contact list, the photo library, and an advertising ID that lets different apps recognize the same phone. Location data sold by ordinary apps has been used to follow named individuals to their homes.
- To anyone nearby. Wi-Fi and Bluetooth signals. Modern phones randomize their Wi-Fi address per network, which helps.
- To fake towers. An IMSI catcher pretends to be a cell tower and collects the identities of phones nearby, usually by forcing them onto old, weak 2G.
Settings that cut most of it
- Review app permissions. Give location “only while using” and approximate where exact isn't needed. Remove contact and photo access from apps that don't need it. Delete apps you don't use.
- Remove the advertising ID. Android: Settings, Privacy, Ads, Delete advertising ID. iPhone: Settings, Privacy & Security, Tracking, turn off “Allow Apps to Request to Track.”
- Turn off 2G where the phone allows it. Android has a switch under SIM settings. On iPhone, Lockdown Mode does it.
- Set the USB port to charge-only while locked. Both systems have this, and it blocks most data-extraction tools.
SIM swapping
In a SIM swap, a criminal convinces your carrier to move your number to their SIM, by social engineering or by paying an employee. Your phone loses service and they receive your calls and text codes. From there they reset your email and drain what's behind it.
- Ask your carrier for a port-out PIN or a number lock. Most offer one and few advertise it.
- Remove your phone number as a login or recovery method on every account that allows something stronger.
- If your phone unexpectedly shows “no service” where there's normally coverage, call the carrier from another phone straight away.
Locked isn't always the same as encrypted
A phone has two locked states, and the difference matters if it's stolen or seized.
| STATE | WHEN | WHAT'S PROTECTED |
|---|---|---|
| Before first unlock | After a restart, before you've typed the code. | Nearly everything. The keys aren't in memory yet. This is the strong state. |
| After first unlock | Locked, but unlocked at least once since starting. The normal state all day. | Less. Many keys stay in memory, and forensic tools aim at this state. |
That's why recent iPhones restart themselves after three days without being unlocked, and why GrapheneOS, a hardened version of Android, restarts after 18 hours by default. If you're about to hand a phone over or leave it somewhere, switch it off.
Fingerprint and face unlock are convenient and they resist someone watching over your shoulder. They can also be used by someone who holds the phone up to you. Both systems have a quick way to require the code again: on iPhone, hold the side button and a volume button for two seconds. On Android, enable the “Lockdown” option in the power menu.
Stalkerware and spyware
Stalkerware is a commercial monitoring app that someone installs on a phone they've had in their hands. It needs the unlock code and a few minutes. A strong code that nobody else knows prevents nearly all of it. Signs are covered in chapter 18.
Mercenary spyware, such as Pegasus, is sold to governments and can infect a phone through a message with no tap needed. Very few people are targets. For those who might be: keep the phone updated, restart it daily (many infections don't survive a restart), and turn on Lockdown Mode on iPhone or Advanced Protection on Android. Both switch off the features these attacks most often use.
When the phone itself is the problem
For a sensitive meeting, the reliable choice is to leave the phone somewhere else. A phone that goes dark at the same time and place as another phone creates a pattern of its own, so leaving it at home, switched on, is quieter than turning it off at the door.
11
Travel, home and the physical world
Digital security assumes nobody has their hands on your device or knows where you sleep. This chapter is about when that assumption is under pressure.
Your address
Your home address is the piece of information that turns online harassment into a physical problem, so it's worth treating as critical.
- Use a business address, a mailbox service or a parcel locker for deliveries, domain registration and anything that prints on a public register. Check what your country's company register shows about directors.
- Turn on WHOIS privacy for every domain you own. Most registrars include it free.
- Shred or tear off shipping labels before boxes go in the recycling.
- Look at your home on street-level map services. Most let you request that the building is blurred.
- Make your fitness app private, and set a privacy zone around home and work.
In public
- Shoulder surfing. People read screens and watch codes being typed on trains and in cafes. A privacy screen filter costs little. Unlock with a fingerprint in public so the code isn't seen.
- Public Wi-Fi. With HTTPS nearly everywhere, the old risk of someone reading your traffic is small. What's left is the metadata in chapter 7 and fake networks with familiar names. Using your phone as a hotspot avoids both.
- Charging ports. A USB port can carry data as well as power. Attacks through public charging points are rare in practice, and a charge-only cable or your own plug removes the question.
- Calls. Don't read a card number or discuss a deal in a lounge. This is the oldest leak there is.
Trackers
Bluetooth tags like AirTags are cheap and get misused to follow people and cars. iPhones and Android phones now both warn you when an unknown tag is travelling with you. Don't dismiss that alert. Use it to make the tag play a sound, and check bags, coat pockets, wheel arches and under seats.
Crossing a border
Border officers in many countries can inspect devices with far fewer limits than police inside the country, and refusing may mean being turned back or having the device kept. Rules vary a lot, so check for the country you're entering. The general approach:
- Carry less. What isn't on the device can't be read. Sign out of accounts you don't need on the trip and remove apps you won't use.
- Consider a travel device. A wiped laptop or phone with only what the trip needs, and the rest reachable from the cloud after you arrive.
- Power off before the checkpoint. That puts the device in its strongest state (chapter 10) and disables face and fingerprint unlock until the code is typed.
- Know your position in advance. Decide before you travel what you'll do if asked to unlock, and whether you're carrying anything, such as client or source material, that you have a duty to protect.
- Treat a device that left your sight as suspect. If it was taken to another room, change the passwords it held and, for high-risk work, stop using it.
Hotel rooms and unattended devices
An evil maid attack is someone with a few minutes alone with your switched-off laptop, tampering with it so that it captures your password the next time you type it. Hotel safes open with a master code and don't change this. If your threat model includes it, keep the device with you. For everyone else: shut down fully when leaving a laptop, set a firmware password so it can't be started from a USB stick, and keep the keys to your accounts (hardware key, phone) on your person.
PART 3
ADVANCED
This part is for people whose threat model includes a capable, patient adversary, and for anyone who wants to understand how the strong tools work and where they stop working. Most readers don't need to run any of it day to day.
12
VPNs, Tor and network anonymity
A VPN moves your trust from your internet provider to the VPN company. Tor is built so that no single party has to be trusted. They solve different problems.
What a VPN does
A VPN (virtual private network) sends all your traffic through an encrypted tunnel to a server run by the VPN company, and from there to the internet. The Wi-Fi owner and your provider see only the tunnel. Websites see the VPN server's address.
It's useful for hiding browsing from the local network, for keeping your home IP address away from sites and from people you call or game with, and for appearing to be in another country. It doesn't make you anonymous: the VPN company sees your real address and every site you visit, your browser fingerprint is unchanged, and you're still logged in to your accounts. A “no logs” promise is a promise. Prefer providers whose claims have been tested by independent audits or in court, that accept payment without a name, and that own up to their limits.
How Tor works
Tor sends your connection through three volunteer-run servers called relays, chosen from several thousand. Your browser wraps the data in three layers of encryption, and each relay removes one layer, which is where the name “onion routing” comes from. Each relay learns only the hop before it and the hop after it.
Where Tor stops protecting you
- You log in as yourself. Tor hides where you connect from. If you sign in to your personal account, the site knows it's you.
- Someone watches both ends. An observer who sees traffic entering Tor from your connection and leaving toward a site at the same moments can match the two by timing and volume. This is the traffic correlation from chapter 7.
- Using Tor is visible. Your provider or your employer's network can see that you connect to Tor, though not what for. On a small network, being the only Tor user at 9:14 is enough to identify you. One of the cases in chapter 16 is exactly this.
- Files opened outside it. A downloaded document opened in a normal app can contact the internet directly, with your real address.
- The browser gets changed. Extensions and custom settings make your copy of Tor Browser different from everyone else's.
Use Tor Browser as shipped, from torproject.org. Where Tor is blocked or you need to hide that you use it, a bridge is an unlisted entry relay that disguises the connection. It's a setting in Tor Browser.
Which to use
| YOU WANT TO | USE | WHY |
|---|---|---|
| Stop cafe Wi-Fi or your provider seeing which sites you visit | VPN, or secure DNS | Fast, works with every app. |
| Hide your home IP from sites, games and calls | VPN | Tor is too slow and is blocked by many services. |
| Read or research without it being tied to you | Tor Browser | No single party to trust, and a shared fingerprint. |
| Run a pseudonymous identity over months | Tor, plus chapters 9, 13 and 15 | The network is the easy part. Behavior is what gives people away. |
| Hide Tor use from your network | Tor bridges | A VPN in front of Tor also hides it, and adds the VPN company to the people who know. |
13
Hardened operating systems
An ordinary operating system is one big box: any program you run can usually read all your files. Hardened systems are built around the idea that something will get compromised, and they limit what it can reach.
| SYSTEM | THE IDEA | GOOD FOR | THE COST |
|---|---|---|---|
| Tails | Runs from a USB stick, sends everything through Tor, and forgets everything when you shut down. | One sensitive session on a computer you'd rather leave no trace on. | Nothing is saved unless you set up encrypted storage. Slow. |
| Whonix | Two virtual machines. One does nothing except route traffic through Tor. You work in the other, which can't reach the network any other way. | A long-running pseudonymous identity on your normal computer. | You have to manage virtual machines. The host system still has to be trusted. |
| Qubes OS | Every activity runs in its own isolated virtual machine, shown with a colored window border. | People who open risky files for a living and need strong separation on one laptop. | Needs suitable hardware and patience. A steep first month. |
| GrapheneOS | Android with stronger defenses against exploits, apps kept in tighter sandboxes, and no Google services unless you install them in a sandbox. | A phone that resists both data-hungry apps and forensic tools. | Runs on Google Pixel phones today, with Motorola devices announced in 2026. A few banking apps refuse to work. |
How Qubes splits a laptop
If a malicious PDF takes over the disposable box, it finds nothing there, and the box is deleted when you close the window. The password vault was never reachable, because it has no network and shares no files.
The same idea without changing system
- Open unexpected attachments in the browser's built-in viewer or an online viewer rather than the desktop app.
- Use a standard user account day to day and keep a separate admin account.
- Use a virtual machine for anything risky and restore it to a clean snapshot afterwards.
- Keep a separate cheap laptop for the bank and the admin consoles if you run a business.
Below the operating system
Disk encryption protects files, but the firmware that starts the computer can be tampered with too. Keep firmware (BIOS or UEFI) updated, leave Secure Boot on so only signed systems start, and set a firmware password so nobody can boot from their own USB stick. Devices with a dedicated security chip, such as recent Macs, iPhones and Pixels, enforce limits on password guesses in hardware, which is what makes a short code survivable.
14
Keys and encryption
Modern encryption isn't what breaks. What breaks is how the keys are created, stored, shared and retired. This chapter is about that.
Two kinds of encryption
Symmetric encryption uses one secret for both locking and unlocking, like a house key. It's used for disks and files. Asymmetric encryption uses a pair: a public key you can give to anyone, and a private key you never share. What one locks, only the other opens. It's how two people who've never met can talk securely, and how software and messages are signed.
Everything depends on the private key staying private. The safest private key is one that can't be copied at all, because it lives inside a hardware security key or the phone's security chip and only ever answers questions.
SSH and code signing
If you administer servers or write code, your SSH key is as valuable as an admin password. Generate a modern one with a passphrase:
ssh-keygen -t ed25519 -a 100 -C "laptop-2026"
Better, create one that only works while your hardware key is plugged in and touched. Malware on the laptop can't copy it:
ssh-keygen -t ed25519-sk -O resident -C "hardware-key-1"
The same key can sign your Git commits, so others can check that a change really came from you:
git config --global gpg.format ssh
git config --global user.signingkey ~/.ssh/id_ed25519_sk.pub
git config --global commit.gpgsign true
Encrypting files
- A single file to send or store:
ageis a small, modern tool.age -p secret.pdf > secret.pdf.ageencrypts with a passphrase. 7-Zip with AES-256 and encrypted file names also works and is easier for the person receiving it. - A folder that travels: a VeraCrypt container, or an encrypted disk image on a Mac.
- Cloud storage: Cryptomator encrypts files on your device before they sync, so the cloud company holds only scrambled data.
- Email: PGP still exists, and it's fragile in practice. Use it where a contact requires it.
Secrets in code
API keys and passwords committed to a code repository are one of the most common ways companies get breached. Bots scan public repositories around the clock and find a published key within minutes.
- Keep secrets out of files that are tracked. Use the operating system's keychain, a secrets manager, or environment variables injected when the program runs.
- Run a scanner such as gitleaks or TruffleHog before each commit, and turn on your code host's push protection.
- If a secret is ever pushed, rotate it first: revoke the old key and issue a new one. Deleting the commit isn't enough, because copies exist the moment it's public.
- Give each key the smallest permissions that work, and an expiry date where the service allows one.
Forward secrecy
With forward secrecy, each conversation uses temporary keys that are thrown away afterwards. If someone records your encrypted traffic today and steals your long-term key next year, they still can't read the recording. Signal and modern HTTPS have it. PGP email doesn't: one stolen key opens every message ever sent to it.
Quantum computers
A large enough quantum computer would break the asymmetric algorithms in use today. None exists yet. The present-day risk is harvest now, decrypt later: an adversary records encrypted traffic now and waits. For data that has to stay secret for decades, that matters already. NIST published the first post-quantum standards in August 2024, and Signal, iMessage and the major browsers already combine a post-quantum algorithm with a classical one. Symmetric encryption with 256-bit keys isn't seriously affected. For you, the practical action is to keep software updated so you get these changes as they roll out.
A key's life
- CreateOn the device that will use it, or inside hardware. Never have someone email you a private key.
- Back upRecovery codes and a second hardware key, stored in a different place from the first.
- RotateReplace keys on a schedule, when someone leaves, and immediately after any suspicion.
- RetireRevoke it everywhere it was trusted. A forgotten old key on a server is an open door.
15
Fingerprinting and behavior
Once the network address is hidden, what identifies you is your equipment and your habits. Careful people get caught here.
Browser fingerprinting
A website can read your screen size, time zone, language, installed fonts, graphics card, and how your browser draws a hidden test image. Combined, these often form a fingerprint that's unique among millions, and it works with cookies cleared, in private mode and behind a VPN. The EFF's Cover Your Tracks site shows you yours.
There are two defenses. One is to look like everybody else: Tor Browser and Mullvad Browser give every user the same fingerprint, which only works if you leave the settings alone. The other is to give different answers each time, which is what Brave does. Adding privacy extensions to an ordinary browser usually makes you more distinctive, because few people have that exact set.
Writing style
Stylometry identifies an author from their writing: favourite words, sentence length, punctuation habits, recurring spelling mistakes, British or American spelling. With a few thousand words of known writing to compare against, it's accurate enough to pick one author out of a large group. If two identities must stay apart, they shouldn't write the same way, and the simplest route is for one of them to write very little.
Patterns in time
- Posting hours show your time zone, your working day and your holidays.
- A pseudonymous account that goes quiet exactly when you're on a flight, at a wedding or in hospital narrows the list of candidates.
- Two accounts that are never active at the same moment, and often active one right after the other, look like one person.
None of these is proof alone. Investigators and determined amateurs build a case by stacking many weak signals.
Redaction
People publish secrets they believed they'd hidden.
- Black boxes in a PDF. A rectangle drawn over text leaves the text underneath, ready to copy and paste. Use a real redaction tool, or flatten the page to an image.
- Blur and pixelation. Blurred text and numbers can often be recovered by software. Cover with a solid block.
- Cropping. In 2023 a flaw nicknamed aCropalypse meant cropped screenshots from Google Pixel phones and the Windows snipping tool still contained the cropped-away part. Fixed since then, and a good reason to re-export an image before sharing.
- Highlighter in a phone's photo editor. It's partly transparent. Turn up the contrast and the text shows through.
- Document history. Tracked changes, comments and earlier versions stay in the file.
Printers and screens
Most color laser printers add a faint grid of yellow dots to every page that encodes the printer's serial number and the time of printing. Organizations also log who printed what and when. Assume a printed document can be traced back to a machine and a login.
Money
Card payments tie a purchase to your name. Bitcoin isn't anonymous: every transaction is public forever, and exchanges that check identity tie addresses to people. For a purchase you want kept apart from your name, such as a domain or a VPN, a prepaid voucher bought with cash is simpler and more private than most cryptocurrency.
16
Seven real failures
Most of these people were criminals, and the reason to study them is that their mistakes are so well documented in court records. What unmasked them is the same thing that would unmask a whistleblower, a journalist's source or you. In every case the strong technology held, and an ordinary slip did the damage.
| WHO | WHAT THEY USED | THE SLIP | THE LESSON |
|---|---|---|---|
| Ross Ulbricht, Silk Road, 2013 | Tor onion service, Bitcoin, a pseudonym. | Years earlier he had promoted the site on a forum under a username, then used that username to post a job ad with his personal Gmail address. | The first days of an identity count as much as the last. Old posts don't expire. |
| Hector Monsegur, “Sabu”, 2011 | Tor for all chat connections. | Press reports at the time said he logged in to a chat server without Tor once, which showed his home IP address. He disputes that account and says he was identified from personal details leaked years earlier. | A rule that depends on remembering will fail once. Make the safe route the only route. |
| A Harvard student, 2013 | Tor and a disposable email service, to send a bomb hoax and get out of an exam. | He used Tor over the university's Wi-Fi. Network logs showed who on campus was connected to Tor in that window. He was one of very few. | Tor hides what you do. It doesn't hide that you're using it. A small crowd is no cover. |
| Alexandre Cazes, AlphaBay, 2017 | Tor onion service, cryptocurrency. | Early welcome emails from the site carried his personal Hotmail address in the headers. | Metadata and default settings leak. Read what your systems send. |
| “Guccifer 2.0”, 2016 | A VPN, to pose as a lone Romanian hacker. | Reportedly the VPN wasn't switched on for one login, leaving a Moscow address in a company's logs. | Anything optional gets skipped once. |
| Reality Winner, 2017 | She printed a classified report and posted it to a news outlet. | The outlet showed a copy to officials to verify it. The page had been printed, the print log listed six people, and one of them had emailed that outlet from a work computer. | Access logs narrow the field. Then one extra link finishes it. |
| Strava users, 2018 | Nothing. They went running. | A default setting published their routes. | You can leak without doing anything. Check the defaults. |
The patterns
- Contamination at the start. The real identity touched the new one on day one, before the person took it seriously.
- Protection that could be forgotten. Protection that depends on remembering a step gets skipped eventually. Build it so the unsafe path doesn't exist: a machine that can only reach the internet through Tor, a browser profile that only holds one identity.
- A small crowd. Being one of a handful of people who could have done something is nearly the same as being named.
- Someone else's records. A forum, a mail server, a printer, an employer. You control your own device and none of the others.
- Time. An investigator needs one mistake, and can wait years for it or go back years to find it.
PART 4
TEAMS AND COMPANIES
A company's OPSEC is the sum of everyone's habits plus a few written rules. This part covers what a business leaks, how to notice an attack early, and what to do in the first hours after something goes wrong.
17
OPSEC for a company
Attackers research a company before they contact it. Almost everything they need is published by the company itself or by its staff.
What a company leaks
| SOURCE | WHAT IT GIVES AWAY | WHAT TO DO |
|---|---|---|
| Job adverts | The exact software, versions and security products you run. | Describe skills. Leave product names for the interview. |
| Who works in finance and IT, who reports to whom, who started last week and doesn't know the procedures yet. | Warn new starters that fake “welcome” messages from the boss are common in week one. | |
| Public code repositories | Keys, internal hostnames, staff email addresses in the commit history. | Secret scanning (chapter 14). Review what's public every quarter. |
| Certificate transparency logs | Every HTTPS certificate is recorded in a public log. Search yours and you'll see every subdomain, including staging, vpn and old-admin. | Search your own domain at crt.sh. Shut down what you'd forgotten. Don't count on a hostname being secret. |
| DNS records | Verification entries name the cloud services you use, which tells a phisher which login page to imitate. | Remove stale records. Assume your vendor list is known. |
| Out-of-office replies | Who's away, until when, and who covers for them. | Send details to internal senders only. |
| Published documents | Author names, usernames, software versions, file paths. | Strip metadata before publishing (chapter 7). |
| Photos from the office | Whiteboards, screens, badges, sticky notes, the model of the door lock. | Look at the background before posting. |
| Public boards and shared links | Task boards, documents and cloud folders set to “anyone with the link.” | Audit sharing settings. Make private the default. |
A critical information list
Step one of the five-step process, done for a business. Write one page listing what an outsider must not learn, who's allowed to know each item, and where it lives. Typical entries: customer and supplier bank details, who can approve payments and up to what amount, admin credentials, client data, unreleased prices and deals, staff home addresses, the layout of your network. Share the list with the team. People protect what they've been told matters.
Rules that prevent the expensive incidents
- Payment changes are confirmed by phone. Any new payee or changed bank details gets a call to a number already on file, made by someone other than the person who received the request. This one rule defeats most invoice fraud.
- Resets require proof. Whoever resets passwords, in-house or at your IT provider, verifies identity by a method agreed in advance, such as a call back to a known number or a video call with a manager. A caller in a hurry gets no exception.
- Phishing-resistant login for admins. Passkeys or hardware keys on email, the domain registrar, banking, payroll and every admin console.
- Least privilege. Each person has the access their job needs and no more. Admin rights are a separate account that isn't used for email or browsing.
- Joiners and leavers. A checklist for both. On the day someone leaves: disable accounts, remove them from shared mailboxes and password vaults, collect devices and keys, and change any shared passwords they knew.
- Company data on managed devices. Encrypted, updated, and erasable remotely if lost.
- Email authentication. SPF, DKIM and DMARC are three DNS records that stop strangers sending mail that appears to come from your domain. Set DMARC to
p=rejectonce legitimate mail passes. - Suppliers. List which vendors hold your data or have access to your systems. Their breach becomes yours. Ask how they log in and what they do when someone leaves.
- No blame for reporting. The person who says “I think I clicked something” within five minutes saves the company. If reporting gets people punished, they stop reporting.
Need to know
Information should reach the people who need it for their work and nobody else. It can feel unfriendly in a small team. In practice it means a salesperson's hacked laptop exposes sales files and leaves payroll alone, and a phished intern can't approve a transfer.
18
Knowing when you're being targeted
Many attacks go unnoticed for weeks. A handful of cheap tripwires and a short list of signs bring that down to hours.
Tripwires
- Login alerts. Turn on notifications for new sign-ins on email, bank, cloud and social accounts. Read them when they arrive.
- Breach notifications. Subscribe your addresses, or your whole company domain, at Have I Been Pwned.
- Unique aliases. When the alias you gave to one shop starts getting phishing, you know that shop leaked or sold it.
- Canary tokens. A canary token is a fake file, link or key that emails you when someone touches it. Put a document called
passwords.xlsxon the shared drive, or a fake cloud key in a code repository. Nobody has a legitimate reason to open them, so an alert means someone is looking around. They're free at canarytokens.org. - Search alerts. A search alert on your name, company name and domain catches impersonation and doxxing early.
- Bank alerts. A notification for every transaction on business accounts.
Signs an account is compromised
- Password-reset emails you didn't ask for, or login prompts you didn't start.
- Sent messages you didn't write, or contacts replying to something you never sent.
- Mail that should have arrived and didn't. Check for forwarding rules and filters that send messages to the archive or the trash.
- Devices, sessions or connected apps you don't recognize in the account's security page.
- Changed recovery details.
Signs a phone is monitored
- Someone knows things they could only know from your messages or your location.
- Apps you didn't install, or an unknown app with accessibility or device-admin rights on Android.
- A configuration profile or device-management entry you didn't add on iPhone.
- An unfamiliar device in your Apple or Google account, or location sharing switched on with someone.
- The phone was out of your hands, unlocked, with someone who might want to do this.
Battery drain and heat are poor signals. Phones do that for plenty of ordinary reasons, and good spyware doesn't. Apple and Google both run built-in checks: Safety Check on iPhone and Play Protect on Android. If you believe a partner or family member is monitoring you, read the note in chapter 2 before removing anything.
For state-grade spyware, Apple sends threat notifications to users it believes were targeted, and Amnesty's Mobile Verification Toolkit can examine a phone backup for known traces. If you get such a notification, take it seriously and contact one of the organizations in chapter 2.
Signs you're being researched
- A burst of connection requests from thin new profiles, or strangers asking colleagues about you.
- Calls to the office asking who handles payments or IT.
- Password-reset attempts across several of your accounts in a short time.
- A newly registered domain that looks like yours with one letter changed.
These often come a few days before the real attempt. Tell the team when you see them.
19
When it goes wrong
The first hour decides how bad it gets. Have the order written down in advance, on paper, because the device you'd read it on may be the one that's compromised.
An account has been taken over
- Use a clean deviceIf the computer might be infected, a new password typed on it is stolen too.
- Email firstNew password, sign out every session, then check forwarding rules, filters, connected apps and recovery details.
- Password managerNew master password. Review which devices are signed in.
- MoneyCall the bank on the number printed on your card. Freeze cards. Go through recent transactions.
- Everything elseWork through the accounts that shared the password or can be reset from that inbox.
- Warn peopleTell contacts that messages from you may be fake. Report it to the platform and, for money, to the police.
Email comes first because of the diagram in chapter 3. Until the inbox is yours again, the attacker can undo every other change with a reset link.
A device is lost or stolen
- Mark it as lost with Find My (Apple) or Find Hub (Google). That locks it and shows a message.
- If it wasn't encrypted, or was unlocked when taken, start a remote erase and change the passwords of accounts that were signed in.
- Remove the device from your accounts so its sessions stop working.
- If it held your authenticator app, use backup codes to get in and move the second step to a new device.
- Tell your carrier to block the SIM. Report the theft with the IMEI number.
A key or password has leaked
Rotate first, investigate second, tidy up third. Revoke the old key and issue a new one. Then read the service's logs for what the old key was used for while it was exposed. Clean it out of code history last.
You've been doxxed
Doxxing is someone publishing your private details, usually your address, to invite harassment.
- Record it. Screenshots that show the address of the page and the date. Platforms and police both ask for this, and posts get deleted.
- Report it to the platform as a privacy violation. Most large ones ban posting private addresses. Ask search engines to remove the results.
- Lock down. Set accounts to private, check logins, and warn family members, who often become the next target.
- Tell your employer if they may be contacted or sent false complaints.
- Contact the police if there are threats. In countries where hoax emergency calls are used to send armed police to a victim's home, tell your local station in advance that you may be targeted.
- Stay out of the thread. Replies feed it. Ask a friend to monitor it so you don't have to read it.
A company incident
- Contain. Disconnect affected machines from the network. Leave them switched on if you can, because memory holds evidence.
- Keep evidence. Don't wipe and reinstall until someone has decided what needs keeping.
- Talk on another channel. If email is compromised, the attacker reads the discussion about the attack. Use a group chat set up beforehand.
- Call for help early. Your IT provider, your cyber insurer's hotline if you have one, and your national cybercrime reporting service.
- Check your legal duties. Under the GDPR, a breach of personal data that puts people at risk must be reported to the data protection authority within 72 hours of your becoming aware of it.
- Tell the people affected clearly and early. Say what happened, what data was involved and what they should do.
- Review it afterwards. Work out how they got in and which control would have stopped it, and fix that. Leave blame out of it.
20
Keeping it going
OPSEC is something you keep doing. A setup that was right two years ago has gaps today, because your life, your tools and the attacks have all moved.
Do less, and keep doing it
The most common failure is tiredness. Someone builds an elaborate system in a burst of worry, finds it exhausting, and drops all of it, including the parts that mattered. Match the effort to the threat model from chapter 2. A password manager and a hardware key you actually use protect you better than a hardened laptop left in a drawer.
A review every three months
Half an hour, in the calendar.
- Has anything changed in what I'm protecting, or from whom? A new job, a new business, a breakup, a move, a public profile.
- Check sessions, connected apps, forwarding rules and recovery details on the main email account.
- Restore one file from each backup.
- Search my own name and address again.
- Check that every device still receives updates.
- Remove apps, extensions and accounts I stopped using.
- For a company: compare the list of people with access against the list of people who work here.
The people around you
Your security depends on the people who hold information about you: family who post photos, a partner who shares your streaming password, a colleague with access to the shared drive. Helping them with part 1 protects you as well. Start with their email account and a password manager.
The whole course in six lines
- Decide what you're protecting and from whom before choosing any tool.
- Lock down the email account that can reset everything else.
- Use login methods that a fake page can't capture.
- Assume what you publish stays public, and that separate pieces will be put together.
- Keep different parts of your life in different boxes, and never let two boxes touch.
- Build your setup so that the safe way is the only way, because anything optional gets skipped once.
A
Glossary
- 2FA / MFA
- A second proof at login besides the password: a code, a prompt, a passkey or a hardware key.
- Adversary
- Whoever you're protecting something from.
- Adversary-in-the-middle (AitM)
- A phishing setup that relays the real login page and captures the session.
- Alias
- A forwarding email address used for one purpose, which can be switched off.
- Bridge
- An unlisted Tor entry relay, used where Tor is blocked or its use needs hiding.
- Business email compromise (BEC)
- Fraud by email that impersonates a boss or supplier to redirect a payment.
- Canary token
- A decoy file, link or key that alerts you when someone uses it.
- Compartmentalization
- Keeping activities in separate boxes so one failure doesn't spread.
- Credential stuffing
- Trying email and password pairs from one breach on many other sites.
- DMARC
- A DNS record that tells receiving mail servers what to do with mail that fakes your domain.
- DNS
- The system that turns a site's name into a numeric address.
- Doxxing
- Publishing someone's private details to invite harassment.
- End-to-end encryption (E2EE)
- Encryption where only the participants hold the keys, and the service can't read the content.
- Evil maid attack
- Tampering with an unattended device to capture its password later.
- EXIF
- Data stored inside a photo file, including time, camera and often location.
- Fingerprint (browser)
- The combination of browser and device traits that identifies you without cookies.
- Forward secrecy
- Use of temporary keys so a key stolen later can't open past conversations.
- Full-disk encryption
- Encryption of everything stored on a device, unlocked by your password.
- Hardware security key
- A small device that proves it's you at login and can't be phished.
- IMEI / IMSI
- The permanent numbers of a handset and of a SIM, both seen by the mobile network.
- IMSI catcher
- A fake cell tower that collects the identities of phones nearby.
- Metadata
- Information about a communication: who, when, where, how much.
- Onion service
- A site reachable only inside Tor, at an address ending in .onion.
- OPSEC
- Operations security: finding and closing the ways information about you leaks.
- OSINT
- Open-source intelligence: information gathered from public sources.
- Passkey
- A login credential stored on your device that replaces the password and can't be phished.
- Phishing
- A message built to trick you into giving access or money.
- SIM swap
- Fraudulently moving your phone number to an attacker's SIM.
- SNI
- The field that reveals a site's name at the start of an HTTPS connection.
- Social engineering
- Manipulating a person into doing something that helps the attacker.
- Stalkerware
- A monitoring app installed on someone's phone by a person with physical access.
- Stylometry
- Identifying an author by analyzing their writing style.
- Threat model
- Your written answer to what you protect, from whom, and how far you'll go.
- Tor
- A network that routes traffic through three relays so that none knows both source and destination.
- TOTP
- Time-based one-time password: the six-digit code from an authenticator app.
- Traffic correlation
- Matching traffic at two points by timing and volume to link sender and destination.
- VPN
- An encrypted tunnel to a provider's server, which then connects to the internet for you.
B
Further reading
- Surveillance Self-Defense, from the EFF. Step-by-step guides by tool and by situation.
- Privacy Guides. Community-reviewed tool recommendations with the reasoning shown.
- Tor Project support. What Tor does and doesn't protect, from the people who build it.
- NIST SP 800-63. The US standard on passwords and authentication.
- Have I Been Pwned. Check and monitor addresses for breaches.
- Cover Your Tracks. Test your browser's fingerprint.
- Canarytokens. Free tripwires.
- Access Now Digital Security Helpline. Free help for journalists, activists and civil society.
Liaison is a weekly email about the security flaws and scams that affect the software small businesses run, with what to do about each one.
GET THE WEEKLY EMAIL