Welcome to the KuppingerCole Analysts chat. I'm your host. My name is Matthias Reinwarth. I'm an analyst and advisor with KuppingerCole Analysts. My guest today is Martin Kuppinger, and that will be a longer introduction. So first of all, hi, Martin. Good to have you.
Hi, Matthias. Pleasure to be back here. Great to have you because we want to cover a topic which is a bit apart from what we have recently covered when we talked about AI security fabric, the identity fabric. But we are still in the area of identity management and we want to raise a topic that is a bit different. So as a starting point over the last years, and this is something that you covered as an analyst as well in your leadership compasses in your research, the direction for identity management was pretty clear. Everything moved into the cloud wherever possible.
And for many users and for many applications and use cases, that is the right call. And then there are these issues when it is not the right call, when there are DNS issues that kill complete sites of cloud service providers as a whole, so that many organizations are locked out for hours while the applications are still running, but you just can't get into that. And in general, for an office, that is an inconvenience or it really costs money, but that is different for a production line, in an energy grid, in a smart city, for a military organization.
So this is something completely different there. Their identity has to keep working even if DNS is not working. And it is always DNS as we have learned. So it has to remain usable enough so that nobody falls back to shared passwords or other scenarios where you break glass. And that exactly is our topic today. We want to talk about identity management that works disconnected, where the cloud is the right paradigm and especially where not and where an architecture can deal with that.
And after that long introduction, Martin, you have covered that actually in your EIC 2025 opening keynote and where you called it autonomous identity. Why is it back now? Why is this still a topic? Why are vendors looking into that? I'm not sure how many vendors really are looking into that. So I think they had their first signals on the market in a broader context, but I think it's worse that vendors start looking at it.
And maybe it's also because that I didn't pay that much attention or didn't follow up that much on it because there were so many other things to talk about, including building all the fabrics and the signal aspects, the mesh aspects, the orchestration aspects within the fabric, a lot of other things. And this is something we spotted as a trend or a longer term trending theme. And so it just came back to my attention recently. I thought it might be really good to bring it back. And in a sense, it's not really new.
We were discussing some of this for four years with our clients and advisory when they look at how can I build a resilience strategy for my operation technology environment, for my super critical data, for other areas that is not dependent on one of these leading cloud services for authentication or for other identity services. So take privileged access management. I think for IGA, it's less critical because if your IGA is down for a day, it's more an inconvenience than a real runtime issue. But for all the services that go into runtime, it is.
And even when we think about dynamic authorization, they just don't. Anyway, I think to your introduction, I would look at it a bit broader because there's definitely also a security angle. There is an angle of environments that are per se disconnected. So air capped environments where you don't want to open doors just because of identity. So I think we should look at it broadly enough. And I think there's a need for I believe there are a lot of use cases for an autonomous identity approach, something that is really separate from the traditional approach we take.
In a sense, when I go back a couple of years, I had conversations with one of the very large vendors where they thought about an autonomous version of their very well-known product, which a lot of have in place. I can't say names here. There are a couple of large players and one of these came up and said, yeah, we may have scenarios. Back in the days, it was maybe also not only defense or OT, might have been a cruise ship.
Right now, probably with satellite internet, this has changed a bit. But for other scenarios, it's still valid. And so I think we see a shift in the market here. Not a shift, but we see a gap in the market where there's a potential for doing identity management different for various reasons. Right. And I think making identity different is especially important when it comes to these use cases. And we just said autonomous or disconnected, but this is more than just one challenge. There are different use cases and you've hinted at them already. Maybe if you just walk through them a bit.
So one reason for these autonomous or disconnected use cases might be just sovereignty. I've talked about that with Alexei, but you want to be connected, but maybe you don't want to have any foreign influence onto that system. So it's disconnected from that area. Resilience, you've mentioned that. So there is this another good acronym, but I like it. It's DDIL. It's denied, degraded, intermittent, or limited connectivity. So that would be a factory floor. That would be something that is just so far disconnected that it's out there in the field for an energy provider.
True air gap, military, never connected classified networks. And on the other hand, there's an increasing number of edge autonomy because of latency. Because just because you are out there, you really do not want to phone home for each authentication authorization. So that is where decisions must happen locally and fast. For example, and fast is important there for traffic control. So these are very different use cases and we all put it in that one basket that's disconnected autonomous, but actually that's the reason for different architectures.
And it's not necessarily about creating the identities. It might be about authorization, which is an element of that. I would even add one more, which is in a sense, close to the DDIL, the resilience aspect, which is there may be also the fallback scenarios saying, okay, what if my standard service fails and I need continuous access, continuous identity decisions, whatever, for at least a certain set of my environment.
Again, operational technologies are very common for that because you may have situations where you must ensure that your production line continues working, that certain types of machines don't stop unexpectedly because there is a safety element behind it, not just a security element anymore. The other thing is, and you brought it up with traffic control, there are a lot of scenarios.
If we think beyond the human identity, when we think about the workload identities to connect things, all of that, then obviously also the picture is changing again because then we have way more use cases where this is absolutely relevant because there are these disconnected, these air-gapped scenarios and there are quite a number of them. Once you start thinking about it, more and more of these scenarios tend to come to the mind.
I think that is most probably also an exercise that many organizations have not yet done for themselves to really identify these areas in their own organization where they need autonomy and should have. I wouldn't say it that way. I think quite several organizations have thought about it. Some have some solutions for something. The bigger challenge, obviously, is that it means you have either two segregated or an overlaying solution. It could be that you say, this is the world I handle, so to speak, traditionally with our cloud-based identity management. This is the world I handle differently.
Or it's, I try to handle most with our standard identity management, but I have another solution on top which I use for specific use cases. In both cases, obviously, the complexity, the effort is higher. Doing that in a smart manner is a way is not easy from an architectural perspective, from a conceptual perspective. Maybe also one thing, this entire story is not about going away from cloud-first. There's the logic for cloud-first for many scenarios, but just not for all and not everywhere. I think this is what we really need to see.
It's about specific scenarios, but there are more of these than that we just could say, okay, these are some edge use cases we don't care much about. It's more than that. It should be an element we think about.
Obviously, a lot of these solutions are not in the narrow sense what is done today or what we think about is not in the narrow sense autonomous, but there could be scenarios where identity nowadays also with all the technology we have acts way more autonomous because also when we look at this, all this workload identity stuff comes in. We handle that and AI identities as well. We handle that via automation and obviously, automation is a prerequisite for autonomy in that field. So only then it can work. When we look, you've mentioned that already. So how would architectural patterns look like?
So in the end, the identity fabric as we think of it is a central control plane. So we want to have something that understands the identities at every point in time, whether they are connected or not. But on the other hand, to bring them to life in the individual use case that we just talked about, that would require more than one implementation of authentication, authorization, provisioning within these overarching identity fabrics. So two speeds, more than one delivery chain for this one central control plane.
So as you said, many use cases are fine with having a cloud-based, I don't know, Entra, Okta, whatever, but others just demand for a different service, a different speed, a different way of dealing with these challenges. So I think the design for an absence of connectivity is what really is sometimes missing. So we really need to factor that into our requirements, don't we? I think we need to think about where do we have which use cases and probably look a bit more on resilience failover and segregation as well, leading to a certain decrease of autonomy.
I think, as in many cases, it's not likewise. So there are probably different levels of autonomy, like full autonomy and temporary autonomy, in a sense. We need to look at where do we really need it from a functional perspective.
But also, can we handle it? Does it really help? I think there's still a bit of way to go there. An advantage I see nowadays is a bit that the standards, the technology, the way how systems can work, are way more evolved than back in the days when on-premise identity management started. So I think we can create also with standards-based connectivity interactions, communication, we can probably better create these types of, as I said, a broad range of systems. Right.
And I think we are benefiting also from the evolutions that we just see from agentic AI, from automation at a very high volume, high volatility. The standards that are coming into existence there and are tested and interoperable, they might, and I expect them to help also for this autonomous identity or disconnected identity very much because they actually solve parts of the issues that we see right now. So to get to a smooth handover between disconnected and connected when necessary, that would be a starting point. And because we had offline IAM back in 2005, but it was just a pain.
And now we have technologies that can make things much easier and support these new use cases, smart cities, where there is not a single user, but a shared infrastructure from a municipality or just a platform provider. I think these are the new challenges. Yeah.
I think, as I said, I think the scenarios are not gone. I think we try to ignore them at our best. But I think also when we look at identity as an element in cybersecurity, there are enough scenarios where we definitely need the ability to decouple certain environments and still ensure that they can keep running. So I think nowadays it's very common that we, after following certain incidents, these are incidents impacting an organization or incidents impacting a service provider.
A lot of things basically first go out of operations until someone feels they are secure enough again and then can come back, instead of saying we go into isolation. And also, so from that, also cybersecurity is something that drives this evolution. I would say, and I would be, I'm also very fair. This is still a bit early thinking in a sense of how this autonomous identity looks like. And I would really love to get a lot of from listeners to our podcast on that, their thoughts, their ideas to further shape this, because as I've said, I think we're just starting here.
On the other hand, my feeling is that it's worse to bring this back. So as I said, I started talking about this some 18 months ago almost. And so we must not let this just fall behind and ignore it because there's more. And I think, as I've said, all the turmoil, the uncertainty, but also all the able youth we are facing in the market and in the scenarios require it.
And latest, when we, as I've said, when we think beyond how do we create an identity, when we think into authorization use cases, into dynamic access, all that stuff, when we think about all types of identities, about use cases that are outside of the enterprise use cases, then the broader we think, the more scenarios arise where an autonomous or at least a fully autonomous or an identity fabric that can run autonomous when it needed could be of huge value. Probably also in quite a number of form factors.
Right, exactly. Also covering these different use cases that I've mentioned earlier.
So to say, okay, a fully disconnected network will need different capabilities than something that is just prepared for being disconnected for a limited amount of time. So that will be different capabilities. So as a starting point at the end of this episode, what should a listener do concretely? I suspected that maybe not everybody is yet fully prepared for understanding what their requirements are. So what would be a good test, a good starting point for our listeners to say, okay, maybe I need something that supports me in these autonomous use cases?
Think about where identity, authentication, authorization, all that stuff plays a role in your world outside of the office environment. I think that's the starting point.
That's OT, that might be connected things. So how do you properly handle access in a vehicle with different drivers and people in the car using certain capabilities, et cetera. At the end, we have an identity problem and that may also be quite a bit changing. So who can control the system and more, how do we do it really properly? I think there are a lot of things to think about in this world when we just sit back. So maybe a good glass of red wine or whatever you prefer and then sitting back and just thinking wild, in a sense, might be a very good starting point.
As you said, when you have these wild thoughts prepared and fixed on a piece of paper or in a mail, share them with us and then we are happy to build that, factor that into the next version of what we think a capable architecture in identity should look like. So not everything is AI today, not everything is agentic, but there are other challenges still needing to, solutions still needing to be solved. I think this autonomous disconnected identity paradigm is something that we should look into. It's an extension, it's not something different, but it should be factored in.
Martin, thank you very much for sharing your thoughts with us. I think that will be a topic that won't wait another 18 months, right? I don't know. But I think we need to work on it because I believe there's a pressure, there's something really, really big behind. And it's obviously an identity perspective beyond our traditional, let's say, workforce identity management, a broader perspective. But identities are ubiquitous and we need to deal properly with them across all the different use cases. Maybe this helps opening quite a number of new fields of projects, business, things to do. Right.
And maybe industry and critical infrastructure can learn, sometimes that is even possible from the military who have most probably prepared for that much longer than many other industries as of now. So again, thank you very much, Martin. Looking forward to having you back soon. And as always, it was a great conversation.
Thank you, Martin. Yes, thank you.