Slashdot Mirror


Starting Today, Google Chrome Will Show Warnings for Non-Logged SSL Certificates (bleepingcomputer.com)

Starting today, Google Chrome will show a full-page warning whenever users are accessing an HTTPS website that's using an SSL certificate that has not been logged in a public Certificate Transparency (CT) log. From a report: By doing so, Chrome becomes the first browser to implement support for the Certificate Transparency Log Policy. Other browser makers have also agreed to support this mechanism in the future, albeit they have not provided more details. This new policy was first proposed by Google engineers in 2016, and was scheduled to enter into effect in October 2017, but was later delayed for 2018.

37 of 172 comments (clear)

  1. this is starting to feel like email by cascadingstylesheet · · Score: 2

    You'll need an SPF record ... oh, and DKIM ... oh yeah, and DMARC ...

    1. Re:this is starting to feel like email by Anonymous Coward · · Score: 3, Insightful

      Gotta keep throwing up those barriers to entry. Can't have the small fry getting in on Google's take, now can we?

  2. Another Google metadata sink? by Anonymous Coward · · Score: 4, Insightful

    So how is this going to be implemented? Every SSL cert is going to be sent to Google for "verification" or is the CT log going to be local and the browser will just search it every time?

    1. Re:Another Google metadata sink? by squiggleslash · · Score: 4, Interesting

      Also what about locally signed certificates, using a corporate or Intranet CA, that's installed on all computers that might use those certs?

      That was, at one point, considered a best practice, but I assume this'll break that.

      --
      You are not alone. This is not normal. None of this is normal.
    2. Re:Another Google metadata sink? by houstonbofh · · Score: 4, Insightful

      I am ready to break all cert warnings entirely. As a networking professional logging into self signed certs all day, the cure is FAR worse then the disease!

    3. Re:Another Google metadata sink? by Curunir_wolf · · Score: 3, Informative

      Also what about locally signed certificates, using a corporate or Intranet CA, that's installed on all computers that might use those certs?

      That was, at one point, considered a best practice, but I assume this'll break that.

      This, from TFA (I know, right?): "Google engineers have also added a Chrome policy flag that allows sysadmins to disable the CT log-checking behavior in instances Chrome is deployed inside an intranet."

      --
      "Somebody has to do something. It's just incredibly pathetic it has to be us."
      --- Jerry Garcia
    4. Re:Another Google metadata sink? by Bert64 · · Score: 4, Insightful

      You shouldn't be doing that, your networking devices should have proper certs or your client system should be configured to trust a corporate CA and the network devices then have certs from that. Chances are your job requires you to have elevated privileges on all manner of important networking devices, if someone was able to MITM you they could steal some powerful credentials.

      The only times you should be logging into a device that uses a self signed cert are:

      1, initial configuration
      2, testing

      In the same vein however, it seems browsers are disabling support for weaker algorithms by default, and preventing you from turning them back on. This represents a significant problem, as there are all kinds of old devices which we still need to access for various reasons.

      --
      http://spamdecoy.net - free throwaway anonymous email - avoid spam!
    5. Re:Another Google metadata sink? by caseih · · Score: 5, Insightful

      Are you joking? Self-signed certificates are secure, arguably more secure than commercial CA-signed certificates because I had to register each and every one with the browser. I created the certs myself. A MITM attack is *instantly* detectable to browsers (and to me), unlike a MITM attack using bonafide signed certificates from a breached certificate authority. Browsers make using self-signed certificates somewhat awkward, which is unfortunate. Firefox tells me, incorrectly, that my self-signed certificate is not secure. That is complete nonsense of course.

      Another secure method is to sign with your own certificate authority. Then you just have to convince the browser once to take your CA cert. Like the self-signed certificates, MITM attacks are instantly detectable. This method is preferable to self-signed certs when you have deal with more than a few.

      In my mind for internal servers and devices, my own certificate authority is far more secure than using something like Let's Encrypt.

    6. Re:Another Google metadata sink? by Junta · · Score: 4, Interesting

      The CA model is particularly bad for 'internal' devices.

      So one, for internal communications inside a home network, the warning is so scary that some devices decline to support https just to avoid the support call because a web browser called the device 'insecure'. Note that https with bad cert is considered 'terribly insecure to the point of blocking the site' and http without any cert whatsoever is 'ok'. Home networks are not going to go through the rigmarole of all this.

      For another, my internal IT department is given the ability to sign a certificate for *any* site I visit to provide support for internal devices. I am not empowered as a user to elect to impose my own nameConstraints as I import the certificate, so to secure access to router.internal.mycompany.com, I give them access to impersonate my.bank.com

      Even when the IT department has ability to sign certificates, either it's going to be uselessly lax (automatic granting of certs for whatever reason) or impractically difficult (every sign requires tedious interactions). Companies are terrible about implementing the right balance internally.

      Assuming you overcame all this, you *still* will get a warning, because that internal IT department isn't going to have it registered in a CT log. If they *do*, then others can audit that log to discern details about their network and they have *another* class of security problem to tackle. Or chrome is deployed with a policy to disable this feature for the sake of the internal devices, *again* coming back to fixing internal network behavior requiring reducing security for the wider internet.

      The problem is that roughly all discussions on this front focus on the typical 'internet' usage and fail to conceive of approaches that would make sense for internal networks.

      --
      XML is like violence. If it doesn't solve the problem, use more.
    7. Re:Another Google metadata sink? by Junta · · Score: 2

      One amendment, that CT policy is better than I presumed it would be:
      http://www.chromium.org/admini...

      Would have been nice to link in the article, it took me a while to find it. So this provides a more targeted way to relax the CT, which can in turn limit the efficacy of that internal CA, so it seems to be a step in the positive direction.

      Good to see progress being made in limiting the collateral damage enabling https internally can inflict, but it's still in many ways convoluted and an ill fit for how many teams do internal IT.

      --
      XML is like violence. If it doesn't solve the problem, use more.
  3. Re:Er, what about LetsEncrypt by crow · · Score: 4, Insightful

    No, we need warnings for certificates that aren't trusted. Otherwise SSL does nothing to prevent man-in-the-middle attacks.

    What would be ideal is to support secure DNS with certificates in the DNS. Then you know you have the right certificate and don't need any certificate authorities at all. Of course, you have to trust the secure DNS. so it's just pushing the trust problem down the road.

  4. Registrars treat DNSSEC as an upsell ($) by tepples · · Score: 2, Informative

    No, we need warnings for certificates that aren't trusted. Otherwise SSL does nothing to prevent man-in-the-middle attacks.

    But without a fully qualified domain name, CAs shall not issue a trusted certificate. So we also need a reliable way to provide trustable names for devices on a non-technical users' home network that have a web-based administration interface, such as a modem, router, printer, or NAS.

    What would be ideal is to support secure DNS with certificates in the DNS.

    I agree that DANE would be ideal. But DANE relies on DNSSEC, which has faced practical problems that hinder adoption. Before about a year and a half ago, DNSSEC's root zone key was too short (1024 bit RSA) for browsers to accept as part of a certificate chain. And many domain name registrars bundle zone hosting with a domain, but a lot of these (such as GoDaddy) have charged more for zone hosting with DNSSEC than for zone hosting without DNSSEC.

    1. Re: Registrars treat DNSSEC as an upsell ($) by Monster_user · · Score: 3, Informative

      Why do home devices need to have trusted SSL certs? They are not web facing, and if they have remote capabilities they are typically routed through a service provided by the manufacturer. There is no reason to go through the trouble of key generation and registration against a global root CA.

      Besides, how is a global root CA supposed to verify the connection to a device on a non-routable IP/Subnet?

    2. Re: Registrars treat DNSSEC as an upsell ($) by JesseMcDonald · · Score: 2

      Besides, how is a global root CA supposed to verify the connection to a device on a non-routable IP/Subnet?

      Technically they don't need to. A Domain Validation certificate proves that the certificate holder controls the domain, not the server. Provided you have a (sub)domain you control, you can get a valid DV certificate through a DNS challenge without involving the device at all; the domain's A and/or AAAA records can point to a private or otherwise publicly-unreachable IP address.

      This does require a domain name, however, which is extra hassle and expense which most small-network operators shouldn't need to deal with. For the most common cases this could be provided as a service (like DDNS) by the device manufacturer. However, my proposal would be to bypass the CAs altogether: If the first part of the domain is the base-32 encoding of the fingerprint of the certificate (which may be self-signed) then browsers should automatically consider the certificate Domain Validated. The domain name itself identifies a particular certificate—what more evidence is needed? The "fingerprint domains" will be long, of course, but the initial discovery can be handled via an HTTP redirect or mDNS. Afterward, assuming the user bookmarked the full domain name, they can be sure they're connecting to the same device (trust-on-first-use).

      --
      "The state is that great fiction by which everyone tries to live at the expense of everyone else." - Bastiat
  5. Re:So....F U Proxies and Internal CAs. by tepples · · Score: 2

    If they did care about the end user's security, they wouldn't make stupid changes like not trusting end-user / admin installed CA certs by default.

    Chrome opts in through its network security config file, and Firefox has its own TLS engine. So this affects mostly native apps that use Android's TLS engine.

    Since when does removing / forbidding the user's input on trust somehow boost their security?!?!?

    Malware has in the past added certificates as a means of intercepting apps' traffic. So have governments.

  6. Re:Scam by Anonymous Coward · · Score: 2, Informative

    You know you can get a ssl cert for free and have it configured in like 30 seconds with certbot, right?

  7. Let's Encrypt supports Certificate Transparency by tepples · · Score: 4, Informative

    All websites with a fully qualified domain name qualify for a domain-validated certificate without charge from Let's Encrypt. Every certificate that Let's Encrypt issues is logged in CT.

    1. Re:Let's Encrypt supports Certificate Transparency by tepples · · Score: 2

      Anonymous Coward wrote:

      Which is exactly why let's encrypt should have their trust revoked. Such lax policies mean even illegitimate sites can get certificates, just look up all the paypal scammers using let's encrypt to show up as trusted.

      By the same reasoning:
      Which is exactly why domain name registrars should have their trust revoked. Such lax policies mean even illegitimate sites can get certificates, just look up all the paypal scammers using domain names to show up as trusted.

    2. Re:Let's Encrypt supports Certificate Transparency by tepples · · Score: 3, Informative

      Does Let's Encrypt verify identity, I can't find anything on their site about it.

      Let's Encrypt is a domain-validating certificate authority, which issues domain-validated certificates. Every such CA verifies that the person requesting a certificate is the same person who controls the domain's DNS. What other sort of "identity" did you have in mind?

      If a CA is not verifying identity then what use is their certificate?

      If a domain registrar is not verifying identity then what use is their domain?

    3. Re:Let's Encrypt supports Certificate Transparency by JesseMcDonald · · Score: 2

      Let's Encrypt doesn't issue EV certificates, so no, they don't verify real-world identity. They verify control of the domain name, just like everyone else issuing non-EV certificates. (Put another way, for DV certificates the domain is the identity.) The distinction between DV and EV certificates long predates Let's Encrypt, and their policies regarding domain validation are no more lax than most CAs'. Stricter, actually, since with LE you have to prove that you still control the domain at least once every 90 days.

      --
      "The state is that great fiction by which everyone tries to live at the expense of everyone else." - Bastiat
    4. Re:Let's Encrypt supports Certificate Transparency by AmiMoJo · · Score: 2

      If a CA is not verifying identity then what use is their certificate?

      It allows your connection to the web site to be encrypted, preventing ISPs and other nefarious agents from spying on you to a limited extent.

      --
      const int one = 65536; (Silvermoon, Texture.cs)
      SJW, n: "Someone I don't like, and by the way I'm a fuckwit" - AC
  8. If you use LE you're fine by tepples · · Score: 2

    A lot of people, including myself use LetsEncrypt on a CPanel based hosting account to generate certs for a website.

    Are those local, self-signed certificates or something that is registered somewhere?

    You could answer that question with five seconds on a search engine. Google Search for let's encrypt certificate transparency produces, as its first result, a document stating the following: "We submit all certificates to Certificate Transparency logs as we issue them."

  9. Re:Er, what about LetsEncrypt by swillden · · Score: 3, Informative

    In answer to your subject, from https://letsencrypt.org/certif...:

    We are dedicated to transparency in our operations and in the certificates we issue. We submit all certificates to Certificate Transparency logs as we issue them. You can view all issued Let’s Encrypt certificates via these links:...

    So LetsEncrypt certs will work fine with Chrome.

    --
    Note to ACs: I usually delete AC replies without reading them. If you want to talk to me, log in.
  10. CAA records by tepples · · Score: 2

    IMO all certificates should be EV in the current internet if we want security.

    I thought EV certificates were available only to corporations or LLCs, not to individuals. If someone puts up a site to show her personal portfolio, would you prefer to require her to incorporate first?

    But the current situation - Let's decrypt being able to issue a DV for any EV issued domain, is completely wrong.

    That's what certificate authority authorization (CAA) records are for. If a domain owner publishes a CAA record that doesn't include Let's Encrypt, Let's Encrypt will not issue a certificate for that domain.

  11. Re:Er, what about LetsEncrypt by hispeedzintarwebz · · Score: 2

    Ctrl-F "letsencrypt" Worries eased.

  12. Re:Er, what about LetsEncrypt by Anonymous Coward · · Score: 2

    Certificate authorities are entities that grant certificates. If certs were passed with DNS, a CA would still be needed, even if it's just the DNS server itself.

    Of course then the CA could not be airgapped, and the system as a whole would be more interesting to attackers and have a much larger attack surface.

    Right now there are warnings for certificates that are not trusted. If they do not have a path to a root of trust there is a warning. If they have been revoked there is a warning. If they are self-signed there is a warning.

    This logging system, it does not appear to provide any new services of meaningful value aside from making moderate-knowledgeable people more able to understand a cert and query it's nature. Sounds good but who decides what constitutes a threat? The cert logging policy website indicates CA certificates are strange. They are not. They are mandatory for trust. Apparently the people who made this logging feature get to decide what is and is not a concern to the rest of us, without accepting the fundamental nature of PKI as being strange.

  13. Secure Contexts by tepples · · Score: 5, Informative

    Why do home devices need to have trusted SSL certs?

    Because Service Workers and several other web platform APIs are restricted to secure contexts per W3C's spec. For example, a browser may restrict the Fullscreen API or Presentation API to secure contexts as a mitigation against phishing by replicating the chrome of the operating system and web browser. In such a browser, the web interface of a NAS on which video is stored will not be able to present the video in the full screen.

  14. Re:Er, what about LetsEncrypt by Anonymous Coward · · Score: 4, Interesting

    Perhaps, on the other hand, without letsencrypt most of us would not have websites. The poor people of the world would be completely cut off from having their own website, that was not the dream of the internet.

    We cannot be putting restrictions that cut off chunks of the population because they do not meet our criteria, the internet and having your own website should be free and open to all.

    In the bad old days you could only get an SSL certificate if you were incorporated, provided your contact phone number, real name, address, and pay a hefty sum of money. This was completely unacceptable and went entirely against the whole point of the internet. With letsencrypt the playing field has been leveled and this is a good thing and it is keeping the internet operational in the hands of the people.

    Honestly though I am still of the opinion that we should completely eradicate centralized certificate authorities. The certificates should be there to provide encryption which they do whether they come from an authority or not. We should allow free self signed certificates with no warnings. I should not have to link myself up to some 3rd party of any kind to operate my website.

  15. Re:Er, what about LetsEncrypt by tepples · · Score: 2

    Just to get a ballpark figure to guide further discussion of your opinion on Let's Encrypt: How much do you think someone ought to have to pay per year in order to host a personal portfolio site?

  16. Re:Er, what about LetsEncrypt by tepples · · Score: 2

    Let's Encrypt certificates are issued under an intermediate that has always been cross-signed by IdenTrust, an older and more established CA.

  17. Re:Er, what about LetsEncrypt by sjames · · Score: 2

    Like a standard cert from someone else requires anything beyond rudimentary photochop skills?

    Pinning would do a LOT more for security than the CAs ever have, but since that doesn't present any exciting new business opportunities, it remains unimplemented.

  18. Multiple routes, expiry, and CT block that by tepples · · Score: 3, Informative

    CAA records are useless when I hijack the DNS in the first place.

    I'm interested to see how you plan to hijack the DNS in a way that evades the following three defenses:

    Breadth of hijacking At what point would you hijack the DNS? A domain-validating certificate authority queries DNS through several Internet routes. How had you planned to hijack them all, especially if the domain's authoritative DNS servers are on different /16s? Expiry Certificates issued by Let's Encrypt expire after 90 days, and organizations may renew them at 60, 73, 85, or whatever day intervals. How long do you plan to keep up the DNS hijacking? Certificate Transparency If a CA issues an certificate to a hijacker, the domain's rightful owner can check CT logs and find your certificate. The policy change described in the featured article encourages CAs to keep their CT logs complete.
  19. Re:Self-signed certs don't block first-visit MITM by houstonbofh · · Score: 2

    A DNS entry like with e-mail would work.

  20. Re:Er, what about LetsEncrypt by JesseMcDonald · · Score: 3, Informative

    This logging system, it does not appear to provide any new services of meaningful value aside from making moderate-knowledgeable people more able to understand a cert and query it's nature.

    The point of the Certificate Transparency logging system is to make it extremely difficult for any CAs to get away with quietly issuing extra certificates for your domains to state actors and others to enable them to carry out MitM attacks. Since any CA can issue certificates for any domain, this is a real threat which undermines confidence in the entire CA system; it's only as strong as its weakest link. With browsers enforcing CT logging this attack is no longer possible; the certificates will not be accepted unless they are first made public, and any CA that issued such certificates openly would immediately lose its trusted status and be finished as a CA.

    --
    "The state is that great fiction by which everyone tries to live at the expense of everyone else." - Bastiat
  21. Re:internal apps / ipmi / other things that are no by dgatwood · · Score: 2

    That's partly a reaction to browser implementation of Secure Contexts [pineight.com], a W3C spec that reserves certain web APIs for HTTPS sites.

    I think it's more a reaction to browsers popping up security warnings on all non-HTTPS sites.

    On the one hand, getting public websites to use HTTPS is almost inarguably a good thing. On the other hand, getting intranets to use HTTPS is nearly useless, and getting mDNS devices to use HTTPS is impossible. That last one is going to be a real problem, and I'm really not sure how the industry is going to solve it. The only way I can think of would be to:

    • Define a new mDNS device name field in the certificate spec.
    • Require that all browsers and devices that implement that field use key pinning in lieu of any guarantee of global uniqueness to the names.
    • Require allowing multiple certs with the same name to be pinned with multiple keys (after issuing a stern warning that they are trusting a second device with the same name as an existing device).
    • Possibly require browsers to ignore the expiration date for those certificates (because those devices may not be directly connected to the Internet, and thus might not be practical to re-cert).
    • Possibly define a standard whereby a browser accessing a device with an old certificate can automatically obtain an updated certificate from the same registrar (or its designated replacement) and upload it to the device without user interaction.

    Either way, I'm pretty sure it can't be done practically without making some sort of changes to the standards themselves. That said, I can't be certain of that, because contrary to security best practices, the people who designed the X.509 specification will not make the specification available to security researchers unless they pay $130. So I can only speculate on what the standards say. Aren't standards grand?

    --

    Check out my sci-fi/humor trilogy at PatriotsBooks.

  22. Re:Er, what about LetsEncrypt by Actually,+I+do+RTFA · · Score: 4, Insightful

    No, we need warnings for certificates that aren't trusted. Otherwise SSL does nothing to prevent man-in-the-middle attacks.

    Sometimes, not protecting against a MITM attack is fine and I don't need to worry about preventing it. Examples include "being on a LAN and accessing something that is required to be behind https by W3C standards" or "local development of secure services before they're uploaded to test".

    --
    Your ad here. Ask me how!
  23. Re:Er, what about LetsEncrypt by JesseMcDonald · · Score: 3, Informative

    how hard would it be for them to work a command line flag like -gov to not log the certificate they are forging?

    Not hard at all, but it doesn't matter since browsers won't accept the certificate if it isn't in the log. That's the point of making CT logging mandatory.

    --
    "The state is that great fiction by which everyone tries to live at the expense of everyone else." - Bastiat