Vectra AI Platform

Negative

We are using it for our SOC services. We are also using it for our clients. We have our monitoring setup for our SOC staff.

There are many detection features available. There are extensive out-of-box detection capabilities. I cannot mention just one or two at the moment. There are multiple detection rules, and its integration with ADR and Office 365 AI is very nice, to be honest with you. It is scalable, and they have their own appliance that can handle multiple locations. You can deploy it for enterprises with multiple sites.

The advantages of the integration are not entirely out-of-the-box. You have to do it manually. When I'm doing tier response, an out-of-the-box solution is not available. You need to have a Linux server, and from the Linux server, you must perform AI tasks, and there is a lot to be handled in the back end. This is a major consideration about them. The recall feature, if it can be placed in some areas instead of the cloud, and charged for, would be better. Recall the storage where you watch all the traffic, and you can recall it and try to analyze it in the back end. It’s cloud-based. If they offer it on-prem, it would be better. I think they have a solution, but I have never tested it, to be honest with you.

I have been using the solution for years.

It is scalable, and they have their own appliance to handle multiple locations. You can deploy it for enterprises with multiple sites.

They are supportive. From a support perspective, they are supportive, to be honest with you.

I am using something else. I am using Vivo, Vixstrap, Vextra AI, Vectra, and Security Onion as open-source. It depends on the clients.

It is very acceptable when you compare it with Darktrace, for example.

At the end of the day, it's written rules in such a way. The trend in the market is something I did not consider much. The detection rules are written in the back end. There is something happening in such a way to do it again. AI is mentioned too much, and for me, it is only marketing talk. At the end of the day, there is no one hundred percent AI in security. Detection requires manual writing at times. They already handle back-end processes but vendors won't show this. AI is not targeting a specific vendor. AI, for me, is just a trend. It depends on the client. I tailor solutions to client requirements. For visibility and monitoring, I choose the best products. Every application, every NDR solution has its capabilities. It varies by client because I must advise clients on solutions they can use and benefit from. I sometimes advise clients about Vectra as it still serves my clients well. It's fair enough for now. The overall product rating is seven out of ten.

As an end user, I do not have to commit manpower to manage Vectra since most of their use cases are managed by them. It's a hands-off kind of deployment.

The deployment is hands-off, which means it saves us manpower resources since Vectra manages the use cases.

Most of their use cases, including deployment, are managed by the tool itself, requiring less manual input from our team.

Neither Vectra nor Darktrace  have a function like a status health check on my log sources and traffic sources.

I have been working with Vectra for one or two years.

It's pretty good with no major issues.

The support is quite reliable depending on the service engineer assigned. I would rate them between eight and nine.

Positive

We are also working with Darktrace .

The setup is generally straightforward.

Vectra is cheaper in terms of pricing and features compared to Darktrace.

Vectra was compared alongside Darktrace.

Vectra serves its purpose well and does not require much manpower for updates.

I'd rate the solution eight out of ten.

We had another product with Vectra AI and used the MDR solution as an add-on. Initially, it wasn't fully appropriately configured, so we didn't get the expected results. Even once configured correctly, we weren't fully satisfied with its response. The issue was both with their service response and the product's capabilities.

The solution's weekly reports needed to have more explanations. However, we needed more explanations because the reports provided were mainly statistical. We were looking for more analysis and insights.

I have been working with the product for less than a year. 

The initial setup was pretty straightforward. 

The solution's pricing was 50 percent lower than the other vendors shortlisted. 

I wouldn't recommend the product to others. We are moving away from it. I rate the overall solution a six out of ten. 

We use Vectra AI for endpoints where we are unable to install agents, like endpoint agents, EDR agents, or antivirus tools. For example, BYOD devices or routers in our network. We don't have any control over those, but we need monitoring capability. 

Vectra AI can monitor the traffic from the wireless router to the firewall or any outgoing traffic. It can give us an idea of whether there is any C&C or C2 communication or any botnet activity from those source IPs. Without having any agents in the endpoint, it is a network monitoring tool. We use this tool to detect threats within the environment where the assets are unmanaged. 

Also, since we tap into certain network points such as firewalls or IDSs, we get more visibility from managed assets as well. So before the endpoint notices the behavior, Vectra notices some of the exfiltration techniques and alerts us.

Overall, it is good and has reduced our time in identifying the system. It is for unmanaged devices. Previously, if we got an alert from the firewall, it was very difficult to find that particular asset. But with the help of this tool, we can simply run a packet capture and immediately get the hostname and know which user is using it. 

It has greatly reduced our time to remediate the situation. We can identify the user, block their account immediately, and sometimes kick that device off the network completely.

It has a confidence level of around 60% to detect insider threats of anomalies, but we mostly need to fine-tune the product. We are still in the fine-tuning process. Even though it has been one year since we implemented the product, the first six months were spent integrating various log servers and determining where to tap. 

For the past three months, we have been actively investigating the alerts. When we investigate some of the insider alerts, most of the time it is a false positive because the domain is allowed. Vectra does not know that those are allowed domains, such as OneDrive and SharePoint, to access our network devices. 

It considers it malicious because a huge amount of file uploads is seen, according to Vectra. But we know those are known URLs and known behavior. When we slowly started whitelisting, the threat confidence level increased. So right now, for insider threats, it gives around 60% confidence, but around 80% of the incidents were false positives because we are still in the fine-tuning process.

The packet capturing feature is very useful, and as the name suggests, AI uses models to detect abnormal behavior. Some of the patent-matching algorithms they use are very advanced and detect threats at a very early stage.

For me, detections from unmanaged networks are one of the greatest values. You can identify threats from BYOD or even mobile devices, which were not handled before.

The detection algorithms can be improved at the sensor level rather than doing all the things at the brain. For example, if the sensor has some directional algorithm or detects repeating traffic, it can drop those packets at the beginning itself. There is no need to send that traffic to the brain in order to reduce the bandwidth.

AI is picking up a lot now. There is no manual intervention needed. Whenever a detection happens, it can automatically summarize and give it to you. But Vectra doesn't have those kinds of capabilities. It still needs manual intervention to analyze, and they don't have a summarized kind of output. So that can be improved. But apart from that, the detection models and all the other categories have good support for that.

In future releases, I would like to see Vectra AI to generate a summary of the instance.

I have been using it for a year. 

I would rate it at eight. The remaining two points I'm not giving because it's a fairly new product. So far, it is good as per our test and it is able to scale as well.

The only limit is you need to increase the sensors when you have more traffic. For example, the current sensors can handle up to 50 GBPS of traffic per second. If you need more traffic to be utilized, then you need to buy additional sensors to handle the traffic.

From a technical perspective, there is not much more possible, because there are some hard limits in the hardware. You cannot increase the bandwidth. They have other options to increase with more sensors, but it ultimately ends up being a cost factor.

If you have more money, you can buy more sensors and do it.

In our organization, we are an MSSP provider. We use Vectra, and our entire SOC team, which is around 20 people, uses Vectra for our MSSP. We have two customers who are also using this product. Two of the largest telecom industries in Thailand are using this product to understand their behavior as of now. The approximate number of users in those categories will be around ten.

The customer service and support are good. So far, we have not faced any issues at all.

The setup is a very straightforward process. You need to tap the network traffic at your desired point, and it has two components: a sensor and a brain. The sensor collects the logs and forwards them to the brain, which does the detection and everything. They offer a virtual appliance that you can run in your environment. 

The setup process is usually very simple. It took only two days to set up. But, initially, deciding the location of the sensor and other factors took more time. The threat team at Vectra AI engaged with us effectively, provided all the support, understood our architecture and advised us on placing the sensors.

The licensing is on annual basis. 

I would rate it at nine out of ten. The one point I'm reducing is because the model can learn itself. If no one is fine-tuning it, for example, every time we find a huge number of alerts, then only we go and look it up and fine-tune the product. 

If no one is acknowledging it or it seems like regular traffic, then the product can understand that behavior and have a feedback mechanism to correct it, mark it as a false positive, or whitelist it.

My recommendation: 

Understand your network first, and place the sensors in the correct position to receive all kinds of traffic: THC, PDNS, and all those things. If you place the sensors at the egress traffic, you may not receive some of the packets, and you will not have overall visibility. 

So the placement of sensors is very important; you need to understand your network to place them correctly.