VMware Explore 2025 | Umesh Mahajan, Broadcom & Gillon Helman, Cole Engineering

Clips
More from VMware Explore 2025

Umesh Mahajan

VP & GM, Application Networking & Security Division

Broadcom

Gillon Helman

VP of Engineering

Cole Engineering

play_circle_outline Umesh Mahajan highlighted new VMware vDefend capabilities for quicker customer deployment.

play_circle_outline VMware Explore Insights: Advancements in Cybersecurity, Application Networking, and AI/ML for Enhancing Productivity and Securing AI Workloads

play_circle_outline Gillon Helman explained Cole Engineering's role in cybersecurity training for US Cyber Command.

play_circle_outline Enhancing Cyber Mission Forces: Training and Security Advancements to Combat Sophisticated Attacks and Protect Critical Infrastructure

Info
Transcript

Umesh Mahajan, Broadcom & Gillon Helman, Cole Engineering

Umesh Mahajan

VP & GM, Application Networking & Security Division Broadcom

Gillon Helman

VP of Engineering Cole Engineering

At VMware Explore 2025, theCUBE hosts John Furrier and Rebecca Knight sit down with Umesh Mahajan, VP & GM of Application Networking and Security at Broadcom, and Gillon Helman of Cole Engineering, to discuss the state of cybersecurity, AI-driven security advancements and training innovation for cyber mission forces.

Mahajan highlights Broadcom’s latest updates to VMware vDefend, including simplified deployment through AI/ML-powered automation, prescriptive security stages to accelerate time-to-value and new initiatives for securing agentic AI worklo... Read more

explore Keep Exploring

What are the new capabilities announced for VMware vDefend and the key takeaways regarding customer deployment? add

What are the main areas of focus for improving customer deployment and productivity with the product? add

What is the background and primary focus of Cole Engineering? add

What recent advancements and strategies have been implemented in cyber training and protection to enhance readiness against sophisticated cyber attacks? add

bolt Powered by CUBE AI

Umesh Mahajan, Broadcom & Gillon Helman, Cole Engineering

search

Rebecca Knight

>> Welcome back everyone to theCUBE's live coverage of VMware Explore here in Las Vegas. I'm your host, Rebecca Knight, alongside my co-host and co-analyst, John Furrier. John, let's talk cybersecurity.

>> Yeah. I mean, AI is one of the hottest areas and everyone knows that it's very vulnerable at many levels. So this conversation is perfect. Obviously, balancing out the traffic, understanding all the nuances. MCP servers are hot. It looks good on paper, but there's a lot of things to watch. This should be a lot of fun.

Rebecca Knight

>> The bad guys are everywhere. So with that, I'd like to introduce our next guest. We have Umesh Mahajan, VP and GM Application Networking and Security Division at Broadcom. Welcome back to theCUBE.

Umesh Mahajan

>> Thank you so much.

Rebecca Knight

>> , Umesh.

Umesh Mahajan

>> I'm so happy to be here.

Rebecca Knight

>> And Gillon Helman, Cole Engineering. Thank you so much for coming on theCUBE.

Gillon Helman

>> Thank you for having me.

Rebecca Knight

>> So Umesh, I'm going to start with you. You've just rolled out some new capabilities. This show, there have been so many new product announcements, but in particular capabilities for VMware vDefend. Tell our viewers a little bit about what's new and what the big takeaways are.

Umesh Mahajan

>> So there were two, several announcements, but two major areas. One is what we want to get our customers deployed quickly and get the value, because they buy the product and then they take a long time to typically get deployed, because they're limited in staff. So we are doing a lot of work in the product itself, ask a few simple questions, use some AI/ML techniques, and then come up with the firewall rules in a step-wise manner so that they can get deployed. Second area of focus is AI and GenAI. How do we do two things with it? We kind of raise the productivity. We come up with intelligent assist or co-pilot for our products so that the customers can be more productive as they use our product. And the other one is secure AI and agentic AI workloads, so we have a plan and a strategy how to do that. So those are our big focus areas.

Rebecca Knight

>> Can you walk us through? I mean, as you said, you oftentimes come out with new products and then there's a lag time between how customers deploy them. Can you walk us through a little bit how it's happening faster and easier now?

Umesh Mahajan

>> So what we are doing is, first, we want to give them visibility or figure out what is my risk posture. Where am I with security? Because many people, sometimes people are in a false sense of that thing that, "Hey, I'm secure because I did a few firewalls." No, you're not. So we can show them this is where you really stand, this is your risk posture, this is your security score. From there, then we can tell them, "These are the exact steps you can do to increase your security posture." And then we have come up with three prescriptive stages, stage 1, 2, 3. You secure your infrastructure services, rather easy to do that. So rogue, ransomware attacks don't happen by masquerading as RDP, DNS and all that stuff, so we lock that down. Second is you reduce your tax surface by doing zoning or macro-segmentation. That again is easy. Production is separate from development, is separate from DMZ. Third is where you get to the application level segmentation, and that's, while it's a little bit more data you need and you need to know what you're up to, again, it can be simplified and automated, and that's exactly what we are doing. So instead of customers taking years, we think they can, within weeks and months, they can get it completely deployed.

Rebecca Knight

>> And Gillon, for those who aren't familiar, tell our viewers a little bit about Cole Engineering.

Gillon Helman

>> Yeah, so Cole Engineering has been around since 2004. We're primarily a training software company. However, back in 2018, we started in the cybersecurity realm. We are the primary on a contract called Cyber TRIDENT, which is under US Cyber Command. We run the PCTE, which is the Persistent Cyber Training Environment. Primarily, what we do is we provide individual and collective training to all the cyber mission forces under US Cyber Command. So that's Army, Navy, Marine Corps, Air Force, Coast Guard, National Guard, the works, right? And our foreign partners as well. And so it's a full life cycle training system. We train everything from your fresh out of basic training. What is a firewall? How do you implement firewall rules? What is TLS? Things like that, all the way up to, you're now part of a cyber protection team and you need to execute an offensive or a defensive cyber operation. How do you do that as a team? And we provide the environment inside a VMR platform to execute those training missions and mission rehearsals.

>> You have boot camp and full forces. You're training everybody.

Gillon Helman

>> Everybody. Yes.

>> Obviously, that's huge pedigree and credentials. Congratulations.

Gillon Helman

>> Appreciate that.

>> Take us through the state of the market. Because cyber is always ... I mean, it's not for the faint of heart, that's for sure. We know that.

Gillon Helman

>> Yes, absolutely.

>> Take us through some of the training, things you're working on now, because the bad guys get a lift, the good guys get a lift, it's a ratchet game of cat and mouse. It's been talked about. There's a lot of defense. There's also strategic, proactive things you can do. What are some of the state-of-the-art things you're working on?

Gillon Helman

>> Yes. I mean, so a few things. So over the past year or so, we've gone through a massive expansion of our platform. So the idea is that there's multiple cyber protection teams out there. And the more people we can get through training, the better prepared they're going to be for that sophisticated attack from some nation state out there. And so over the past year, we've expanded our platform. We've quadrupled in size. We've implemented new multi-tenant security measures using Avi and vDefend and things like that to make those internal systems more resilient. We are able to execute malware in our systems, offensive cyber attacks in our platform and maintain that security of our platform as we move through it. So we're really trying to keep up with the state-of-the-art of security within the platform so that we can allow our cyber operators to execute really the full range of cyber attacks and defense inside the platform while maintaining that multi-tenant secure environment.

>> And the relationship to VMware vDefend, and you said Avi?

Gillon Helman

>> Yes, absolutely.

>> How's that going?

Gillon Helman

>> So, yeah. I mean, I've been in IT training for the Department of Defense now for about 26 years. During that time, I've gone from your hard-wired legacy systems to primary virtualization to software-defined platforms. The type of data that we deal with is obviously very sensitive. So keeping that secure is primary to us. Putting that in the public cloud is something that we don't want to do. And so when we looked at this, we kind of understood the scale of what we're going to be dealing with over a period of about eight years. And having something that we can deploy, the software-defined data center and the scale out, both our load balancing, security, screen policies and protocols. And as we expanded and our user base expanded, the load on the platform expanded was primary to us. And I'll be honest, I mean not, just because we were Explore, but there's not another competitor on the market that can scale with us and provide that automatic scale-out security that we get with VMware.

>> Really? That's news to your ears there. Testimonial right there in theCUBE.

Umesh Mahajan

>> Yes, exactly.

>> He's all smiling.

Umesh Mahajan

>> Thank you.

>> You can't pay for that. All right, talk about application security in the AI world. I mean, model poisoning, it takes one record, that's all it takes. Huge vulnerability. All kinds of other techniques. What are some of the things? Give us some confidence. We're already enthused, the hype cycle is so high. What are some of the things you guys are working on to lock that down in a good way, give the confidence so that these models just don't go off the rails?

Gillon Helman

>> Yeah, I mean, we've put a lot of effort into identifying how do you trust the model, right? Developing a standard. We're actually working with some universities and other companies to develop a standard of how you would test the model to develop trust in that model. Not just the model itself, but the data that it's been trained on. Is it biased? Does it hallucinate? How does it hallucinate? Things like that. And then actually, I haven't seen anything novel come out of the AI world as far as cyber attacks, which should put a smile on your face. However, agentic AI is allowing that malware to move at a much more rapid pace. So our defensive operators have got to be able to keep up with that. But I'll tell you the one thing that keeps me up at night is 10 years ago, if you had said to me, "I'm going to put some software in your platform and I'm going to give it all the administrative rights to mess around with your identity manager. Actually, run, modify commands inside your system to troubleshoot and correct issues." That's giving software administrative privileges in your system that you never would've done before. But now, since it's backed by AI, it's okay. And even things like network traffic analysis out of vDefend, it's going to have real problems identifying when somebody has created a vulnerability to get into either your model or the platform that's running that model. You no longer have that network insight to see that something in the regular chain has been broken. Because part of that regular chain is giving that model the ability to run those administrative commands in your platform. So most of our standard tools today aren't going to be able to realize that something is wrong.

>> It looks like an agent doing its job.

Gillon Helman

>> Exactly.

>> All right, so let's talk about the MCP I mentioned that on the open. A2A is another protocol. This has been a great organic development in the community, it comes from the open source. And so people like it because delegations involved, there's some trust. Again, quotes, trust. There's a lot of action. Tool chains are getting easier to build with GenAI. So the tools to do the security also enables the tools to fight against for the hackers. MCP, is it secure? How do you guys look at that? What's needed compared to A2A, for instance? Because it's not just an API call, this is delegation, this is like you said, "Here are the credentials, go do things."

Gillon Helman

>> Right. Exactly. So I look at MCP like this. We've actually got a steering committee right now looking at MCP to kind of understand what it is, how we're going to bring it into the platform, how we can secure it. I don't think the security measures that we have in place today are good enough. We're actually looking at building our own AI gateways specifically to enhance the security for MCP before it goes back to the models. If you look at what we've done in the past, MCP is what? It's basically a USB port for ML, for AI models. We didn't trust USB systems in the past because you can plug something in and it just propagates through your platform. We aren't going to run into this trusting MCP either. And to the rest of your question, how do you trust it? How do you get to that level of trust? I don't know yet.

>> So it's a software problem.

Gillon Helman

>> Absolutely.

>> This is going to be ... So I want to get back, you mentioned identity before. We were talking about cryptography before we went in. There's a lot of math out there right now, proofs that it's becoming very valuable in saying who you are. So in cryptography, there are ways to prove who you are without any kind of knowledge of who you are.

Gillon Helman

>> Absolutely.

>> How is all that playing in? I don't know if that's maybe quantum question, but it's more of a cryptography question. Is that coming in? Do you guys see any of this coming down?

Gillon Helman

>> Take that one first.

Umesh Mahajan

>> So I think yes, because our customers are asking for that support. Because with the quantum computing, today's cryptography, eventually, you can break it through enough quantum computing. And say, "Oh, it'll take you so many competitions, five years." No. Quantum computing, you could zoom down and do it much faster. NIST has published some new higher entropy algorithms for cryptography, post-quantum computing. And a couple of large banks and other government agencies want to be on the forefront of utilizing the technology. So we have rolled it out on the RV front, we're supporting it. Broadcom, NICs already supported, but that's in hardware. We have to do it in software also at layers above.

>> So you kind of got the readiness thing going on with quantum readiness inside the platform.

Umesh Mahajan

>> Yes. I don't know.

Gillon Helman

>> Oh, we're not ready.

>> No, no one. It's right. They're trying to get ready, but again, you have to start doing it?

Umesh Mahajan

>> Yes, because we want to make it available for the early adopters, which are some ... I don't want to name the banks which have asked for that support that's intended that feature.

>> In long lines.

Umesh Mahajan

>> And some government agencies. So they're going to be the early adopters and then the rest will follow as the technology gets more compute intensive, because we have to spend cycles to do the encryption today.

>> Yeah.

Rebecca Knight

>> Well, what I'm hearing from you both is when you are thinking about AI and talking to customers about it, are you more excited about the potential, or are you more concerned about the increased surface attack? Because when it comes down to it, we've not yet had the big one. We've not had a Pearl Harbor cyber attack and we've had incidents here and there. Of course, they get a lot of media attention, but we've not had a 9/11 in cyber. Think about it how do you have conversations with your customers about it.

Umesh Mahajan

>> So customers are concerned and it's both a concern and it's an opportunity. If we can solve that problem, clearly, customers will be more interested in our product, so we do want to solve the problem. The way I look at some of these attacks is, Gillon said they might not be able to be able to crack things, but they can keep probing forever in the agentic AI kind of a thing. And you keep looking for that little hole someplace and get in, which normal attackers, it may not have been possible to do it. They'll just keep running around. And so it could be the next version of bot, but more powerful as they are attacking. And eventually, they will find a side door and get in. So you have to find ways to protect it. And they will be advanced attacks, but at least let's stop all the basic ones completely so that then you can dwell on the super advanced one, the new attacks would show up. Because today, when the normal ones or the so-called slightly advanced ones are not protected. So let's get all that locked down completely so that then teams can focus on the super advanced attacks.

Gillon Helman

>> Yeah. From my perspective, I am super excited about AI for many reasons. I mean, it brings the world's knowledge into your hands, right? It's incredible. It's a level setter across social spheres. However, from a cyber perspective, I'm not completely terrified of that big one. Eventually, something will happen that will go down in history as the attack that changed everybody's mind, and I see that. From my perspective, and I'm going to pat ourselves on the back from cyber command on down, I have faith in the cyber protection teams and cyber mission forces that have been trained in this specific area. They've done a great job. I mean, just recently last month, Saint Paul, Minnesota, they got hit with a ransomware attack. It was a very sophisticated attack, lots of resources behind it and the city just couldn't get over it. They were down for weeks and they called in the cyber protection team from the local National Guard in Minnesota, and they were able to get them back up and running. So they were down for weeks, but they called in the guys who've been trained to do this sort of work and keep America safe. And even though our enemies are very sophisticated, I think that our teams with the right training, I think that they are better.

>> And what you guys do is so important because we've never had a cyber military, because attack digitally is like troops landing on our soil.

Gillon Helman

>> Exactly.

>> Why wouldn't you deploy a team to assist?

Gillon Helman

>> Exactly.

>> Versus companies building their own militia.

Gillon Helman

>> Right.

>> Which is, they still do, but this is where the platform advantage is coming. So I think the scale piece is also interesting. Any thoughts on some of the data scale as you get more and more data coming in? Because there's a lot of sharing in the cyber community.

Gillon Helman

>> Absolutely.

>> People share information.

Gillon Helman

>> Yes.

>> I heard that there's a day minus zero where it's like you can start seeing the formation of a team being constituted in the dark web. I've heard things like that. I mean, this is kind of the new data coming in. What's just the reaction to that? What's your thoughts?

Gillon Helman

>> Yeah. I mean, there's a lot happening every single day. It's literally impossible to keep up with everything that's happening. I think that our platform really provides some of those cyber mission teams an advantage. So we have people that can go out into the field, they can get information, they can bring it back in, they can test it in a safe environment. They can figure out what it does and maybe even use it to their own advantage. And then from just a data perspective, so I know that for us, we have petabytes and petabytes and petabytes of storage space and data, and it's being distributed across multiple of our environments that people can go and use, but this is very specialized information. And so yeah, it's-

>> AI helps you too....

Gillon Helman

>> overwhelming.

>> There's AI for security and security for AI.

Gillon Helman

>> Absolutely. I actually just saw that a minute ago. I don't know who coined that term, but that was great. But, yeah, so just being able to access that data isn't good enough. We actually need to use AI to basically build massive RAG platforms to consume that. Right? So somebody will have to go in there and be able to do a natural language query to figure out what they're looking for. Right? There's just too much for any one human being to go through and figure it out.

Rebecca Knight

>> Exactly. Exactly. Well, this has been a fascinating and wide-ranging conversation, Umesh and Gillon. Thank you both so much for coming on theCUBE.

Gillon Helman

>> Thank you, guys.

Umesh Mahajan

>> Thank you so much.

>> They're closing the place down here. Look at this, beautiful.

Rebecca Knight

>> I know. I know. And I'm Rebecca Knight for John Furrier. Come back tomorrow for more of our live coverage of VMware Explore. You're watching theCUBE, the leader in enterprise tech news and analysis.

>> I really appreciate the incredible coverage theCUBE provides. Not just for this show, but for-