Podcast
Root Causes 450: 2025 Predictions
Hosted by
Jason Soroko
Fellow
Original broadcast date
December 23, 2024
We make our 2025 predictions. Topics include maximum certificate term, AI, post-quantum cryptography (PQC), deep fakes, and more.
Podcast transcript
TCTim CallanLet's go Season 2. And it is near the end of the year, and we are all ready to do our predictions for 2025.
TCTim CallanThere is. And you made a list. I made a list. We didn't compare our lists. We decided yesterday, coordinating for this, they were gonna operate off your list, and then if you hit an item I had, we'll just cross it off. If you hit an item and I said the opposite, we'll have a nice debate. And then afterwards, I'll just pick up anything that is different between them.
TCTim CallanThat works. So, Jason, Prediction #1.
TCTim CallanHere's what I wrote. A concrete schedule for TLS lifecycle reduction is in place and starts in 2026. So that's one of my predictions. Yes.
TCTim CallanI think 2024 we saw a great deal of progress in this area. And leaving the year, we are poised for something to be codified, either - and I would say preferably through CA/Browser Forum ballot. I would really like it to be through a ballot, but if needs be through a browser root program requirement. And I don't think that the major browsers are just going to let the year go by if a CA/Browser Forum can't get there. I don't think they're just going to let it go. I think they're going to take their own action.
TCTim CallanI don't know if it would be a first at all, because I don't remember off the top of my head whether the three-year restriction passed by ballot or whether it was a root program requirement first.
TCTim CallanBut it's definitely rare.
TCTim CallanSo in all fairness, capping certs at all happened by ballot. Like the three- year EV requirement happened by ballot. Or the two-year EV requirement, rather, the three-year code signing requirement happened by ballot. And the three-year moving down to two-year S/MIME requirement happened by ballot. So it seems that two/three-year kind of time frames are very passable in the CA/Browser Forum. I think it's the shorter stuff where we're finding that it's not.
TCTim CallanSo that's one we're in alignment. We both had the same thing, and I can cross one off my list.
TCTim CallanIt's interesting. We've seen this with other things. And I think when you say that, I immediately think about discussions you and I were having about four years ago, about what to call, what ultimately wound up being called, PQC. And I'm not sure that PQC was the best choice, but it's what everyone settled on. I don't know how these things happen. There's some secret cabal decides and we all have to do it or something, but I think yes, you and I should be the secret cabal. I agree. So then when you do get these kind of warring sets of terms, you have to collapse on one, or it makes discussion really hard.
TCTim CallanSo are you picking a winner or are you just saying that it's gonna have to get settled?
TCTim CallanGood. That's prediction number two. We're gonna settle our taxonomy around human and non-human digital identities.
TCTim CallanSo sovereign AI, meaning AI that is operating independently of human oversight?
TCTim CallanNow this, this is shades for me, of a lot of conversations we've had before.
And I think you can probably predict where I go on this one, which is, I don't trust their confidence and I don't trust their good intentions.
TCTim CallanI think kind of in a free economy, if you want to make your AI however you want to make it, people can choose to use it or not, and they can live with your bias or not, I get nervous about this force. This control by Fiat of which technology solution we will use under which conditions. That's where we get into ground that I get extremely uncomfortable with.
TCTim CallanAll right. That's a good one, Jay.
TCTim CallanI mean it's interesting because if you think about there was a, let's say a six-month period where it seemed like the progress in AI was ridiculous. If you compare AI today to what we had a year ago - not super different.
TCTim CallanSo, and I don't know if I'm going to step on another one of your predictions here. In terms of the core tech, I agree with you. I do think over the course of 2025 though, we are going to see the complete, the end, of something that started in 2024 which is the complete and utter - what's the word I'm looking for? Applicationization. Is that a word? The complete and utter integration of AI into every conceivable use case, even the ridiculous ones.
TCTim CallanSo if I can trick, just to put it in layman's terms, if I can trick the AI, I now have this very empowered agent with a huge amount of privileges inside the target, that if I can get that to do the wrong thing, if I can get that to do the things that I want it to do, which are not in the best interests of the owner of that AI, that is a major potentially powerful attack surface.
TCTim CallanSo are you predicting that we're gonna start to see news articles about this occurring in the wild in 2025?
TCTim CallanI think so.
TCTim CallanThere's another one.
TCTim CallanThey did publish a major update pretty recently.
TCTim CallanI know where you're going with this. It's a call to action for subscribers, for users, to cease using public certificates in use cases that are more appropriate for private certificates. Is that your point?
TCTim CallanThis is definitely one of the things that the Chrome team has made into a pretty strong statement in recent times, during 2024, because and the reason for that is this goes back to conversations we've had in the past, and certainly we'll have in the future about the Bugzilla blood bath, because a big element of the plague of DelRev incidents that is still going on, the DelRev epidemic that is still occurring, is supposed situations where they're too fragile and critical for public certificate life cycle requirements. And we're hearing this in terms of delayed revocation. We're also hearing this in terms of shortening certificate lifespans, where as we start talking about making public certs shorter, the similar hue and cry comes up where people on the subscriber side or CAs, who want to represent that viewpoint come up and say, well, they can't cycle these certificates out in this kind of timely fashion. And so the response for both of those coming from Chrome is to come back and say, well, then you're using the wrong technology. There is technology that does this. That's technology you should be using. Instead, you're using the wrong technology.
TCTim CallanYou don't need to.
TCTim CallanThat's one where there isn't a case - if it's real client auth - There isn't a case where you need public certs. There just isn't because you own the authorizing system.
TCTim CallanAnd I will layer onto that. I agree with you completely, and I agree that that will become a major point of discussion in the industry. And I will layer on to that, that some of the actions that we are seeing proposed in the public certificate space, including reducing certificate lifespan, including requirements and penalties around delayed revocation are in part to address this exact problem.
TCTim CallanIt solidifies the 2030 date as a real date.
TCTim CallanI mean the 2030 date, like how do you pull it in? You can't pull it in.
TCTim CallanIt's only five years away.
TCTim CallanSo 2025 is the year that we have broad acceptance among serious, informed thinkers, it's not a science project, it's an engineering driven future. And that this is a thing that we must prepare for, because it will happen, whether we want it to or not.
TCTim CallanAre we still on the PQC topic?
TCTim CallanI didn't know that. Well, and rightly so. So, one of the things that I've maintained is that 2023 and I think this part has held up, 2023 was the year for the industry, for industry, for the technology industry, to wake up to PQC. 2024 was the year for enterprise IT departments to wake up to PQC. 2025 from my list, is the year that we see PQC in production.
TCTim CallanAnd I can run that on existing, established, cheap, reliable, ubiquitous architecture.
TCTim CallanSo you're talking about, basically, implementations of mostly ML-KEM, broadly. That?s kind of what I meant with my shorthand of PQC in production. Like we did see some of that in 2024. We had Cloudflare. Now you still won't be able to do that with a public cert, but I think 2025 is the year where you will be able to run ML-KEM, ML-DSA, in your internal, entirely owned walled garden environments. And we will see that happening in real, real world production systems, starting with the ones that are most sensitive, highest risk. Starting with obvious candidates. Think about a major global financial institution or a government, but yes, I do believe we will. We will be seeing that occurring in real production in 2025.
TCTim CallanOh, and what is it?
TCTim CallanCan all be blockchain.
TCTim CallanFascinating.
TCTim CallanWe are.
TCTim CallanA lot of people speculate that this is basically a mutual assured destruction situation? I can take down your power grid, but you're going to take down mine. You think we start to see those things really happen?
TCTim CallanThat's a tough one. That's a hard one. That's not so easily actioned.
TCTim CallanThat's interesting. I sincerely hope that we're sitting here a year from now, going wrong, but - -
TCTim CallanBut I think there's a very real risk of that.
TCTim CallanSo I think I'm going to tie this into the first one I'm going to bring up.
TCTim CallanWhich is, I'm predicting a general tightening up of the WebPKI. I think the WebPKI was extremely loose. Was badly out of control in the early teens, the 20 teens, and we got some basic control and expectations and normative behavior in place. In the last few years, maybe the last five years, the progress on upping the ante on security and reliability and predictability and trust has gone down tremendously. We've seen a lot less progress, and we've seen pressure building up around that which blew up in the Bugzilla blood bath, the DelRev incidents, the pressure on shortening certificate lifespans, all of that. I think we will see a general across the board, tightening of the WebPKI, in a number of ways in 2025 as various parties, driven by the browsers, where the power is, say, look, we've had enough. You guys are like a bunch of unruly teenagers, and we're going to put our foot down, and we will see a tightening of the WebPKI in a variety of ways across the boards. This isn't being me some great seer, because there's been a lot of conversation about this. Like we're seeing an appetite from the browsers to push these things in a way that we haven't seen in the last few years. And I think part of that was COVID. I think when COVID rolled around, everybody got real conservative about this kind of thing for a while. Now that we're coming out of it, there's been a dearth of progress, and there's an appetite to make up for it. As part and parcel of that, yes, the question has been asked, do we really need 80 CAs? And it's important to remember that every CA is equal in its level of threat to the WebPKI, regardless of the amount of the real usage that it has. So it would be a fallacy to say, well, if you take four or five CAs, they account for 99.99% of the certificates out there. So as long as those four or five CAs are good, you're okay. That's a fallacy.
The way to think about it is, if digital identity is a wall that keeps out the invaders, then every CA is a gate in that wall, and every gate is equally large, and every gate is equally suitable for invasion. And every gate must be protected. And if a gate is well protected, like, let's say, Sectigo, but a gate next to it is entirely unprotected, like some of the ones that were distrusted in 2024, then the invaders will just go through there. And so if you can't protect all the gates, another thing you can do is brick them up.
I do think that there will be more distrust events by root programs in 2025.
TCTim CallanI do not think there will be a wholesale house cleaning. I do not think we're going to suddenly go down to 30 CAs. But I do think there will be more distrust. And I think this conversation about why on earth do we have 80 CAs, when we probably could do it with 10 is a conversation that will continue.
TCTim CallanSo there you go. There's my WebPKI tightens up. Shall I go down a few of mine? So, excuse me if we return to the AI topic real quickly. I want to go back to deep fakes. I think 2024 was the year that deep fakes moved into the popular understanding. And we predicted this. I think 2025 is the year that, in general, we give up on any concept of believing in the veracity of a so called recording without independent confirmation.
TCTim CallanAnd just, just kind of fundamentally be a skeptical of a 'recording' as you would be of hearsay. Oh, can you believe that such and such said this? I don't know if he said it. And that's the same way we're gonna get even if you supposedly see the person saying it on video. I think that is 2025 is where that's gonna kind of move into the common understanding.
Another one. I think 2025 certificate automation, CLM, ACME, all of that. Let's call it certificate automation and certificate agility becomes a de facto boardroom topic. What I mean by this is that if it's not a boardroom topic or not already solved, you are in the minority. That is my prediction for 2025.
TCTim CallanThey're gonna flip. It was a little wedge of people who were thinking about it, and a giant pie of people who weren't. They're gonna flip in 2025.
TCTim CallanThere'll be a little wedge of people who aren't thinking about it, and a giant pie of people who either are or have already addressed it.
TCTim CallanAnd outages, like, they keep right on going. Still see headlines. Like, all of these things are gonna combine. This is the year where we see that occur.
TCTim CallanThen connected to that, this is a topic you and I have touched only a little, but I still think it's really interesting. I think 2025 is the year that we see the emergence of a bona fide category that I'm gonna call namespace management.
TCTim CallanWe've talked about it like one or two episodes.
TCTim CallanBut we need to talk about it more.
TCTim CallanI mean, it's a big need. It's a big opportunity. We've seen tech providers emerge who can do this. It has real utility. There are attacks, real world attacks that are coming around weaknesses and namespace management. I think this is the year that we see this emerge as a bona fide category, well, it'll still be forward thinking organizations that are doing it but where it we all start to look at is okay, this is a real thing that is on my security playlist.
TCTim CallanI agree. I agree. Vendor assessments. Tim's favorite hobby horse, except for delayed revocation. Vendor assessments.
Vendor assessments become basically ubiquitous. They get longer. They get more detailed. They get more exacting, and they just are everywhere. I think 2025 is the year of the vendor assessment. It's getting to the point, I mean, there's new legislation coming out for financial institutions, setting minimum standards for supply chain assessment, and they are very exacting standards. And this is happening in multiple countries for regulated industries, and it's going to be very, very normal.
Now, what I'm not predicting for 2025 which we need, is a standardized way to do this. Like there is absolutely no reason why there couldn't be a list of 1,000 questions. They're the same 1,000 questions, and we all know what they are, and we all go to get an answer once, and then everybody just gets our list. We do this exactly once. Same questions in the same order, with the same numbers next to them. That is something we need. I do not believe that's happening in 2025. I think it really should, and I would like to see that, but I don't think that's going to come together this year. But I think this year, this is going to be just everywhere. And as part of that, there will be the emergence of this as a major pain point, as a crisis that needs solving for the enterprise, as we find ourselves in this situation where you can't operate because of the vendor assessments you're being subjected to.
TCTim CallanSo that's one. So returning to AI and PKI, there was a conversation that you and I had prepping for this, and I think we want to bring it up here, which is, I came in with a prediction which is that PKI is going to continue to be relatively unaffected by AI. And then you push back on me.
So let's start with my prediction. My prediction is that of all the things in our grand world that are being transformed by AI, PKI has come out pretty much unaffected. And I predicted the same thing for 2025 and then you said, what?
TCTim CallanAI is great for things that are fuzzy, nebulous, things that, for want of a better word, I'll say, require judgment. And PKI is the opposite of that.
TCTim CallanPKI requires exactitude. And exactitude, sorry, is one of the things that our LLMs aren't all that good at.
TCTim CallanThat's like applications and use cases of PKI, which is important, it's not basically the same as the actual bits and bytes, the ones and zeros of the PKI.
TCTim CallanThey go to baseball games.
TCTim CallanAnd they go to weddings and they sleep, and they do all these really kind of disruptive things that prevent them from being productive.
TCTim CallanWell, first of all, it won't sleep. First of all, it'll work around the clock all the time. Second of all, just throw more chips at it. And instead of having 8 people, you can have 800 people.
TCTim CallanSo you can imagine a scenario where there's this army of helpers who are investigating every squirrely angle you can think of, no matter how unlikely, and coming back with everything that seems interesting and then those eight people who really know this, can sort through it, pull those lumps of gold out of the sand, and that could be a real, a real game changer in terms of attacks against cryptography.
TCTim CallanIt's interesting. Again, let's hope not, but it's something to be worried about.
TCTim CallanClassical mathematics with AI-assisted classical techniques break it. I love it.
And then the last one, just because I have to say it, because I always have to say it is, I think once again, 2025, as every year as long as we've been doing, this is going to continue to be the year of government versus the internet. There's no sign of anybody letting up on their viewpoint that governments want to control and break the internet, and they're trying to break the wheel of technology continually, and they're going to keep right on doing it.
TCTim CallanYes, yes. That's a little teaser for a feature episode, third phase. So that's it. Bunch of predictions. And as you and I have decided to do, a year from now, we'll come back and we'll see how we did.
TCTim CallanThanks, Jason. This has been Root Causes.