Schneier on Security

Clive Robinson • September 29, 2026 6:26 PM

@ Billy Jack, ALL,

With Regards,

“I know some people who are going to hate learning that Signal is not the ultimate in security.”

It never was and I said as much, much to many peoples surprise.

The point that every one was forgetting is that Signal and all the other “security apps” were just one small part of the system. Just one link all be it a strong one in a chain that had very weak links.

As our host @Bruce had noted some years before Signal was even named,

“A chain is only as strong as it’s weakest link”

Or more technically the design of smart phones as such is that,

“You can from the communications end point, reach around to the user interface beyond the security end point.”

Thus no matter how strong and clever the cryptography and other security features built into the Application are, an attacker can simply bypass it all from the “Over The Air Interface” directly to what the user sees and types etc of the User Interface.

It is as true now as when I said it on this blog some years ago…

Can it be changed?

Yes, but it won’t be convenient for the user… And as “convenience” is what the majority of smart device / mobile phone users want it won’t get fixed…

Because what you have to do is move the security end point beyond the reach of the communications end point. And that cannot be done securely or reliably on a commercial or consumer grade device as the fundamental design will not support it “on device”.

Which means you have to have another device that takes the security end point and following user interface off of the smart device / mobile phone…

Fun thing is you can do it using a pencil and paper with a “hand cipher” implementation of Claude Shannon’s “perfect secrecy”.

You will find I’ve given details of how to do this, in quite some detail on this blog a couple of times in the past. Including how to do it in a deniable form.

Do I think anyone actually does that?

No of course not…

Why?

Because it’s just too inconvenient to do…

Which is just what the Guard Labour and other authoritarian entities want you to do, due to the usefulness of attacks like the one this thread describes.

However the new game in town is not defeating “End to End Encryption”(E2EE) because even the US SigInt agencies and FBI have admitted that it is to important.

No… the new game was actually started by Apple to try and stop the “think of the children” type attacks that were very clearly going to be Guard Labour and Authoritarian agency next type of attack via CSAM accusations.

So Apple came up with the first “Client Side Scanning” that ran on your smart device / mobile phone and if it recognised certain types of flesh coloured images it would report them back…

The thing is once you have the “Client Side Scanning” on your device effectively embedded in the OS it can look for anything not just CSAM.

So text messages and the like or anything else visible in the device User Interface gets scanned.

Microsoft has gone way further with it’s AI nonsense and “ET Phone Home” every few seconds with everything on screen, “just so you don’t forget it” and Microsoft steal all your IP to put in their ML systems…

The only way to defeat “Client Side Scanning” is with “off device security end points”… Unfortunately, sending ciphertext is a “red flag” and any half assed Client Side Scanning will spot it and report it.

It’s one of the reasons I came up with “deniable encryption” that just sends what looks like “ordinary plaintext” not the gobbledygook looking “ciphertext”.

It also has other advantages as it helps protect a communicating “first party” from betrayal by the “second party” to a “third party” such as Guard Labour and the like.

The funny thing is that quite a while after I first posted how to do “deniable encryption” somebody showed how to do it with quirks in AI.

This in turn tells you why “prompt injection” attacks will always be able to bypass Guard-rails on both the Inputs and outputs of LLMs.

But… It also shows by the same logic, that AI agents will always be able to escape/bypass sandboxes.

So yes I can see all sorts of problems ahead and no way to stop them unless AIs are run fully segregated on private systems. Which is not what the “investors” of “AI for rent” companies such as Anthropic, OpenAI, etc want you hearing as the ROI flies out the door.

Likewise the other companies such as Google, Microsoft, Meta, etc that are reliant on “spying on users” and “selling them on as product” want you to hear either again because their Shareholders will not want to see share price and dividends to drop either.

If you look at what has happened to Oracle in recent times you can see what might happen to other large Silicon Valley Corps. And worse what that will do to the US economy….

Clive Robinson • September 30, 2026 6:59 AM

@ Anonymous, ALL,

With regards,

“I’m assuming that a series of apps will emerge which provides encryption/decrpyption of custom text. Messaging apps like Telegram, Signal, Whatsapp will just become a pipe of encrypted base64 text what will be decrypted by the third app on receiving.”

There is already a way to do this that is cryptographically secure, uses just paper and pencil, and importantly looks like inane “plaintext” language not suspicious base64 or other coding.

Search this site for my name and deniable encryption to read how to do it.

But remember your modile phone or smart device is a pile of monkey crap when it comes to security.

Any security app on it can be easily bypassed through the OS so an attacker can over the air bypass a security app and go directly to what the target sees and types.

So you have to get things “off device” and the advantage of paper over electronics is it only needs a match and a few seconds[1] to destroy it beyond usable evidence.

[1] Look up “flash paper” it’s used by magicians as “part of the act”. There are two types, one that is nitrocellulose based and the other cellulose washed in nitrate. The first uses sulfuric acid and should be avoided as it is a low grade explosive and is without care poisonous. In essence the second sort is very fine paper that has been washed with food grade nitrates as an oxidizing agent to make the paper burn very very quickly and completely to ash finner than dust. The reason for “food grade” chemicals is that at a push you can swallow the paper instead (ie same nitrates as used in curing meat).

Though I would not swallow some nitrates as they are also heart medications and can give you a “Nitrate Headache” that anyone who has used a little to much “NG Spray” under the tongue can tell you all about as “migraine city” or “suicide headache”.