Computer security / safety

This are only notes from Nov 2024, and may be extended later.

Exclamation marks are used to express cynicism.
Most people underestimate the complexity of security. Most website authors simplify the security thematic so hardly, that the outcomes are wrong and make security weaker for you. I will show you why I think another way. I concentrate on RSA-like asymmetric algorithms, because you need that for world wide safe communication.

Security - fundamentals

Todays (2024) security is based on asymmetric encryption (including electronic signatures). The encryption algorithms are not proven safe, but they are believed to be hard to break. But there are some "small" assumtions which must be fullfilled to reach the hard-to-break state.

• your "privat" key must be privat - this is a very difficult problem for most people. All (!) your hardware and software on the encryption device which can access the privat key must be trustfull and errorfree. To have a chance, use minimalistic systems with minimum complexity for encryption devices.
• your "public" key must be public, which means, nobody should be able to take over your identity and mimic its public key as yours and do a man-in-the-middle attack.
• Of course you need help to do the crypt math. This is done on a reliable encryption device. This device is assumed as safe. That means it always does only the intended encryption math and algorithms not something different.

This has some "minor" consequences ...
Your personal computer (PC) is a bad device for encryption. If you not belong to the privileged group of people which spend half of there lifetime to explore the secrets of computing, you are depending completely on your providers and its quality.

What mostly goes wrong?

On SSH (secure shell) on the first connection, the software presence you the fingerprint of the servers public key and requires you to check its correctness. I do not know anyone how does this always and often wether the fingerprint is publicised nor the servers admin is reachable for asking directly. This method follows the Trust On First Use (TOFU) principle. But after that, ssh public key is saved localy and you are save until the algorithm is known to be broken for that keylength someday in the future or the server updates its OS (Operating System) updating and not backup and restore its old key pair.
For TLS (Transport Layer Security) used in the world wide web, another way was choosen. Here we have electronic certificates which give garantie for correct public keys. The certificate is created by a verifier and is finaly a electronic signature which must be checked too. This can be done by a higher trust instance, so you have chains of trust. Finaly there is still the problem, that you have to check one of this chain member signatures and additionally have to trust the lower chain members to verify correct. Very likely you will delegate this task to your software vendor. The problem is, that as more institutions and people are involved in that trustchain, the more errors or bad manipulations can occur. The first way to fix this problem, was to add a way to revoce a signature, but that involved a complex way to distribute revocations with a lot of bad site effects (later OCSP etc, see EFAIL 2018 certs CRL as malicious back-channel). At last the browser industry tries to fix it with shorter lifetime of certificates, which has also its problems. Today the short lifetime mostly brakes the original way to save the public key of the server and compare against it. This is because your browser aves the certificate which is the public key plus some attributes. One of the attributes is the lifetime. Automaticly updated certifcates have a lifetime of 3 months only. It tells you or your software to throw the certificate away and expect a new one. The new one could have the same or another keypair and this will be accepted by your browser before the old certificate will expire. It again fully trusts the certificate chain and never presents you a warning if the public key (the inner cryptography core) have changed. Today (2024) the distributers put a couple of (long living) root or authority certificates together with their software. I count 353 default Authorities within Firefox Certificate Manager from all countries or companies over the world. The security is as weak as the weakest authority in that list. A weak authority may be likely, if this weak authority uses systems managed by other vendors which purchase services from further vendors. One of this vendors may depend on cloud providers and than another inextricable worldwide tree you depend on.
If one of this authority companies will fail, by beeing hacked, for example, it does not mean that it will deleted from that list. To-big-to-fail companies are left in that list to minimize the worldwide trouble, causing such a incident. so there are political decisions which overgo security rules. This is why for my opinion, TLS is proved to be broken by design.
So its a security vendor tree where the weakest point you depend on. Also your OS or Software may update at any time you are online and have automatic updates enabled, or let vanish one of that authorities seconds after used in a bad way.

The TLS system is so broken, that certification authorities must be notify the big browser companies for every new certificate to give them a chance for extra checks. This is mostly because their own web-certs could be taken over. This was already happen. If you wonder why your new installed server is been scanned immediately after getting a web certificate, may be its because of that report chain.

The jabber.ru-hack 2023, a man-in-the-middle attack with regular letsencrypt-certs also shows, that TLS and its certificate trust world is broken.

I suggest changed browser software. Throw the default root authority certificates away and let the user verify the fingerprint of the public key on first visit. The websites must publicise there fingerprints and hold the public keys as long as possible and sign new keys with the old one. I think there is no good way to revoke leaked keys. The new website should do it like pgp with an advanced generated revoke-cert which is send directly to the users browser which has to store it together with the stored public key. This is more simple than the current cert chain world with less dependencies.

inlined external web-content - correct tls-connection?

firefox: ctrl-mouse-right + Q network reload Domain.lock-symbol Fingerprints,noPubkey

When you browse a website you get a lock symbol in the headline for that website. It is colored and you may click on them to get more information about the connected security. But what about the inlined code from other websites (use noscript to be informed about the fact, that there are such sources)? I see no easy way. So simply distrust websites, which do include sources from other sides. I know the web is full of such sh*t.

automatic software updates - safe?

An update is a chance for the volatile in-memory malware to silently go permanently!

You may know the "rule" Update often, update immediatly, no excuse. But this is only a good rule within a trust-your-vendor idealized world. But your vendor may first interested in making money our saving money, security interests are more or far later or are skipped for money efficients reasons even for companies naming their products as security products. Security products mostly are snake oil products, sometimes creating security holes wider and more catastrophic than the original system has had. Today we have the vendor attack-chains and some "minor" design decisions against security.
There is another rule: Software has always (security) bugs. This is not exact true, but the bigger the software, the more likely bugs are in there. Simple software can made safe. Software which has security in mind is therefore made after the KISS principle. That means "Keep it simple, stupid!" and other similar phrases pointing to simplicity as design principle for safety in all forms (for aircrafts, spacecrafts ... and ... software).
If you want to make security you need KISS systems. If you do not, you have to do updates as a way to fix the security holes inside your system ... and create new holes!
Even the update itself can be the hole. Todays updates are mostly cryptographical secured with a attached signed hash sum. The updating software has a public key to verify the signature. So far so good. But if you need the update for security reasons, there is a hole which possibly already was used and you can not trust the update process anymore.
Do not underestimate this. The internet is faster than ever and your system is surely put in one of the databases for marketing reasons where all accessable data is stored too with minimum your browser and OS version and selled to other money makers down to the dark site of the net. If a security breach method becomes known to the web an army of interested people sends code to the bots and hunts for every accessible system using all available databases. Everything may happen within seconds.
If you have a read-only system media (KISS) or the trusted platform functionality (TPM - the complex way with its security holes) you must reboot before doing the update! Your system could be infected in memory and you have to reboot in a clean state before update. Booting after the update is useless from security point of view. After that you need a safe (bug-free) update-process where you must set your system-media temporarly to read-write mode (KISS) or trust the software do the complex resigning of the boot system if that is included.
If you haven't that read-only functionality, you must reinstall your system to a fresh state or the last backup after the last update before starting the update process.
You do not go this way? Congratulation! Your system may still be infected and is not safe.
If your system is always online and attackable? Congratulation again! You need to be offline during reboot or reinstallation und during the update process.
Additionally some vendors deliver you broken updates, which let your system in an unusable state. Welcome to the real life. I guess this vendors do not follow the KISS way. Unfortunately even the big linux distros leaved the KISS way and replaced initV (KISS) by systemd (not-KISS). So if the systemd updates itself (to fix its bugs) and you have not enough memory or disk, you will left with an unbootable system. Well done!

If you need help, search people which use KISS-software (p.e. linux without systemd). You may recognize them if you got HTML-free emails (KISS) with no attachements from them.

If you do not know, which system is more KISS than another, look at the overall system size. Smaller is more likely simpler.

Separation as security method

If you want to build a secure device, take a old computer, add a systemd-free linux distribution like devuan from www.devuan.org to it, install gnupg and openssl and take it offline or use a bootable memory stick with a similar system. Transfer simple formats (ASCII-text is the best) to that offline system for encryption/decryption via writable transfer-stick.
Do not use complex file formats (like pdf) for signing, make detached signatures (its a signed file-hash only). You must be able to verify the text or file or its hash which you want to sign or encrypt. This is against unwanted changes. Therefore encryption devices need to have a human readable display and at minimum a yes/no keyboard.
Such an extra encryption device is the way using separation principle for security. The extra device is the secured zone which does all encryptions or signing. If you do this, you go from 1% safety to 99% safety. The often proposed two-factor authorization is more like going from 1% safety to 2% safety. Its only for the web-login-safety as a small part of the overall safety.

For webtraffic I do not know a practical solution. Install NoScript-plugin and disable all default trusted website entries as a minimum. Never execute code (JavaScript et al.) from the web. Also disable all unneeded webbrowser functionality which causes additional traffic for marketing and pseudo-safety which is difficult. Most webbrowser are the opposit to KISS and websites do there own, to require bad functionalities. If you need that bad websites, access them from extra machines made for the dark world (bootet from read-only media).