this post was submitted on 16 Dec 2024
353 points (97.8% liked)

Technology

59963 readers
3387 users here now

This is a most excellent place for technology news and articles.


Our Rules


  1. Follow the lemmy.world rules.
  2. Only tech related content.
  3. Be excellent to each another!
  4. Mod approved content bots can post up to 10 articles per day.
  5. Threads asking for personal tech support may be deleted.
  6. Politics threads may be removed.
  7. No memes allowed as posts, OK to post as comments.
  8. Only approved bots from the list below, to ask if your bot can be added please contact us.
  9. Check for duplicates before posting, duplicates may be removed

Approved Bots


founded 2 years ago
MODERATORS
(page 2) 48 comments
sorted by: hot top controversial new old
[–] jdw@links.mayhem.academy 8 points 1 day ago* (last edited 1 day ago) (6 children)

Don’t certs just create an ephemeral key pair that disappears after the session anyhow? What does cert validity period have to do with “This is a big upgrade for the security of the TLS ecosystem because it minimizes exposure time during a key compromise event.”

I mean, it’s LE so I’m sure they know what their talking about. But…?

[–] frezik@midwest.social 2 points 21 hours ago

The key pair you're thinking of is just a singular key for a block cipher. That key needs to be generated/transmitted in a secure manner. Meaning that its security is dependent on the cert. The expiration time of that cert is what they're aiming at.

[–] treadful@lemmy.zip 5 points 1 day ago (5 children)

I'm far from an expert on PKI, but isn't the keypair used for the cert used for key exchange? Then in theory, if that key was compromised, it could allow an adversary to be able to capture and decrypt full sessions.

[–] snowfalldreamland@lemmy.ml 5 points 1 day ago

Im also not an expert but i believe since there Is still an ephemeral DH key exchange happening an attacker needs to actively MITM while having the certificate private key to decrypt the session. Passive capturing wont work

[–] TheKMAP@lemmynsfw.com 2 points 1 day ago

Have you read about perfect forward secrecy?

load more comments (3 replies)
load more comments (4 replies)
[–] 486@lemmy.world 107 points 2 days ago (3 children)

I understand their reasoning behind this, but I am not sure, this is such a good idea. Imagine Letsencrypt having technical issues or getting DDoS'd. If the certificates are valid for 90 days and are typically renewed well in advance, no real problem arises, but with only 6 days in total, you really can't renew them all that much in advance, so this risk of lots of sites having expired certificates in such a situation appears quite large to me.

[–] jlh@lemmy.jlh.name 18 points 1 day ago (2 children)

That's true, but it would also have to be a serious attack for LE to be down for 3 entire days. There are multiple providers for automated certs, so you could potentially just switch if needed.

[–] sugar_in_your_tea@sh.itjust.works 17 points 1 day ago* (last edited 1 day ago)

The attack would only need to last for a day or two, and then everyone requesting updated certs when it stops could push enough people outside the 6-day window to cause problems. 6 days is probably long enough to not be a huge issue, but it's getting close to problematic. Maybe change to 15 days, which should avoid the whole issue (people could update once/week and still have a spare week and a day to catch issues).

[–] derpgon@programming.dev 22 points 1 day ago (1 children)

Most companies are not really suited for instant switching to a different cert service.

load more comments (1 replies)
load more comments (1 replies)
[–] hsdkfr734r@feddit.nl 88 points 1 day ago (3 children)

When I look at the default list of trusted CAs in my browser, I get the feeling that certificate lifetimes isn't the biggest issue with server certificates.

[–] errer@lemmy.world 21 points 1 day ago

The sites I have most frequently have had to add expired certificates to use are US government websites. Particularly those affiliated with the military branches. It’s sad.

[–] rottingleaf@lemmy.world 25 points 1 day ago

People who'd abuse trust into centralized PKI system are not real, they can't hurt you, because if they abuse it, said system's reputation will fall to zero, right?

Except it's being regularly abused. LOL. And everybody is using it.

[–] jagged_circle@feddit.nl 1 points 1 day ago (2 children)

Yes X.509 is broken. If you're a developer and not pinning certs, you're doing it wrong.

[–] oldfart@lemm.ee 3 points 1 day ago (1 children)

Yeah, now imagine pinning certs that change weekly.

My first thought is that old school secure software (like claws-mail) treats a cert change as a minor security incident, asking you to confirm every time. Completely different school of thought.

[–] jagged_circle@feddit.nl 1 points 20 hours ago* (last edited 20 hours ago) (1 children)

You can pin to your own CA. Then it doesn't matter if you want to update your certs frequently.

[–] oldfart@lemm.ee 1 points 20 hours ago
[–] stinky@redlemmy.com 7 points 1 day ago (1 children)
[–] jagged_circle@feddit.nl -4 points 1 day ago* (last edited 1 day ago) (1 children)

What part are you confused about, and are you a developer?

Edit: why was I downvoted for asking this?

[–] semi@lemmy.ml 12 points 1 day ago (1 children)

I'm a developer and would appreciate you going into more specifics about which certificates you suggest pinning.

[–] jagged_circle@feddit.nl 6 points 1 day ago* (last edited 1 day ago) (2 children)

I'm saying that if you're a developer of software that communicates between two nodes across the internet, you shouldn't rely on X.509 because the common root stores have historically been filled with compromised CAs, which would let someone with that CA decrypt and view the messages you send with TLS.

You should mint your own certs and pin their fingerprints so that your application will only send messages if the fingerprint of the cert on the other end matches your trusted cert.

[–] semi@lemmy.ml 6 points 1 day ago

OK, so cases where you control both ends of the communication. Thanks for the clarification.

[–] Pieisawesome@lemmy.world 3 points 1 day ago (2 children)

And your software stops functioning after X years due to this.

Don’t do this, this is a bad idea.

[–] jagged_circle@feddit.nl -4 points 1 day ago (1 children)

Yeah, fuck the users. We can just slap "100% secure" on the box and who cares if some woman is raped and murdered because we decided not to follow best security practices, right? /s

load more comments (1 replies)
load more comments (1 replies)
[–] M33@lemmy.sdf.org 12 points 1 day ago (1 children)

Digicert, Sectigo, Globalsign: hold my beer, 1 day certificate, even better: on the fly certificate per client 😂

[–] AAA@feddit.org 4 points 1 day ago (2 children)
load more comments (2 replies)
[–] sugar_in_your_tea@sh.itjust.works 14 points 1 day ago (1 children)

Interesting. I use LetsEncrypt largely for internal services, of which I expose a handful externally, and I've been thinking of only opening the external port mapping for cert renewals. With this at 90 days, I was planning on doing this once/month or so, but maybe I'll just go script it and try doing it every 2-3 days (and only leave the external ports open for the duration of the challenge/response).

I'm guessing my use-case is pretty abnormal, but it would be super cool if they had support for this use-case. I basically just want my router to handle static routes and have everything be E2EE even on my LAN. Shortening to 6 days is cool from a security standpoint, but a bit annoying for this use-case.

[–] kurt@lemm.ee 32 points 1 day ago (2 children)

You can use DNS challenge to renew your certificates without opening ports! Have a look at acme.sh for automation.

[–] treadful@lemmy.zip 4 points 1 day ago

certbot has modules for most DNS providers as well.

[–] sugar_in_your_tea@sh.itjust.works 7 points 1 day ago* (last edited 1 day ago) (1 children)

Oh, awesome! I thought that was a manual process, so I've been using the regular method.

Looks like I have a new project for this weekend. My DNS is currently hosted at Cloudflare, so this should be pretty straightforward.

[–] Rogue@feddit.uk 5 points 1 day ago (1 children)

Caddy with the cloudflare module makes TLS with DNS verification insanely simple

https://github.com/caddy-dns/cloudflare

Nice! I use Caddy, so this should be a snap.

[–] Laser@feddit.org 16 points 1 day ago

It's kind of in line with their plan to get rid of OCSP: short certificate lifetimes keep CRLs short, so I get where they're coming from (I think).

90 days of validity, which was once a short lifetime. Currently, Google is planning to enforce this as the maximum validity duration in their browser, and I'm sure Mozilla will follow, but it wouldn't matter if they didn't because no provider can afford to not support chromium based browsers.

I was expecting that they reduce the maximum situation to e.g. 30 days, but I guess they want to make the stricter rules optional first to make sure there are no issues.

[–] paraphrand@lemmy.world 9 points 2 days ago (1 children)

I wonder how short this could conceivably go…

[–] Rivalarrival@lemmy.today 26 points 2 days ago (1 children)

Dynamic generation. There is no certificate until user request.

[–] groet@feddit.org 25 points 1 day ago (1 children)

Perfect, let's also bind the certificate to a user session that is derived from a user fingerprint. That way the CA can track users across all sites

[–] stinky@redlemmy.com 11 points 1 day ago

I just want to serve https, not get someone's dick permanently installed in my ass

[–] gencha@lemm.ee -5 points 1 day ago

Increase how often the drones call the mothership, excellent.

load more comments
view more: ‹ prev next ›