Cybersecurity

A lot of what goes wrong online never needs an exploit, because a reused password or a message that makes you panic will do the job. So that's mostly what I write about here, along with what software can and can't see.

Scams and takeovers

The “I accidentally reported you” scam

A sorry stranger warns you that a ban is coming, then points you to the one agent who can stop it.

Read the post

Takeover without prior intrusion

A site can take your account with no malicious code at all, as long as you've reused your main email password.

Read the post

The “just trying to make friends” dropshipping scam

A stranger wants to be friends, and half an hour later a sales screenshot has you opening a Shopify account.

Read the post

What software can see

What Chrome will and won’t let an app hide

Two screen-sharing restrictions shaped Perch, and I only found both by trying it the other way first.

Read the post

Every way to install a Chrome extension is closed except one

An afternoon of measuring showed which install routes Chrome has shut, and the one that's still open.

Read the post

Why loading an extension needs the DevTools pipe

Perch drives its own Chrome through a pipe rather than a port, which any local process could otherwise reach.

Read the post

My own tools

Threat model

This is how my privacy tools treat your data, tool by tool and in plain words.

Read the threat model

Pioneering

Here's the suite those tools belong to, along with the problem behind it and what I built.

Read the case study

Report a problem

If you've found a way one of my tools leaks, breaks or oversteps, please send me the details.

Email me
Step by step

How the same trick plays out

These two look nothing alike, and yet both end with you handing over your own account. Pick one to see how it goes, step by step.

Check a message

Paste a message that worries you and it gets checked against the patterns from the scams I've written up. Check here runs only in your browser, so nothing you paste leaves this page unless you also ask the AI.

0 / 4000

Ask the AI as well sends the message to Groq, which runs the GPT-OSS 120B model, through this site's server. Emails, phone numbers and code-like numbers are swapped for placeholders first, and nothing is stored here, but still leave out passwords and anything personal.

Checks

If a message worries you

The whole thing depends on you acting before you look, so these all come down to looking first.

Don't add anyone

Skip the “agent” they point you to, because real platforms don't run support through a stranger's DMs.

Don't share anything

That means no codes, no passwords, no tokens and no payments, however official the person sounds.

Check it yourself

Open the official app or type the address yourself, rather than following anything you were sent.

Report it, then block

Use the platform's own report option, which on Discord means holding the message and choosing Report Message.

Warn your friend

If it came from someone you know, tell them somewhere else, since their account may have been taken over.

Prepare in advance

Use unique passwords, keep your main email password to itself, and move two-factor authentication from SMS to an app.

Frequently asked questions

Does HTTPS mean a site is safe?

No, because HTTPS only shows that your connection is encrypted, so it doesn't tell you anything about who's running the site or what they do with your password.

Would Discord staff ever message me about my account?

Discord's help centre says staff won't contact you directly through the app for support matters, so anyone doing that in your DMs isn't Discord Support.

Can I trust it if the message comes from a friend?

Not automatically, because these messages often come from hijacked accounts. It's worth checking with your friend somewhere other than that account.

What if I've already handed something over?

Please don't be embarrassed, and move quickly. Change your password if you still can and warn your friends, and if you've been locked out of Discord, its recovery form is at dis.gd/hackedaccount.

How do I report a problem with one of your tools?

Email me the details along with the steps to reproduce it, using the button below. My threat model explains what each tool does and doesn't do with your data.

Found something wrong? Tell me.