Thank you so much for being here this early in the morning. A lot of people think there are no mornings in Vegas. We're proving them wrong consistently. And anyway we're going to talk about measuring the IQ of your threat intelligence feeds. If you are from the internet and you do the Twitters we have, like, a hashtag so you can hash tag the talk and making fun of us. Please do because I want to laugh at everything later because I can't check Twitter here. It's kind of lame but that's the way it is.
I don't like talking about myself. I just brought here a very special guest. This is Wendy Nader. She going to do the introductions for us.
>> You didn't bring my fez and my cup of rum. All right.
Good morning again everybody. I want to introduce these awesome guys. I am a little annoyed with them because in my analyst role I cover threat intelligence and in this talk is just something that I wanted to do. But they probably did it a lot better anyway.
We have Alex Pinto otherwise known as the brain from Brazil. Is that right? He's been giving talks in all three conferences this week. And people just can't get enough of him. And his hacker spirit animal is decaffeinated Kathy Barra. Kyle Maxwell, on the other hand, is a math smuggler. His spirit animal is the axiomatic armadillo. I'm here to tell you he is a fugitive time traveling Karl Marxs and there's a reason why his initials are KM. Sorry. I had to out you like that. So enjoy the talk.
>> Thank you so much. Anyway, we're going to cover a bunch of stuff. We're just going to give a brief introduction about threat intel and really go into what is the crazy idea about measuring threat intelligence. Anyway I'll just ask Kyle to do his magic.
>> Good morning everybody. So we're going to talk real quick about threat intel. We didn't want to do threat Intel 101 for those who have thought about it before. Honestly the stuff we talked about today, talk about that what that really means. When we talk about threat intel, or intel in general there's two things we really actually care about. It's not IP addresses. Its capability and intent, right? What can your foes, I hate the word "adversaries" and all of that. So I'm going to bring it back Viking style. What are your foes able to do and what do want to do? That's what you really want to know so you can prepare on your end for whatever it is that you need to do.
So that's the core idea that what we're trying to measure, honestly most of the time what we look at threat intel is capability. What are they doing? What are they able to do? We don't spend enough time looking at what they can do or rather what they're planning to do. So we'll talk about that a little bit. There are a couple of key concepts that I want you to be able to take away where there are some dichotomies within this field. So we're going to talk about kind of these cage match between things that oppose each other. One of them signatures versus indicators. So Dave Vital for example has made the point very correctly that a lot of times the way people use indicators or indicators of compromise sounds like signatures, right. But that shouldn't be. If you are using them that way, you are doing it wrong.
The idea is that indicators rather a signature says this is definitely something that happened here. It's, like, a fingerprint, right. You have a very, very high degree of confidence that this means X. Whereas an indicator says look over there, there's something fishy going on. And you may have different levels of confidence. You may even have a pretty dang high level. But the reality is that it's not supposed to be used for blocking. It's not to be used for complete confidence. It's supposed to indicate something that you need to go look at.
The other thing I'm going kind of deflate a little bit here is the difference between threat data and threat intelligence and data and intelligence in general. Now we talk about threat intel feeds here because that's what vendors typically label them as. But they're generally speaking they're doing it wrong. It's not threat intel because it's data. Data without context is UGH. If you want Intel you need to do a lot of things, right. You need to be able to establish some context around it. You need to be able to add some of that intent and capability understand and we talked about it a minute ago. And as we're going to see you kind of escalate up the abstraction of what you are looking for. But since this is what vendors sell it as we want to make sure people know at least what we're coming to talk about.
There's also a difference between tactical and strategic intel. Tactical you can think of as the how and the what. Like what they're attacks, how they're going about it. Whereas strategic intel and this is a continuum not a binary either or, towards the strategic end, you have the who and why. Who are my foes? Why do they care? And why do I care about them? This continuum is important to understand what type of Intel you are getting. And then very briefly we talk about atomic indicators versus composite. Atomic indicators are an IP address. A packet stream you are looking for. A hash And composite is when you pull these things together and now you start to get some value out of them because I really don't give a SHIT about the IP address but I do if I see it combined with other things.
A colleague of mine has a concept he calls the pyramid of pain! And I always use it because I like to say it like that. The pyramid of pain! And the pyramid is basically this idea. At the bottom of the pyramid and you can quibble and Alex and I do, about what is easier and what's more trivial, IPs or domain. The honest truth is it really depends on content if you are talking about domain generation algorithms versus phishing. or what have you, but the idea is Either way they're kind of at the bottom. It's pretty easy to get a lot of IP addresses. Right? It's also easy for foes to change their IP address. As you move up this pyramid of pain you get network artifacts that are a little harder for you to analyze and figure out what to look for. And a little bit harder for the attack to be able to change on his end. Whereas now when you start to get to specific tools of TTP whic are tactics, techniques and procedures, how these guys go about what they're doing, that's much harder. It's much harder to change your process than it is to flip a bit and have the hash come out completely differently.
Now, that said, we talked about simple, easy.and trivial We're going to see it's not always they're simple and easy. But relative to these other things, right. With that said, I'll hand back over to my brother here to go into his analysis.
>> So, I mean, just a side note. I think if you did like do the pyramid of pain in a falsetto voice it would be much more funnier. But that's just me.
>> That would hurt my throat.
>> Okay it's your throat.
I guess the point here is like Kyle was talking about, I mean, should we care more about the IP addresses and domain names if that's all you got? Ideally all of our intelligence would be a bunch of contacts in it. We'd all have animal names attached to all of them and it would be awesome. We would really understand what we're up against. I really want to work with IP addresses here. They have a bunch of interesting stuff we can look for. And the thing that is interesting for me, it's finite resource people are always bitching about iPv6 stuff like that. The fact is there is an intrinsic cost of getting and changing an IP address. Some places it's easier or less easy for you to get an IP address to do that stuff. Right. So people are quite on to it. Some data centers, some providers, if you try to do something a little weird they'll shut you down almost immediately. Others you like to have to talk to them we've been here for three months while they're ignoring you and stuff stay there's forever. Maybe there is some structure, some patent there. I'm actually proposing that if we're getting a bunch of these IP addresses why don't we try some experiments in them to see if we can learn a little bit more about what thus threat intelligence landscape is, I don't say that in the threat intelligence landscape like all capital letters like what the world is doing against the world in cyber war. But who is actually trying to get to you. That's the thing you should be most concerned about.
Anyway let's try to do some science here. Can we measure how much a specific feed is going to help? How it's going to help us in our strategic defense of practice we're trying to do in our organization. Can we actually get something out of them. Are they just a bunch of numbers with DOTS in them? How can we actually measure? I have some ideas and I wanted to share those ideas with you.
So actually I put together all of this is GPL3, a bunch of statistical tests and data visualization that you can do, You do have to use R. I almost did it on purpose so you guys start to get to know some statistical languages. Get with the times. In five years it's all going to be data. I'm sorry but that's pretty much it. There's some phyton dude beside me who is just going to be bothering me the rest of the talk.
All the code is there. The sample data that i used for the talk is there. So you can rerun everything and say there is a bug here youre such a stupid guy, and stuff like that. But look how I know how to use this library much better than you. You can do all this awesome stuff. Also, all the graphs and all the data that I show here is in the mark down file. So it's pretty much here is blah, blah, blah. Here is the code, the results. you can actually follow thru if you like It's all in the repository. There's a link to everything. Let's get started. I think.
Just for you to have a general idea how did I structure this data, how did I try to, like, create a sample database or something like that? First of all, I was trying to extract what I am calling the raw information from this feed, right. And I extract both IP addresses and host names. Just a side note, all of this data is publicly available, right. So all this stuff that I've gathered, and I'm going to show here, you might as well just gather. And we actually have a treat for you by the end of the talk that we might just need to help you a little bit with that.
But the point here is, okay, you got your iPv4s you git your FQDNs I did this weird thing. It may sound weird at first when I am calling stuff inbound or outbound. And it's pretty much I'm just trying to differentiate very broadly between the types of threats, let's say, that these IP addresses and domain names provide me. not so much domain names for inbounds but uuhhh So think about the way the data flows from your organization or your IP or your dingly tingly internety thingy think about all the direction as things go in and go out, right. So, if it's someone that is spamming you I am calling that an inbound feed. If you have to access this website to download this malware or it's a dropper or CNC I am calling that outbound. and thats pretty much the list of the feeds that i am harvesting, There is a mistake in this slide so you are free to call me out on the internet if you figure it out. But the point is, and all of this is documented which feed is which but if you guys have been playing around with this you should recognize most of them almost immediately.
How did we prepare this data for the experiments we're going to talk about? First of all, we I only wanted to work with IP addresses. First because there's more interesting topological stuff that you can work with as opposed to domain names. I do plan to do something similar to this using who is data in a future installment of this specifically for host names. But I was only willing to work and frankly only had time to work with IP addresses. So I just ran all the domain names through the passive DNS database. Particularly the far sight. For that day that I got this indicator, what were the IP addresses active that day, right. Assuming you had the data and were taking some measure of that data, that would be the IP addresses you have been flagged as weird, malicious or suspicious or however you take this threat intelligence. So we also removed non-public IPs we dont care if the IP address results to local hosts that's obviously a parking technique but it wasn't really conductive to what we were trying to do here.
Anyway, we got this domain names and IP addresses and we made them a bunch of IP addresses. After we do that we went to do some enriching on these guys. So we want to make sure we were able to identify what were the ASs that was associated with the IP addresses on that specific day. So there's a period of time that we have the data for and that's when we did those, we did those enrichments right. Also country from geo occasion databases. We also resolved the remote host also using passive DNS at that time. And I had some plans for it but we didn't wasn't really interesting. The data was too sparse. There was not a lot of cool stuff to be shown.
Just a side note as well. Although we are doing geolocation there will be no maps at this presentation. Maps are not a good way to look at data. Please, stop doing that.
I mean, if the country is big you think they're more of a threat. That's how the brain works. They look at stuff that's bigger and more important. Stop using maps. Okay let's move on
So, yeah, this is a sample, right. And you can see exactly how easy or not it is to actually use the code that I put together to actually okay I'm going to load in from my database and I want to see it and see what data is in it, et cetera, et cetera. You can see there some examples of things that people thought to be bad in a specific moment and time.
All right we got our data. So what are we going to do with it? I propose three different experiments here, right. I'm calling them novelty. It is a pun because it has both meanings. You can use it because that's how much this feed is refreshing itself as time goes by. Or maybe it's just like I am going to add a bunch of stuff and take a bunch of stuff off. and make it look really active But anyway you got to know how your feeds are updating themselves. You want to make sure that whomever you are getting them from be them open source or commercial, the guys actually putting some work into it. Because, I mean, things move around.
There's an overlap test which is one that I mean, it's the most simplest thing ever. Just, like, let's compare everyone against everyone and see how much they have the same indicators, right. This is specifically useful if you are trying to buy something, right. Just get all the open source feed that you can handle, compare them against the commercial feed, and how much am I actually getting for X. I completely lost it. But like this ridiculously amount of money I am paying for this. The one that is much more exploratory there a little bit more math than the other ones which I'm calling the population test. Which is can the data we're getting actually teach us something about our adversaries? Can we actually learn something from them? Sorry. Sorry. Foes. Foes. Can we actually learn something from this? Can we actually find some patterns on this attacks? Anyway, population is tricky. I'm doing some examples here where I'm comparing against the whole population of IP addresses where they should be at least. And also should ideally mean your data. So get your data and compare how are the IP addresses that your companies accesses or accessed by your company against actual threat intel feeds. Anyway, we'll get there. I was looking for an IQ. I wanted a grade. No we're not going to get there. Specifically, I want to make sure that you guys understand that the idea here is to make sure that you guys are able to reach your own conclusion, right. Whatever I think is right for my company as far as the specific mix of threat intelligence or i should be focusing on this or that is going to be very, very different from whatever you are doing at yours. I just want to make sure that you understand, feel free to publish your own Metrics. Feel free to tell X Y X sucks. I'll try not do this because I kind of like data and I don't want to piss of everyone at the same time. I try to get on their good side. if you wanna meet and do this work for you i can help you out
>> youre so nice.
>> Well I don't know.
Anyway, I do intend to--I will do more private research on this because there is all of this comes from the other researching that I do. So I actually had to go through all of this. I'm explaining which were the threat intel pieces I would use in my models. Just thought it would be interesting to share.
Anyway, novelty test. So we're measuring pretty much added and drop indicators, right. So how much stuff is being added day by day and how much stuff is being removed day by day, right? And if you run this kind of thing you will get something like this. This is specific for inbound. So alienvault is actually divided because they got both stuff in there. It's majorly inbound. There's a lot of inbound stuff. And it's because it's fairly easy to get like you just run a couple of honey pods and suddely you are like threat intelligence all the way to the bank. That sounded worse than I meant it to sound. But anyway. The point here is you want to make sure that the stuff that you are doing is actually refreshing itself with something. So you can see there just like a 20 percent refresh in one of them. Sometimes 50 percent. You definitely do not want the guy on the lower right. I mean, they're not dropping any IP addresses, right. The red line is how much it's being dropped. So it's an upside down and the blue graph is how much is being added. This guy is adding a minimal ratio of stuff. this is only percentage, this is not an absolute values And it's not removing anything else. I'm not sure the guy is doing the work. It looks abandoned to me. So it might not be a good one to actually keep using as you go forward.
But not updating every time does not mean it's bad, right. So this is an example from outbound. Also you got stuff moving up and down. You got something adding or not. And as a rule of thumb if something is being updated hourly or being updated daily there's a lot of out related stuff going on there and it's potentially getting information from other sources as well. And it's being presented to you with little or no human interaction. If it's something that's more like weekly or something like that, maybe there's some people actually doing the work and computing and everything. Of course when I say all of this I mean specifically about open source, right. If people are getting paid to actually deliver this to you, they might as well give me hourly. humanly created feeds all the time otherwise i should not be paying for them at least thats my opinion
It's cool because it can see patches. How malware domain list. and malware domains they actually have a huge spike in a specific day, If you guys are wondering that's was right about the time we had that no IP thing with Microsoft and people got to know, like, a bunch of new domains and everyone started sharing it. So you can actually see when stuff happened in a way. It's kind of cool. Anyway, do this test. Run this and see how much your feeds are updating and see if yeah does it sound right for the sizes of the company or for the kind of data that your reading that they are actually looking at it It is a lead indicator of quality. It's not the last one.
We can talk about overlap, right. More data is better but there's so much you can give your hadoop vendor to actually store all the data. So let's try to keep it only add data that is not the same or it's useful. I mean, it's easy for me to say that because I don't really care. I do lot of summarization for people the way that i consume threat intelligence some people use the fact that something appeared in multiple feeds as a quality issue. A bunch of people read it. They might not have all done the research. Or just want to leave it out there. But I don't know. I honestly don't know. It's just that you have to make sure what you are looking for. So when you look at inbound you get stuff like this. So the brighter the square there's more overlap between those specific feeds. And this was a specific day in time. And you can rerun this test for all the 30 days that I have on the data set. And you can take your conclusions about, I mean, who is potentially drawing this information from the same kind of honey pod or some kind of research, right? I have no idea what's going on in the sense that this guy did that. And honestly I don't really care. I just want to make sure that if you are ingesting this and if you are working with this you are not just, like, oh yeah I got two minute indicators here. All the IP addresses in the world. But it's, like, yeah a bunch of duplicates and that's not really helping your bottom line.
Something that was actually I found was very interesting and it actually relates to information I heard privately. Is that when you look at outbound, there's much, significantly less overlap, right. So it seems that peoples aren't doing their research or work on some specific issues. And specifically some people they will specialize in specific threat. Oh yeah I'm specific whatever hunter. And but okay. Sorry. And what we get here is a little bit of the domain only you do get a little bit of overlap.
I just wanted to point out that specific day that we had that spike. On that day there was actually a 25 percent of overlap of potential indicators between malware domains and malware domain lists. Because they all updated quickly to add all that information. Which is very good, very positive. Then you see that result also represented here, right. Because you can measure the overlap those feeds have. And also you can also from this graph you can very easily see that a lot of the stuff that the WCH guys put on their Zeus feed is also being used by alienvault as well. Which also I think is good. I know a lot of people who use it and a lot of people who use WCH stuff.
Anyway let's get to something which is a little more interesting I think which is the population test. And I guess it's interesting and this is kind of a cornerstone of the research that I started doing a couple of years ago. Which is, you know, maybe, maybe stuff isn't random, right. Maybe it's not like maybe we can actually track. There's some sort of likelihood that some specific corners of the internet are more likely to be targeting you or not. And all the different information that you get are from the providers, they will give you different information. And there is some sort of alignment that I believe that you should be trying to reach when you are talking about here is my data, right. These are all the IPs address that im communicating to. Potentially the threats that I have, the people who are targeting me or the people who maybe have already reached me, that are exfiltrating data they are probably from that pool. It's not going to be like an IP teleporting thing. I am not talk about which type of animal web or from the from the zoo you come from. I am talking about at the end of the day does it really matter. But there has to be a way that the data is being exfiltrated the IP address is there. The guy who is out to get you, who is potentially messing around with your data is on that list. So why don't you try to tailor the threat intelligence and the research you are doing through the specific countries ans ASNs that actually make more sense to your data, right.
So then we start doing some comparisons and doing some tests, right. And we've come up with some interesting things. I specifically separated the domain, the outbound and inbound stuff here. Of course on the inbound China all the way, right. And I can see some people smiling here. Yeah. And at least in my perspective the actual penetration of those kind of things are people who are doing SSH scans or they have, like, this three year old joomla or exploits they're running against the whole internet. I don't think those guys are really going to get in. I mean, if they do you shouldn't be in this talk you should be in, like, 101 or something. But the point I am trying to make is when you start looking at the outbound data it's not that different from it is different from some but the fact that China wasn't the first one and the first one was the U.S. actually made me chuckle. Actually, if you squint you can see that the actual proportion and I'll get there. The actual proportion of Chinese IP addresses which are actually distributing malware is less than the actual Chinese proportion of IP addresses on the internet. So, I mean, maybe they're less likely to be posting malware than other countries which I think is kind of funny.
Anyway, you can play the same game with ASN, right. And the prevalence of Google there obviously startled me. At first I was looking at that and was, like, okay this is a bunch of punk domains. Everybody knows that. When this stuff, domain names they are they're not being active and used, like, they're changing their infrastructure. Someone just took the BOTnet down and put something else together. just spark it all like 18.104.22.168 or 22.214.171.124.1 or something like that and those are all owned by Google. So at any moment in time like a very high percentage of your domain names from these open threat intelligence feeds will be pointing at Google. Be aware of that.
But, even so, I don't know. dont be evil and all that, im with them The thing is I guess the point I'm trying to make here is this kind of population test, they can also help you to key weigh your feeds. Right. I'll get to the specific of that. I just wanted to show you the graphs and now I'm going to get to the math. I'm sorry about this.
But the point is I don't like to squint, right. So let's do some statistical entrance testing. Let's compare the proportion, right. So, if I know the exact proportion of IP addresses on the US, right, just get any geo IP database and make sure you are, like, keeping with it, make sure it doesn't enrich in one database and use another one we can do an exact binomial test which would be okay assuming that the I can just choose randomly the IP addresses that are going to attack me, is that consistent with how the IP addresses are distributed? Even if I don't have the full population, right, and I'm only like comparing my sample of my data against the sample of the threat intelligence feeds, I can use something that's the chi-squared proportion test which is similar to the independence test. Which pretty much ask the questions these are two different samples of stuff. Do they look like they came from the same population, right. And it's a little bit more fuzzy and fidgety than the other one. That's why the other one is called exact, right. But it's a statistical tool that you may have.
So I really wanted to get an idea of our error measures. I wanted to get confidence intervals and say it's something along this line. But at least I can tell you I'm messing it up here in this calculation but this is pretty much how much I am messing up. I can also use p-values. I personally apologize to Alex Hutton for using P values and not like doing some bayesian evil shit. But he said it was okay. So I'm allowed to do this. Some people have to clear it with their lawyers I have to clear it with Alex.
So the point is, if we're doing p-values we're being very conservative. So for the number of tests we do, so what does that look like? we're assuming 95 percent because it's pretty much what everybody does. If we're doing multiple tests we're actually dividing our critical value or the value that we use to oh there's something interesting here by the number of tests that we do. So what does that look like? It looks like something like this. So I have this I'm using this functionally extracts the population from a threat intel feed. It's all there. And I'm comparing it to my known MMGU database. And lo and behold, it seems that Thailand is a winner right now. Which maybe the proportion is five percent more than the number than their actual proportions on the IP addresses there, right. Followed closely by the US. Russia is there. I'll give you guys that. But you should go to China. China is the second least likely to be hosting something.in comparison to the random proportion, I don't know maybe it was not these feeds. Maybe these feeds were not tracking specific addresses for there. But that's not the point. The point is if that's the data that you have and the data we should be comparing, that's what you should be analyzing. You want to add something? You were just gasping. Did I say something wrong?
>> No you are good. You are good. just keep presenting knowledge bombs
>> Co-presenting is hard. The guy is judging me right beside me. I can't really see you because of the lights. Like, why did I get into this?
If the feed value is higher than critical value which is the last one here it's pretty much we can't really say anything. It looks like it's on par with the randomness that we would get. A challenge we got when we reached this conclusion is that hold on, I don't know what to call these countries yet, right. Come on. I mean, we do not have, like, ugly animal names for all of them. So we propose a guide. So as you are trying to create your own and you are creating your own threat intelligence with those statistical techniques you can actually call people and actually view people to what you are doing. Just clarification. Capybara. is like south south of brazil and toucan is more like the amazon part, its very very different hacker spirit animal just be consious about that.
Where are my Texas people?
>> This is not an exhaustive list. So feel free to tweet your own suggestions at the hash tag of the talk. We got to catch them all.
Anyway, then we get to the not exact test. And this is just a little thing that I did trying to compare how this feed evolved and changed from the 11th of July to the 12th of July. Those are the same feeds. What did change from the one day to the next? You see a bunch are pretty much the time on the bottom side. Thailand again, there's actually a trend from this information you are getting from your threat intelligence feeds. That Thailand is getting more likely to attack. Based only on the data and geo location. So, if you got I can see some of you guys have got, like, the brain thing going on and that's exactly what I'm hoping for. There's a lot of interesting stuff and a lot of interesting conclusions you can get from your own data. Not so much the threat intelligence feeds also trying to do something like this.
While you are at it at the animal thing, we do need some for Thailand. Because they're everywhere on this presentation. theyre obviously up to something bad Tweet it. Tweet it. We can't really hear you.
So then I wanted I was bothered by this Google thing right and I wanted to make sure that I could understand if this is all (inaudible) or something different. So I ran the same test, the same exact proportion test on the AS information as well as the population information. And yeah Google's like ten percent proportion increase. That's crazy. That's too much. That's way, way too much. And then, what?
>> Okay. Where is he? What we get there is if you actually look into the data you will find that maybe not so much, maybe it's not just park data, there's some cleaning that needs to be done. I am not the biggest fan of chrome but I'm pretty sure it's not malware so probably we shouldn't have chrome.google.chrome in our list but Its alright, it happens sometimes. You just got to be sure if you are consuming this you need to look out for information like this. And maybe they can help you do a little bit of triage because no one is going to look through the whole thing all the time. I am talking about a bunch of stuff and I talked about my data. What about your data? How can you get your data to use those? For that I'd like to invite back my good friend Kyle Maxwell.
>> Oh my God thank you Alex.
So as you are pulling in these feeds you want to be able to do this yourself. You want on going data in addition to be able to run these analyses for yourselves you may have other uses for the data. In this case we did a lot of the work for you. But ongoing we need something to help you with that. So what we did is we are releasing this week a tool called combine. You can get it right now. its a TPO ver. 3 And what it is, it is a tool for harvesting and processing these types of feeds.
>> I just want to point out at the ML sec project we're very serious about our needle in a haystack joke. so this is why this is a combined harvester I am not implying, well maybe I am, that's all hay.
>> So there's a few different components to it. reaper pulls it in We really, really like this metaphor here. so we reap it in that just pulls in the data directly, We'll normalize it to a very simple data model. we do a little bit of validation We throw out all the RFC 19, 18 shit stuff like that. Do the AS numbers, the DNS look up. We transform the data into CSV or J song. We haven't released the SY box or sim for splunk As this is an open source project, help us help you. Feel free to submit full requests for it. That's another output form that you want. We're constantly trying to feed more data into it. This can be your own data sources. There's a number of tools to bring in. CSV,XML, parsing HTML, what have you. We have a number of feed sources bundled in in terms of the URLs. We're not saying clearly based on the last half hour that any of these feeds are good or bad these are just the ones that we have and we'll add more as time goes on. When we do the enrichment, if you are not familiar with the term think metadata and metadata accessories. So we're pulling the autonomous system and geo location only down to country level from the maxline database. We do DNS resolution from the far site passive DNS database. That requires an AP IP. If you have one fantastic. Plug it right in. If you don't. Ask them for one. Tell them Alex Pinto sent you and he'll get you hooked up.
>> Alex, really good folks at far sight can help you out with that. Again, all of this is open source software. This is not commercial. This is not us trying to make a buck off these tools. This is, you know, us trying to make thing as little better and help you pull in these feeds and pull in your own data as well.
>> And, honestly, I don't think that's where the cool stuff really is. And I we had to build a lot of this as we were building out the research we were doing for MLSec project. We just figured out yeah but this is very commoditized stuff. or at least it should be very commoditized stuff, So we want to make sure that we are replacing our stuff our internal stuff is being replaced by open source versions and we will continue to support them as we actually need to use them in dot coding and all of that. Talk to us about the project if you are interested. If you like the specific TIQ-test you want to test your data against it. oh but i dont wanna publish my feed We can totally help you out. Either you setting up your own stuff to do that or we can help you out and do the test for you, right. And I just want to very quick takeaways before we are kicked out. Look at your data. That's all I'm asking, right. That's all you should be doing right now. And I will try and as much as I can to keep banging this point forward and actually providing helpful tool that people can use for that, right. i Also feel very strongly about the asymmetry of information. So, if people tell you something is good, don't take their word for it. Just try it, right. Just test it. And make sure that it's appropriate for what you are doing. Anyway, if you just got data lying around, you know, we can pick it up for some exercise. We can play some games.
>> Think about, like, data sitting like with dot sitting. We can totally do some data sitting for you guys.
Anyway, thank you very much.