If you were talking to someone whose organization is considering Vectra AI, what would you say?
How would you rate it and why? Any other tips or advice?
Take time to understand how the triage filtering works and standardize it early on. Use a standardized naming convention and be consistent. It's a very effective tool, but if you don't pay attention to what it's telling you, then it's like anything else. If you don't use it, then it's no good. You have to trust that what it's telling you is correct and then you can take the appropriate action. For the most part, the users who log into it in our company are people on the security operations team. It's pretty much a closed tool. Access is limited to the people in the security center of excellence. In terms of the solution's ability to reduce alerts by rolling up numerous alerts to create a single incident or campaign for investigation, we don't use it that way. We've set up enough triage filters over the course of the last year-and-a-half to get all the noise out of the way; stuff that is either innocuous or really isn't bad. Then we're focusing on what's left, which is typically, for lack of a better term, the bad stuff or the stuff that we need to pay attention to. Regarding the solution's privileged account analytics for detecting issues with privileged accounts, we've used it, but not to the extent that we would like to. We just don't have enough manpower to be able to do that at this point. But it's important because we can see when an account is doing something that it shouldn't be doing, or that it doesn't normally do, or that it's connecting to a place that it doesn't normally connect to, or that it's escalating its privileges unexpectedly. We see all that and then we can respond accordingly.
People do a lot more than we actually see. Looking at the test and development guys, sometimes they do things that they don't understand. So, they will do it because it works. The actual things that are behind the scenes are the sort of things that happen, and they don't really understand. If there's something that's really complicated, they're people that have initiated it that don't really know what it is. That is always a problem, because in our sort of company, we have a lot of developers who are doing a lot of coding and things like that, but they're not 100 percent on all the other things that they affect, such as the supporting applications underneath it. They are making a change on one particular app, but it's using the other apps underneath it to develop that and push that across to something else. All these extra, different steps that they are completely oblivious to where we go, "Actually, you've just done this." They go, "Well, I don't know, I just ran the script over here. I don't know why that would happen." But, it'll do a LDAP lookup or connect to a share. Those are the sort of things that you get a lot of visibility from people who don't understand. So, that can become tricky. That's pretty much par for the course for a lot of security tool sets. Where you have a couple of people who know one particular aspect, but don't really understand everything that's going on. To be fair, IT is a big area. You can't expect everyone to know everything of everything, not when you're not working in a massive IT structure, and the security team is a small department. You need to be quite key on your business case and what you're expecting from it. Be 100 percent sure on your use cases. It's an excellent tool. It doesn't create a huge amount of overhead, but it is a tool that you need to keep on top of. The more you keep on top of it and get it right at the start, the easier it will make your life going forward. Don't just stick it in, then leave it to whirl away as a lot of people do. You have to spend that bit of extra time, and it's not huge amount of time, and leverage other teams. The way they do their customer success is really good. There's nothing bad that I've got to say apart from the costs, but nothing's free, is it? It has to be up there with my favorite security tool set at the moment. I am quite lean on scores, but the solution is definitely nine (out of 10). If I look at all my other security tool sets, this is the one that my guys value the most.
My advice would be to really utilize the support and collaborate with Vectra. The solution requires heavy usage and customization to your environment. They provide the guidelines and you just have to be able to fill in the specifics. If you don't do that, it's not an effective tool. It is a really hands-on tool. Vectra has done a really good job of giving you visibility into the type of behavior into which you want visibility. But reducing the number of alerts really depends more on the analyst who is operating it and working with it. As for its ability to reduce false positives and help us focus on the highest-risk threats, the term "false positive," especially in this scope of machine learning, doesn't seem to me to apply. Vectra gives you visibility into what you want to see. It gives us visibility into the exact behaviors which we sometimes have issues trying to create detections for on the host. And on the network it's collected and brought it all together. We get really good visibility into all of the risky behaviors. Vectra provides the whole context, on the network, of what it sees in terms of a risky behavior and provides a story with it. In comparison to some of the other tools that I've come across in this category, I would definitely give it a 10 out of 10.
One thing we have learned using Vectra is that anomaly detection is a critical component of security; a non-signature-based technology is very critical. It helps pick up things that other tools, which are more focused on active threats, will miss. That is one major lesson that we have picked up from Vectra. My advice would be that you need to focus, because the licensing is based heavily on IPs and area of coverage, although predominantly IPs. You need to have a very clear idea of what areas you want to cover, and plan according to that. Full coverage, sometimes, may not be practical because, since it's a detection tool, covering everything for large organizations is complicated. Focus on critical areas first, and then expand later on. Also, the architecture part needs to be discussed and finalized early on, because there is a limited flexibility, depending on which model you choose to take. The solution captures network metadata at scale and enriches it with security information, but the full realization of that will come with Cognito Stream, which we have yet to implement. Right now we are on Cognito Detect. Cognito Stream is something that we are working on implementing, hopefully within the next month or so. Once that comes online, the enriched metadata will have greater value. As of now, the value is there and it's inside Vectra, but we don't see that information — such as Kerberos tokens, or certificates, or what the encryption is — unless it leads to a detection. Only in that event do we currently see that information. The Cognito Stream can feed into our SIEM and then we will have rich information about all the metadata which Vectra has in our data lake.
Make sure you have a dedicated resource committed to daily use of the tool. Because the selling point is it frees up your time, reducing the amount of time you need to spend on it so you don't have to commit resources. Then, you find yourself in an implementation two years later and you don't have committed resources who use it daily or are committed to it full-time. This means you don't maintain things like the triad rules and filters. Even though the sales material says it makes it easier and reduces alert fatigue, it doesn't give more time. You still need to have a dedicated resource to operate the tool, which we never committed at the beginning. Having an established mature team structure is really important as well. Making sure people are aware of their role and how their role fits into the use of the tool is key. Whereas, we were building a security operation center (SOC) at the same time that we took on the tool, so our analyst activities have evolved around the incorporation of the tool into the organization and it's not necessarily a mature approach. I would rate this solution as an eight (out of 10).
Start small and simple. Work with the Vectra support team. The solution’s ability to reduce false positives and help us focus on the highest-risk threats is the tricky part because we are still doing the filtering. The things it sees are out of the ordinary and anomalous. In our company, we have a lot of anomalous behavior, so it's not the tool. Vectra is doing what it's supposed to do, but we need to figure out whether that anomalous behavior is normal for our company. The majority of the findings are misconfigurations of servers and applications. That's the majority of things that I'm investigating at the moment. These are not security risks, but need to be addressed. We have more of those than I expected, which is good, but not part of my job. While it's good that Vectra detects misconfiguratons, there are not our primary goal. The solution is an eight (out of 10). We don't investigate our cloud at the moment.