All TLS/HTTPS clients have a set of Certificate Authority keys which they trust. Your client will only accept a public key which is signed by a trusted CA's key. A proper CA will not sign a key for a domain when it has not verified that the entity that wants it's key signed actually controls the domain.
this post was submitted on 21 May 2024
95 points (92.0% liked)
Technology
57904 readers
5618 users here now
This is a most excellent place for technology news and articles.
Our Rules
- Follow the lemmy.world rules.
- Only tech related content.
- Be excellent to each another!
- Mod approved content bots can post up to 10 articles per day.
- Threads asking for personal tech support may be deleted.
- Politics threads may be removed.
- No memes allowed as posts, OK to post as comments.
- Only approved bots from the list below, to ask if your bot can be added please contact us.
- Check for duplicates before posting, duplicates may be removed
Approved Bots
founded 1 year ago
MODERATORS
A proper CA will not sign a key for a domain when it has not verified that the entity that wants it's key signed actually controls the domain.
Most browsers trust many certificate authorities from all over the world.
Any of them could...
- be compelled by authority
- be compelled by threat
- be hacked
- have a lapse in ethics
- have a rogue employee
- etc.
...and yes, it has happened already.
HTTPS as most of us use it today is useful, but far from foolproof. This is why various additional measures, like certificate pinning, private CAs, and consensus validation are sometimes used.
I urge everybody to read up on CAA records in DNS and add them to your domains. They basically say what CA the certs for that domain are supposed to come from. Even if another CA issues valid certs for the domain they would be rejected if they don't match the CAA în DNS. It takes 5 minutes.
You can specify the valid CA in the form of its representative domain, for example to allow Let's Encrypt you'd add 0 issue "letsencrypt.org". If you want to allow multiple CA you add multiple CAA records. They enter into effect if at least one CAA record is present. You can also restrict the challenge type, for example 0 issue "letsencrypt.org;validationmethods=dns-01".
Please note that this is worth adding a CAA record even if you don't use your domain for HTTP and you don't issue any certs for it, because a rogue CA can do it for you. You can add a blank CAA record (0 issue ";") which basically forbids any CA.
(And yes, this also applies to email. It's worth adding restrictive records even if you don't use your domain for email.)
They mean CAA records:
https://developers.cloudflare.com/ssl/edge-certificates/caa-records/
Right, I'll fix it.
Thats why we now have certificate transparency reports and CA-records.
Sure not perfect, but at least with a compliant CA it wont just happen in the dark.
At some point you have to trust someone.
It is indeed true that some CAs have seriously misbehaved; however, browser builders are rather strict on the presence of the CAs they trust. Misbehaving or even simple errors are reasons for getting kicked out, after which certificates signed by those CAs are now no longer valid.
The certs are still valid.
They are just not implicitly trusted
You are technically correct, best kind of correct
Misbehaving or even simple errors are reasons for getting kicked out,
That can be helpful if a transgression is noticed, and it's not orchestrated by a higher authority (e.g. government), and the damage isn't already done.
browser builders are rather strict on the presence of the CAs they trust.
Of course, browser builders are vulnerable to influence, attack, accidents, questionable judgment, and blind spots just as certificate authorities are.
I agree, it’s far from perfect.
I'm somewhat skeptical. What if LetsEncrypt decided to misbehave tomorrow? Would the browsers have the guts to shut it down and break all sites using it?
Yes, they will. We’ve seen it before in mostly less serious cases: Diginotar, Türktrust, Symantec, etc. As brittle as the CA system can be, when there is real enough trouble, CAs do get revoked.
Not the browser companies. The parent CA would revoke Let's Encrypt's certs and publish that revocation in the certificate revocation list. When the browser (software, running on your system) downloads the new CRL, they will automatically stop trusting LE.
Many years ago I manually removed all CAs from my trust store and only added those I needed. Turned out that from roughly 160 trusted root CAs I needed about 10 to 12. I stopped because it often was very difficult to figure out which CA signed the cert for an app that was failing. The final nail in the coffin was when I was late for a business meeting and the only way to get a parking space close enough to my destination was by paying with an app I'd never used before and finding the right root CA under pressure was too much. I really wish we had more and easier control over who we trust.
It is possible and it has been done.
You need to get your "hacker" key signed/certified by an official CA. Which is not that difficult with some of them because they are doing it for money.
You don't really 'need to' in a world where a good proportion of people will happily click 'continue anyway' when they get any sort of certificate error
Thats why we have HSTS and HSTS preloading, so the browser refuses to allow this (and disabling it is usually alot deeper to find than a simple button to "continue anyways")
In Chromium browsers you can simply type "thisisunsafe" to bypass even HSTS failures.
people will happily click 'continue anyway'
Not possible without a certificate. There will be no TLS connection, only an error message. No 'click continue'.
It is trivial for an attacker to make self-signed TLS certs, and you can absolutely just click “continue” on sites that use them when you get a warning from the browser
What browser is that?
Firefox, Chrome, Edge, will all warn you about self-signed certs or cert mismatches but allow you to continue. You're completely correct that SSL/TLS needs a certificate, but it doesn't need to be CA issued or in any way legitimate for the encrypted tunnel to be established
I am personally using firefox and referencing my own servers that use their own self-signed TLS certs that I have not bothered to load onto my pc because they aren’t public, but chromium-based browsers aren’t some outlier here
Your own servers probably also dont have HSTS enabled, or clicking continue will be disabled (if not overwritten in your browser-config)
Reading the HSTS spec, it doesn’t work on first connection, and while most people are using websites they access more than once, that notably isn’t all web use.
Indeed, not classically, but there are HSTS preload lists you can put your domain into which will be downloaded by supported browsers.
And via HSTS you can include all your subdomains, which would then force proper TLS connections for those you havent visited before too.
With the new TLS1.3 version we are getting the HTTPS / "SVCB" Record which not only allows ECH but also indicates to the client similar protection policies like HSTS. (RFC 9460)
ECH will then make such attacks impossible on TLS-level, assuming DNSSEC is used and client can make an integrity-checked lookup e.g. via DoH/DoT or validating DnsSec themselves.
The strength of this depends on the security-chain you want to follow of course. You dont need DNSSEC, but then the only integrity-check is between DNS-Service and Client if they use DoH/DoT (which is usually enough to defeat local attackers)
It would be inherently impossible for HSTS to work on first connection, you are correct.
As others have mentioned, a trusted 3rd party signs the correct key so your browser can check the key itself.
However, it should also be noted that your browser must have a list of trusted 3rd parties and their certificates used for signing in order to perform this check. It's entirely possible to modify this list yourself. Some examples include:
- executing your own MITM style "*attack" in order to intercept and analyze local https traffic
- corporate network inspection and monitoring, where a gateway does the above for all devices on the network which have a CA cert pre-installed through some policy
So while it's possible for trusted 3rd parties to issue valid certificates to bad actors, it's also possible to add anyone (you, your employer, or some bad actors) to the trusted parties list.
Add Norton to that list. They also perform their own MITM attack on your pc to ensure your certificates are "safe"...
So there are two things that certificates are for. You already understand the first one, which is the cryptography itself. You can use them to encrypt your traffic so that information sent over the connection is not in plain text.
The second thing certificates do, is the answer to your dilemma. Identification. For your browser to trust a website’s certificate, the certificate has to be valid for that website. What makes a certificate valid? The certificate has to have been signed by a trusted Certificate Authority, and the name on the certificate must match the website you’re visiting. If you were to ask “What makes a certificate authority trusted?” The answer is that your web browser and/or operating system come preloaded with certificates for trusted certificate authorities. These special certificates were used to sign the certificate of the website you’re visiting, which is another thing your browser checks for. A malicious third party can’t (easily) obtain a valid certificate/key pair for a domain that they don’t own. If your browser was presented with a fake certificate from the malicious third party, it would not connect and would warn you that your connection isn’t secure and would explain why.
Now if more specifically, you’re wondering that if a malicious third party takes any given website’s public certificate, can it use that to decrypt your session? After all, that public cert is signed and trusted. The answer to that, is that when a certificate is created, so too is a private key file created. This private key is never presented to the public, and it’s the only thing that can decrypt sessions that were encrypted by its paired public certificate. So that third party could install that certificate on a web server theoretically, but they wouldn’t actually be able to decrypt anything because they don’t have the private key for the legitimate certificate.
So in order for a man in the middle attack like this to work, they’d have to obtain not only a legit websites public certificate, but also the corresponding private key. OR, the third party would need to get access to your PC, and install its own certificate authority signing cert, so that it’s fake, self signed certificates are trusted by your browser. Both of these are possible, but at that point you’re not talking about an unknown man in the middle, the man would have to compromise one of the two ends.
It should also be mentioned that you can get a CA to sign your certificate just before your domain expires, so you can attack whatever website is going to buy your domain next
Nevermind hackers. Look up "corporation in the middle" attacks, which is a prime example of subverting secure channels at scale.
If you don't own the hardware, nothing you do on it is truly private. Ditto if someone else has admin access to your hardware (eg BYOD scenarios) . Inserting a root certificate into the OS is trivial in both cases.
An additionap note on what a certificate is, to supplement everyone here who've desceibe howbthat's the missing piece:
A certificate's first main purpose is being the vehicle vy which the public key is distributed, but additionally it contains information ABOUT the owner. Then the whole thing is digitally signed with the private key (and also a trusted CA's private key), so that a receiver can validate the authenticity of the cert with the public key.
The "info" in the cert can theoretically be anything, but the most important one is the domain. Your browser knows that visiting google.com is secure because it checks the cert it gets from google.com to see if it states that it owns the google.com domain, and then we trust the root CAs around the world to make clients prove they own that domain, before issung a cert for it.
What if a hacker were to obtain the Master Key Certificate to the entire Internet? How much damage could be caused by that?
All the damage.
Which is why that cert (and its private key) are kept offline in a secure facility that takes multiple authorized people to access.
https://www.cloudflare.com/dns/dnssec/root-signing-ceremony/