TRANSKRYPCJA 🔗
WEBVTT
00:00:01.282 --> 00:00:07.709
Hey everybody, Evan Kerstel here today with Michael, Chief Product Officer at Gigamon Michael how are you?
00:00:08.170 --> 00:00:09.653
I'm doing great, evan, thanks, how are you?
00:00:10.855 --> 00:00:13.506
I'm really well Super excited for this discussion.
00:00:13.506 --> 00:00:18.810
I always get a glimpse of the future of cybersecurity technology from you and your team.
00:00:18.810 --> 00:00:28.893
This should be really educational Before we dive into the technology, maybe introduce yourself and your role at Gigamon.
00:00:29.882 --> 00:00:43.482
Yeah sure, michael Dickman, chief Product Officer here at Gigamon, so responsible for product strategy, which includes product management, definition, delivery of what we're doing, also product marketing, our alliances and ecosystem work, and the CTO function as well.
00:00:45.042 --> 00:00:46.972
Wow, that is quite a wheelhouse.
00:00:46.972 --> 00:00:50.326
You're a busy guy, so thanks for taking the time.
00:00:50.326 --> 00:01:05.890
We're here to talk about a topic that's new to me, so I'm going to plead a bit of ignorance here, because I am actually looking to learn myself about something called pre-cryption not encryption, pre-cryption.
00:01:05.890 --> 00:01:07.801
First of all.
00:01:07.801 --> 00:01:17.275
What is it and what is it solved for, and how is it looking to redefine the nature of cybersecurity as we see it today?
00:01:18.299 --> 00:01:21.007
Yeah, sure, so thanks for having me in the question.
00:01:22.239 --> 00:01:31.046
No need to feel bad for being ignorant, because it is a term that we actually created because this technology hasn't existed before, so I think it's a great place to start.
00:01:31.879 --> 00:01:36.566
So the idea here is there's encrypted traffic everywhere, and I think we know that.
00:01:36.566 --> 00:01:53.251
So encryption, depending on what you read, is 90, 91, 95% of all internet traffic, but now it's also similar percentages of an organization's internal traffic, including east-west lateral traffic, as well as north-south, and attackers are hiding inside of that traffic, which I think we know.
00:01:53.251 --> 00:01:59.873
And in fact, today most attackers are encrypting their own lateral movement and data expiltration.
00:01:59.873 --> 00:02:09.645
So they're using the same confidentiality that we all like to use to protect our data, to protect their attacks from our prying eyes, and so the question is how do you solve that?
00:02:09.645 --> 00:02:23.106
So, obviously, one way is decryption to remove the encryption by getting the keys and processing that appropriately, breaking an inspect man in the middle, or with perfect forward secrecy, putting some kind of proxy architecture.
00:02:23.106 --> 00:02:34.487
There's different ways to do that, but there's no really good way today to see security for security purposes, the plain text of lateral movement at scale, especially in the cloud.
00:02:35.840 --> 00:02:52.382
And so that's really what pre-cryption is trying to solve is how can I give security the ability to shine a light on east-west traffic that is encrypted in containers, in VMs, in the public cloud and private cloud in a way that's economical and efficient and simple to do?
00:02:52.382 --> 00:03:08.512
And so that's really what pre-cryption sets out to do, and the reason we called it pre-cryption is because we enable security to see before the encryption or after the decryption, that is already happening anyway at the application level.
00:03:08.512 --> 00:03:22.484
So the difference here is, rather than try to go over the top and break something that wasn't meant to be broken, let's actually grab it at the most base level in a distributed fashion and give it to the security team, leveraging some process that's already happening.
00:03:24.366 --> 00:03:25.567
Wow, very cool.
00:03:25.567 --> 00:03:34.217
How does this differ from traditional encryption methods that we've been relying on for so long, or other kind of solutions?
00:03:34.217 --> 00:03:38.080
Let's say to this problem what's your take?
00:03:38.925 --> 00:03:39.425
Yeah, sure.
00:03:39.425 --> 00:03:48.534
So if you think about you know take it for a given if you will that most data is encrypted so you can decrypt at the firewall.
00:03:48.534 --> 00:03:52.079
You could decrypt in something like a giga mon appliance.
00:03:52.079 --> 00:03:55.181
You could decrypt using a load balancer.
00:03:55.181 --> 00:03:57.108
So there's different places you could decrypt.
00:03:57.108 --> 00:04:15.901
Usually those are at the perimeter, and the way that those would work is you have to configure quite a lot actually, in particular, you have to have a way of doing key management and then you need to spend a lot of compute power to do the actual decryption, inspection and then re encrypts to send the traffic on it on its merry way.
00:04:16.826 --> 00:04:31.345
What this solution is radically different in its approach is you go all the way to the workload and so you actually leverage some native capability inside of the kernel called EBPF, which used to stand for extended Berkeley packet filter EBPF.
00:04:32.206 --> 00:04:33.627
And this is a native capability.
00:04:33.627 --> 00:04:39.994
It's already in very wide use and very stable for observability purposes, but it's less used for security.
00:04:39.994 --> 00:05:07.365
And so we're using that same stable, reliable component to give us that plain text visibility at a very granular level, and then we bring that text back, that payload, rather back into what we call our deep observability pipeline and then we can optimize, process it, secure it and send it on its way for threat detection, inspection, forensics, incident response, anyone who needs to look at that with reasonable purpose.
00:05:07.365 --> 00:05:17.819
So what happens from a benefit perspective is you actually don't need to decrypt anything, so it's like decryption
Load full transcript
without decrypting, and so that makes it much more efficient.
00:05:17.819 --> 00:05:27.040
You don't need to manage the keys, it's very low cost from a compute perspective and you can do this in all environments without any kind of change of routing.
00:05:27.040 --> 00:05:29.365
You don't need to force parameters where they don't belong.
00:05:29.365 --> 00:05:37.197
You can just let developers do what they want, let applications run where they are, but still allow security to shine a light on those threats.
00:05:39.187 --> 00:05:40.228
Wow, that's phenomenal.
00:05:40.228 --> 00:05:43.896
So, you know, the cybersecurity market, as you know, is very crowded.
00:05:43.896 --> 00:05:47.324
Different solutions, different approaches, philosophies to security.
00:05:47.324 --> 00:05:54.338
So who are you exactly competing against in this space with pre-cryption, this new approach?
00:05:55.245 --> 00:05:55.745
Yeah, sure.
00:05:55.745 --> 00:05:59.009
So I mean this approach doesn't exist anywhere else.
00:05:59.009 --> 00:06:01.531
But you know where I would say we're competing with.
00:06:01.531 --> 00:06:10.882
The main way we think about our competition is we're competing with attackers who think they can kind of get away with this encryption of their attacks.
00:06:10.882 --> 00:06:29.365
Because we do find for most customers we're talking to that because this problem has been perceived as so difficult it's generally not solved today, and so really it's a question of, rather than say, hey, why don't you replace decryption, which you're doing for good security purposes at the perimeter, continue doing that.
00:06:29.807 --> 00:06:39.980
You can now really broaden the scope of the security posture all the way to the lateral movement, and so we're actually not really competing with anyone in that today because it is so difficult to do.
00:06:39.980 --> 00:06:43.767
Now there's other ways to check for threats.
00:06:43.767 --> 00:07:00.026
So one might be a heuristics-based threat detection where you say, okay, I'm going to guess that this is bad, right, so I'm going to do some baselining and I'm going to guess that it's bad based on, you know what is publicly visible in terms of the data traffic and that's okay.
00:07:00.026 --> 00:07:19.860
But we've seen research, like from Zscaler for example, that says threat detection is five to seven times more effective when it is decrypted, and so really, we're looking at that ability to enable the existing security stack to be much more effective in a place that really isn't being inspected today but should be.
00:07:21.425 --> 00:07:22.286
That's fascinating.
00:07:22.286 --> 00:07:27.732
So being first and being innovative comes with its own set of challenges.
00:07:27.732 --> 00:07:34.819
Yes, were there any particular barriers or roadblocks you had to overcome?
00:07:34.819 --> 00:07:37.281
I mean, this hasn't been done before, correct?
00:07:37.745 --> 00:07:38.185
That's right.
00:07:38.185 --> 00:07:42.009
Yeah, this is new, yeah, new name or patent and all of that.
00:07:42.009 --> 00:07:51.459
So barriers to overcome, definitely technical barriers, so I can talk about those because we had to make this work and it can't work at HAWK.
00:07:51.459 --> 00:07:57.365
It has to work at scale, in multiple environments and in a way that is easily configurable and controllable.
00:07:57.365 --> 00:07:59.687
So there were definitely technical barriers there.
00:07:59.687 --> 00:08:08.577
The other kind of barrier is just kind of like an awareness or understanding that you don't have to take a heavy approach to this problem.
00:08:08.577 --> 00:08:28.072
It's possible to rethink how you look at the challenge of encrypted traffic visibility and going straight to the source, rather than saying, hey, I can't do anything, I'm just going to try to either guess or ignore the problem or put in a very heavy solution that forces changes to networks and everything else.
00:08:28.072 --> 00:08:29.990
A lot of this is just around.
00:08:29.990 --> 00:08:31.310
Hey, there's a better way to do this.
00:08:31.310 --> 00:08:39.669
I do think we have ahead of us a challenge, but also opportunity, to bring out the message to customers, especially in the security arena.
00:08:41.808 --> 00:08:43.013
Yeah, that's super exciting.
00:08:43.013 --> 00:08:45.009
What about cost?
00:08:45.009 --> 00:08:46.695
When it comes to implementation?
00:08:46.695 --> 00:08:50.394
Is this accessible for businesses of all sizes?
00:08:50.394 --> 00:08:53.471
Is this going to be an obstacle?
00:08:54.253 --> 00:08:57.873
just to be frank, no, I mean it's an important question.
00:08:57.873 --> 00:09:07.433
As I mentioned before, one of the economic benefits, if you will, at the very base level, is we're not requiring compute power for decryption.
00:09:07.433 --> 00:09:14.668
One of the reasons traditional decryption solutions can be a little pricey is they need to process all that decryption, which we are not doing.
00:09:14.668 --> 00:09:27.214
From a customer perspective, this is a matter of buying some Gigamon software, which we obviously do monetize, but we do that based on how much traffic is actually processed and it's leveraging something we already do today.
00:09:27.214 --> 00:09:38.873
Then there's a little bit of CPU overhead on every workload in order to configure and control this function From a Cloud perspective.
00:09:38.873 --> 00:09:43.956
If it was a private Cloud, there's almost no incremental cost at all at the base level.
00:09:43.956 --> 00:09:45.591
If there's public Cloud, it's very small.
00:09:46.426 --> 00:10:01.893
Then, from a Gigamon perspective, we price and package this as a volume-based license, which a lot of our customers are doing anyway, because they need to observe the traffic traversing their networks and now they can leverage that exact same capability even at this higher level with decryption.
00:10:01.893 --> 00:10:04.452
It's quite
affordable.
00:10:04.452 --> 00:10:15.615
I think the other thing, maybe more important than cost, is to think about what is the cost of bringing more and more heavy security tools that actually can't see the underlying traffic.
00:10:15.615 --> 00:10:24.471
I think that's really the question is let's just make things more efficient, let's not waste our time, let's not bring tools that aren't being given the data they need to get their job done.
00:10:24.471 --> 00:10:29.152
Overall, with that greater efficiency, I think you get much better results for the organization.
00:10:30.447 --> 00:10:30.969
I love that.
00:10:30.969 --> 00:10:35.913
As you and I know, integration can be a headache for new technology.
00:10:35.913 --> 00:10:44.917
How do you foresee organizations implementing this tech into their current stack and existing legacy technology?
00:10:44.917 --> 00:10:45.989
I mean what's involved there.
00:10:47.908 --> 00:10:51.390
Yeah, for anyone who's familiar with Gigamon, it's actually pretty easy.
00:10:51.390 --> 00:11:14.376
What happens is you would first set up something we call a V-series, which is a VM from Gigamon, and that will be the place where all of the magic happens in terms of aggregation, optimization and enrichment for processing and then brokering to send the network intelligence wherever it needs to go.
00:11:14.376 --> 00:11:16.028
You'd set that up so that's one thing.
00:11:16.028 --> 00:11:25.191
To configure and deploy Then all of the distributed intelligence that is actually accessing this data using EBPF.
00:11:25.270 --> 00:11:38.294
As I mentioned earlier, within the workload that is then controlled locally in a distributed way, you have something we call a fabric manager and that will talk to these things that we call Universal Cloud Tap or UCT.
00:11:38.294 --> 00:11:48.292
The UCTs will configure this native capability, ebpf, which is not actually a Gigamon technology, that's already in the workload.
00:11:48.292 --> 00:12:02.493
The deployment is actually easy because you're leveraging something that's already there and really, if you will and I hope you don't mind me using this older term here like an SDN fashion of controlling something that's already there in a software-driven model and just telling it what to do.
00:12:02.493 --> 00:12:08.929
Then the raw traffic comes back, gets optimized and enriched and then goes off where it needs to go.
00:12:10.767 --> 00:12:11.307
Fascinating.
00:12:11.307 --> 00:12:21.910
Understanding the level of disruption that could be involved in implementing a new technology like pre-cryption is could be a deciding factor.
00:12:21.910 --> 00:12:29.378
The idea of any kind of rip and replace is a scary term or idea for many organizations.
00:12:29.378 --> 00:12:34.009
So how seamless is the deployment?
00:12:36.013 --> 00:12:39.922
Yeah, I mean you could think of it as seamless as cloud, right.
00:12:39.922 --> 00:12:51.009
So you would have to deploy, like I mentioned,
one VM which and then you have to tell that VM where to send traffic.
00:12:51.009 --> 00:12:53.778
So you set up what we would call policies around that.
00:12:53.778 --> 00:13:00.840
Like, hey, I want to send email traffic to the email security team, I want to send application metadata to my observability tool.
00:13:00.840 --> 00:13:12.620
You'd set those things up and that's very easy to configure in that tool from a central console, again, fabric Manager, and then you would leverage what customers are already doing for cloud automation.
00:13:13.149 --> 00:13:16.399
So usually our customers are telling us Terraform very common.
00:13:16.399 --> 00:13:33.298
They could also use Ansible or other methods, helm charts, yaml, whatever it is that they are preferring to do, and that will configure these universal cloud taps or UCTs directly into the workloads and then just they need to know where that v-series entity is.
00:13:33.298 --> 00:13:48.539
So it's easy to deploy because you can automate the deployment at the edge per workload and then you only have to deploy in a centralized fashion, one place to do the processing and that itself is easy to configure from a centralized management station.
00:13:48.539 --> 00:13:50.456
So it's actually not that bad.
00:13:50.456 --> 00:14:03.120
I will say that we've run into challenges occasionally around permissions and so sometimes you'll have the question of well, I need to have the management station be able to reach and talk to the different entities.
00:14:03.120 --> 00:14:07.682
So that has been a question which just has to be resolved through good planning.
00:14:07.682 --> 00:14:08.975
There's no technical barrier there.
00:14:08.975 --> 00:14:16.518
It's just making sure you thought the eyes and cross the T's around who can talk to whom in the environments.
00:14:16.518 --> 00:14:19.235
But it has not been bad to deploy at all.
00:14:20.778 --> 00:14:21.421
That's interesting.
00:14:21.421 --> 00:14:25.058
So let me ask you kind of a big picture question why now?
00:14:25.058 --> 00:14:30.600
Is this idea of pre-cryption relevant now versus even a few years back?
00:14:30.600 --> 00:14:32.043
What's changed?
00:14:32.572 --> 00:14:34.230
Yeah, I think the answer is, but what?
00:14:34.250 --> 00:14:35.375
particular is changed.
00:14:35.851 --> 00:14:37.618
Oh no, no, it's super important question.
00:14:37.618 --> 00:14:48.216
So I will say we've been thinking about this for a couple of years, but I do think we're intersecting at a time when the threat and the urgency is very high, and it's because of encrypted traffic.
00:14:48.216 --> 00:15:05.921
So, if I look, maybe I don't know 10, 15 years ago, you would have had a reasonable amount of encrypted traffic on the internet, and then you'd have a lot of encrypted corporate traffic as it leaves the parameters, and then, five years ago now, you have a lot of standardized encryption out in the wild.
00:15:05.921 --> 00:15:09.820
I think what's changed in the last five years is everything is being encrypted.
00:15:09.820 --> 00:15:28.962
We're talking to a lot of customers that have an encrypt everything mantra Sometimes it's a mandate, and the view is that when you encrypt everything, you're safer, and I think what we're finding now as I mentioned earlier, with attackers using encryption as well is when you encrypt everything, you do a better job of confidentiality.
00:15:29.674 --> 00:15:50.582
You do not necessarily do a better job at stopping attacks that will steal data in other ways for integrity, availability or even confidentiality, when they can steal the data in a different way, and so not allowing this very powerful tool of encryption to be misused by attackers has really come to the head in the last couple of years.
00:15:50.582 --> 00:15:57.645
So I see the urgency much higher now than I do five years ago for not allowing free reign in lateral movement.
00:15:57.645 --> 00:16:11.561
So another example of why that's so important is you've got the huge rise of microservices and modernization of applications, where most new development is containerized, and you have a lot more of this machine to machine, server to server communication in the last few years.
00:16:11.561 --> 00:16:19.765
That is the perfect way to proliferate and attack and spread, and so not allowing those channels to go unmonitored is, I think, essential.
00:16:19.765 --> 00:16:24.139
So all of that has become much more concerning and important in the last few years.
00:16:25.475 --> 00:16:33.586
Brilliant, so this is clearly an excellent way to look at more preventative, proactive approach to cybersecurity.
00:16:33.586 --> 00:16:38.163
Is that the main benefit, or are there other benefits I'm missing in the solution?
00:16:38.894 --> 00:16:40.181
No, I think you nailed it.
00:16:40.181 --> 00:16:43.600
I think proactive cybersecurity is absolutely number one benefit.
00:16:43.600 --> 00:16:48.162
If I had to give you two more, it would be reactive cybersecurity.
00:16:48.162 --> 00:17:00.120
So if something has happened, you now have the ability to have been tracking application metadata from actual data communications and now you can go back from a forensics perspective or an incident response perspective.
00:17:00.615 --> 00:17:05.403
So I do think reactive cybersecurity is also very valuable in addition to the proactive piece.
00:17:05.403 --> 00:17:09.305
And then the third one is really around.
00:17:09.305 --> 00:17:20.443
I call just sort of classic performance assurance and performance management, because you can have questions of what's really going on and being able to dig into that in detail.
00:17:20.443 --> 00:17:22.382
We have seen a lot of concerns about that.
00:17:22.382 --> 00:17:29.563
So one example not an enterprise example, but I think it'll make the point is within the service provider 5G cores.
00:17:29.563 --> 00:17:48.602
There's actually encrypted data going there and you need to
actually decrypt it in order to correlate the control session of, like you and I are on our phones, we have to correlate hey, it's Evan with this actual data flow is Evans, and so that requires being able to look beyond the encrypted control plane traffic.
00:17:48.602 --> 00:17:54.561
So there's all these other examples where, from a performance perspective, you need to know what's actually happening as well.
00:17:54.561 --> 00:17:59.819
So those would be two other benefits in addition to, I think, the big one of proactive cybersecurity.
00:18:01.075 --> 00:18:09.417
Brilliant, so this is a little off topic, but as chief product officer, I'd love to ask you it wouldn't be a tech podcast if we didn't talk about AI.
00:18:09.417 --> 00:18:16.742
It's a big picture and it's such a game changer in so many areas, including cybersecurity.
00:18:16.742 --> 00:18:24.084
What's the role generally at Gigamod for AI and how excited are you by the possibilities?
00:18:26.237 --> 00:18:26.317
Yeah.
00:18:26.317 --> 00:18:29.125
So, on a personal note, I am really excited about AI.
00:18:29.125 --> 00:18:32.381
I think this has the potential to be just incredibly disruptive.
00:18:32.381 --> 00:19:10.640
We're in that point right now where everyone's arguing about whether it's overhyped or how much it's overhyped, but I do think this is as disruptive as the internet, where you go back in time in terms of what's possible and literally disruption in terms of changing the way work is done and who does what and how you ask questions and how you get information, which does tie in very much to where Gigamod plays, which is watching actual data communications, and so there are things that we're gonna be doing internally to Gigamod to make ourselves more efficient, like using generative AI for things like support or documentation or the way we do QA and our own work.
00:19:10.640 --> 00:19:16.398
But I think the really interesting part is from a customer perspective, as they're adopting AI.
00:19:16.820 --> 00:19:21.071
The signal to noise ratio and the quality of the input data is everything.
00:19:21.071 --> 00:19:29.361
If you train a model on poor quality data or if you train it on incomplete data, then it's really garbage and garbage out.
00:19:29.361 --> 00:19:37.103
Just imagine a cybersecurity system that has the best AI model in the world and they've only ever seen a North-South transaction.
00:19:37.103 --> 00:19:41.346
They've only ever seen whatever happens to be at the perimeter at that moment.
00:19:41.346 --> 00:19:46.685
By definition, that system is not going to detect lateral movement within the environment.
00:19:46.685 --> 00:19:52.685
It's never seen that so sure it does a good job of guessing based on patterns that it's seen.
00:19:52.685 --> 00:20:00.060
But if you don't give it complete visibility to what's happening everywhere, those AI systems for cyber are going to be limited.
00:20:00.060 --> 00:20:01.298
That's just one example.
00:20:02.295 --> 00:20:13.148
Another example in a different way when you think about the GenAI systems, one of the concerns is that they can be tricked into revealing personal information that they're not supposed to do.
00:20:13.148 --> 00:20:14.377
It's all over the news now.
00:20:14.377 --> 00:20:19.519
Some of them are funny, some of them are scary, but we know that that is a real risk.
00:20:19.519 --> 00:20:29.442
With those models, you could guarantee that whatever data egresses, one of these models is run through something like Gigamon to do masking for privacy.
00:20:29.442 --> 00:20:44.650
You could actually guarantee that no social security numbers or phone numbers or healthcare information is actually coming out at a low level like hard control versus a best efforts policy.
00:20:44.650 --> 00:20:48.163
So we see a lot of value on the front end with that signal to noise.
00:20:48.163 --> 00:20:53.365
We also see value in the way the process works as well for privacy, compliance audit and more.
00:20:54.214 --> 00:21:00.064
So this is certainly an evolving area, but anything that involves data communications, data in motion.
00:21:00.064 --> 00:21:03.700
That needs security, which GenAI and AI generally fits into.
00:21:03.700 --> 00:21:12.055
We definitely think Gigamon has a role to play, but the other thing I'd add, evan, if I may, is it has a role to play in making the broader ecosystem successful.
00:21:12.055 --> 00:21:15.845
So it's not like, hey, buy our company's product and you're done.
00:21:15.845 --> 00:21:27.105
There's all of these very smart people at all of the companies in the industry who are working really hard to make technology that can help, but if they don't have that signal, that quality signal, they're not able to do their job.
00:21:27.105 --> 00:21:33.443
So a lot of what we do at Gigamon is to make everybody else better and for cybersecurity, for AI-based detection.
00:21:33.443 --> 00:21:34.667
I think we have a big role to play.
00:21:34.667 --> 00:21:36.434
Brilliant.
00:21:37.217 --> 00:21:43.299
Well, in that mic drop moment I'll thank you for the quick education and update.
00:21:43.299 --> 00:21:49.522
I guess you're going to be on the road talking to customers and partners and press and media and analysts.
00:21:49.522 --> 00:21:52.041
Anything exciting coming up travel-wise.
00:21:54.180 --> 00:21:55.564
Yeah, we certainly have a few things.
00:21:55.564 --> 00:22:00.740
I'm always pretty customer-motivated, so I love talking to the media and others, but my favorite is customers.
00:22:00.740 --> 00:22:11.461
So we'll be meeting with our own Customer Advisory Board in October in New York, which is a number of CSOs and that is one of my favorite events of the year, so very excited to talk to them then.
00:22:12.616 --> 00:22:13.500
Oh, that's fantastic.
00:22:13.500 --> 00:22:15.119
Well, enjoy that.
00:22:15.119 --> 00:22:19.265
Congratulations on the mission, very important, very exciting.
00:22:19.265 --> 00:22:21.520
And yeah, thanks for watching everyone.
00:22:21.520 --> 00:22:26.239
Reach out to me or to Gigamon and Michael and others with any questions you might have.
00:22:26.239 --> 00:22:28.220
Take care Bye.