Azure Sentinel webinar: Cloud & On-Premises architecture

Ryan: welcome everyone we’ll be waiting two to three minutes to get started as a courtesy to those who are still in the middle of connecting Ryan: welcome everyone thanks for joining us and thanks for being part of our community we have a great topic to cover today but first a few reminders please feel free to ask questions at any time by typing them in the IM window be aware that any questions you post will be publicly visible however if you prefer you can post your question anonymously by checking the box right below where you enter. We often get many questions on these webinars and we will do our best to respond to all of them in real time but I want to provide an additional mechanism to ensure any questions we miss get answered if you’ll visit you’ll be able to ask questions on our Azure Sentinel forum, if you’re listening to this after the fact is a recording that’s a great place to ask a question we always love to hear your feedback on whether these webinars are valuable to you and how we can improve them so you can give it to us at and all these links I’m referencing here are in the I am window for your convenience there. Please note that this webinar is being recorded and will be shared publicly we will post the recordings on our community at while you are there please join our community by visiting that’s the best way to ensure you don’t miss any future webinars or major announcements on our community you can speak directly to our engineering teams that create our security products you’ll be able to influence our product designs and get early access to changes by doing things like participating in private previews which you can sign up for at requesting features giving feedback reviewing our product roadmaps attending in-person events or joining webinars like this we believe that the best way to improve our products is by removing any barriers between you and the people that create them so we hope you’ll join us we have a great topic for you today we’ll be talking about Azure Sentinel specifically we’ll be focusing on Architecture both for the Cloud and on-premise you to our presenter Ofer Shezaf He’s principal program manager on

the Azure Sentinel product team so he has deep expertise in this topic without further ado I will turn it over to him over / we got it still on mute Ofer your still on mute. Ofer: I’m better now better now thank you yeah sorry everyone I I did mention now that you can hear me that you’ll spend the next 90 minutes talking about a sending architecture architecture it’s part of a series Azure Sentinel is a SIEM product there is a lot to learn about it and we are holding a number of webinars the first one was last week the recording is already available in the URL that Ryan mentioned and it was about the basics it was a deep dive but if you want to learn about the product go and visit the last week’s webinar as Ryan mentioned today will focus on a more advanced topic and will talk about architecture after the holidays break after everybody enjoys sonification we’ll be back with additional webinars one of them will be about and shock scenario which be more targeted towards analysts and security people rather than architects and we’ll show you how to use sent in in real life stock to detect triage and investigate incidents and lastly response to that and lastly we’ll get back to a topic we’ve revealed through earlier in September threat hunting using sentinel we’re also working to schedule additional webinars to keep tuned without further ado Azure Sentinel architecture. We have two large topics today to discuss in a way when preparing the slides and preparing for this presentation I thought that it might be too much for a single webinar and I’ll try to make sure that we cover as much as possible and hopefully in in a way that would be useful to you around both topics I may cut a few ER as a bit short to make sure that if you came to your boat cloud you learn and if you’re more interested in on-trend collection you also get the information you need as Ryan mentioned if you have any further questions either chat box or later on in the security community and we’ll start with collection Azure sentinel is a team security information in event management a solution it’s there to serve as the nerve center for your security operations center for your sock the most basic functionality provider to be that is to collect on a tree from anywhere in your a ite state it would be cloud microstrip dollars whereas other clouds as well as on Prem a devices since we are dealing with security naturally security devices are a important goal but that’s not the only source of events we want to collect data from for example finally an email activity are very important in order to take a large number of threats as residence supports all of that and the discussion will show different ways in which you can connect a sources and get events into a larger sense you know largely speaking I will divide it into cloud-based collection a non-prime collection I think that you’ll find that the cloud-based collection is extremely easy results are sent and we discuss how and to do that and then I’ll go into on-prem collection content collection is one of the main tasks you need to go for when you implement the same and understand there is no exception and we have all the tools you need for that I would not want to claim it’s simple a lot of tools out there so you want to collect from and it require some know-how which we’ll go through today but I’ll start with cloud as a sentinel Nick and natively collect from hey I wanted to say any I’m probably missing a few a marketing service whether it’s part of a sure where it’s one of our offices five services or whether it’s one of our a windows services the reason is because a we are part of my of management fabric Azure monitor

doctor sending is a solution on top of Azure monitor and the markers of mine dates all service owners within Microsoft to send information to us so even if there is no connector a dowser sending the connector page which says specific service that you’re looking for blue chances that if you go into the instructions for this specific service you’ll find how to send data to us this provides a very good coverage of a cloud Microsoft cloud estate in addition we are not any Microsoft only seen we are there to provide you with solution to a monitor and protect any other IT system and that would extend to additional clouds we already have support for AWS our current support the built-in support is for cloud trails we have developed a custom connector in which we can work with you to implement for cloud watch and we’re working to extend further support for AWS telemetry an interesting thing is that since AWS is a cloud service connectivity is also easy we’ll get to all term long term will require more but our collection from AWS is a cloud cloud integration which means serverless you don’t have to worry about the connector machine or scale or anything like that you setup the connection there is a bit more to do because there is a security authentication to go through it’s all outlined in the AWS paid connector page on National Center and within minutes you’ll be connected and you’ll get your territory from AWS so I don’t want to spend a lot on the easy stuff it’s a deep dive architectural talk you can go to documentation and probably get all those connected very easily so let’s move on and talk about a cloud architecture is so unfair collection the first thing you want to collect would be telemetry from systems and when I say systems actually it’s true for on-prem based systems but it’s also true for VMS in the cloud so is infrastructure as a service the best way to collect Emma tree from a systems with Azure sentinel would be using our agent our agent is it can be deployed on windows machines and Linux machines and collect the system telemetry for example windows events on windows or syslog on linux it should also enable you to collect additional streams from those servers or endpoints such as windows firewalls dns logs from the main controller or DHCP logs in general especially when you talk about cloud readiness using the agent is the recommended way because it can be well embedded into your the hopes cycle so say that you have a deployment process and it’s true by the way for either or an apple yes and you’re spinning in an auto scaling environment VMs or actually it’s true a for a containers as well you don’t want to worry about how those system will deliver telemetry for sentinel you want to make sure that the the image itself out of the blocks is deployed with right agent to communicate and send events that’s why in the cloud the usual tautology or best practices is to use an agent in order to send telemetry to either sent you know I think it’s also a this level of quality readiness for example using arm templates in order to deploy the agent on Azure is one of the things that makes sending them very much cloud ready to monitor your cloud a workloads but I did say that we are talking about long term and the agent works just the same also for on parent VMs you may notice that there is an optional collector proxy if your VMs on the clouds are on Prem do not have direct connectivity to the cloud you could use a collector proxy to centralize the location in which data gets out of the organization if you don’t want to do remote collection or you want to collect from sources that you cannot install an agent on this would be a lot of security systems that we do see slog a or common event format which is a specialized type of syslog very common for sending events to sim or that you have a networking agreement and it sends a events using syslog or that you’re using Linux

machines but do not want or can install an agent of them you can use a syslog connector we’ll get into the internals of the Cisco connector somewhat later in this discussion an interesting deployment option is to install this system connector box in the cloud it’s very valuable if you’re connecting say for a branch office where there might not be a location for a VM or if anyways the branch office is not located close to where you your main datacenter is and then sending the through your own premise center does not make any sense the one thing you have to keep in mind is this is it goes over the cloud you will have to use encrypted syslog syslog over TLS some source devices would support it the other source devices may not support it in this case deploying the collector on the cloud might still be an option using something like our Express route which is a if you have some direct in connectivity to Microsoft Azure so we discussed remote collection for a for syslog and for security and networking systems you may also want to use remote collection for Windows to collect remotely events for Windows systems if you do not want to install the agent on those systems you can use with we can again I’ll show the internals and how you implement that a bit later on you can use a web connector via the ref connector VM is essentially Windows Event collector server it uses the standard window is an informing mechanism and it knows how to forward the events to send we appreciate that many people are accustomed to other types of connectors so we are open and we are not supporting just a the max of connector infrastructure if you prefer logstash we support that we have a plug-in for locks that can send information to other sense you know it it does provide I want to say more flexibility in our connector but certainly more people know how to customise locks – in order to drive things such as for example office gate fields on the fly between the source and 70 now if you already have collection infrastructure that would support the to be based on log stash you can use log stash and just added additional destination and the events will be available with other sentinel the same capabilities also available for flow in d an alternative a log processing in a system my experience out there is that unity is not as widely as log stash has identical it’s like with our support forum CFCs log log star flew in the etc we support a large number of sources we do use it using the traditional way using syslog and as my need for pretty elaborate targets crew slide suggests there is a cost to that the complexity cost a cloud native scene would have loved to get away without having a any on-prem collectors we don’t leave that possible today one of the main reasons is that it’s hard to be this locus the service given that it would require authentication encryption many source devices do not support it in the partial solution system in general is a protocol that does not support authentication hence we would need something in between we do have as vision a goal to work with partners so that they send telemetry directly to sentinel center using our API as you can see the big names have already support for that so for several sources and you know noted lightweight ones you can already use direct connection to Sentinel you’ll go to those product UI you’ll copy two numbers from our and from our portal our workspace key we’ll talk about it a bit later or suppose idea then our secret key and data will flow in we believe is a vision that that’s how connectivity to a clone and if seam should be it’s a long journey it will take time until much of what you need today would use that method lastly you always need custom connectors and we have many ways in which for you to build custom connectors we’ll talk about those a bit later one important thing I want to point out we have the list of connectors in the

Azure sentinel connectors page is limited in length it’s not comprehensive and we have in our community a series of blog posts which are linking and rebuilding from the PowerPoint you can download later on which will guide your additional connection connectivity options the first one is obvious we support generics F and C it’s not going to talk about the difference a bit more later on so any device which can sense these organs and system the common event format overseas long we can support and this specific blog article lists those sources and we traditionally updated as I also mentioned Microsoft services and apps are mandated we send information to us we have not in all cases edited connector on our side you can still usually go to this source service side and define us as a destination the second article here would list all those services and apps sometimes it also extend to some services that you know that for example on Prem sequel server does not send to something but just write events to do Windows Event log so or normal agent can read the events which gets me to the third bullet the agent as a series of capabilities I described a bit of them but there’s more there so the agent collection capabilities are described in a third article and finally if you want to write custom connectors the fourth blog post would help you do that one of the reasons we need those blog posts certainly the last free that’s the first one is because our Azure Sentinel is part of the larger a ecosystem and benefits from a lot of the underlying or a integrated services and they’re from the documentation maybe spread we probably need to better around that but the free blog posts here have a lot of things to the relevant documentation which might not be part of our documentation set of as of today so I discard the Logan Leafs agent I describe the collector I describe a log stash in custom collection let’s try to go for each one of the those and understand a bit more about that the Localytics agent as I mentioned you can install it on a Windows server or on a Linux system actually a Windows Server any Windows system can be a client as well it can be used for more than just what it you know it can collect anything which is Windows events and a lot of things in the windows ecosystem are windows events so for example correcting Active Directory events would be done by placing the agent on your Active Directory you want to click system events that’s also Windows events we also support collecting them and I did mention that sequel server also logs through that in addition to system events windows events are syslog which is the core collection for those from Windows or Linux respectively we also support few other sources DNS events Windows Firewall events is events the agent can collect any file data so it can monitor a folder for new files and get those and send them as events to set you know and behind the scenes internally it’s a based on fluency so it supports fluent deep plugins as I mentioned before through the is a very capable technology but it’s not well known so we find it most customers user agent a season do not use too many plugins though that capability is there another note on longer links agent we have used Microsoft different names for that over the years if you know the term or MS agent or Mme it’s the same agent moreover you need just one agent for all the different max of services that work off this agent it’s not a new agent it also is GA for a while which is more than longer than Sentinel so it’s stable and put ready if you already have it on your systems then collecting to Sentinel is just turning on that on the sentinel side deployment I mean different Linux it’s very easy to add it to your automating install in a sharor for moore’s Azure VMs it was there by default they or at least will be available to you through the virtual machines interface so when you press connect you essentially make sure that the agent is available there

it provides central management so once you install the agent you can install many aspects of it from a central location and any measure already provides proxy support so if your virtual machines are not connected directly to the internet a proxy will be able to be there between them and internet it also support a central management it’s not just for sending events moving beyond the agent let’s talk about safe and seasonal collection which is a core capability that’s used for a lot of it’s it’s sort of the basics of security information event management or sim let’s look at what is our collector our collector is a VM or a host that you can deploy I’ll show this one is an on-prem option and essentially inside it has two services or two dimensions we have one of them would be a standard system server they want that you’ll get when you install a Linux the other would be the same Logan links agent which I described just a minute ago together on a box this would be the collector VM it enables the delonix agent enables is the piece that sends events upwards to the sense in the workspace in the cloud of course it so it’s not by night this is done over HTTP it’s encrypted and the system Devon would be the element a collective or listening to the information from the syslog sources together on the single box they are they are collector VM in between them they communicate using a syslog as well over TCP I mentioned you the two ports because that’s the difference between itself and syslog if the system demon sends data or four to five two to four to the agent it is considered syslog information if over to five two to six it’s considered self information the same collector as I mentioned can be also deployed in in the cloud when it works the technology is exactly the same the difference is that now you send syslog over the internet rather than TLS over into that rather than restful api retailers over the Internet and it’s very good when you don’t have footprint from a VM on your own perm location but this it does require making sure that the lag between the syslog source and the collector VM which is system is well protected might not be trivial how do you deploy this collector so what you’d do to do that and these charges are already the documentation in way more detail but you deploy a Linux game of your choice if on Azure because you want to do remote collection or because you collect information from Azure resources from as Raveena it can be any algebra Linux VM if on Prem whatever Linux VM you like there is a set of requirements set up in the documentation but it covers most Linux flavors as well as I mentioned the two popular syslog demons whilst you do it all you need to do is run a deployment script it’s linked from the collector patron sense you know what the deployment script does it’s installed the other one at the long losing agent it configures it and configure your system demon so even though the system demon was there as part of your installation it’s very much part of the area that we configure for you also later on I will not get into details today through the UI for Sentinel you can configure this collector VM and you configure both things which are within the agent and both things which are within the daemon they work as a single system to make sure that you get your safe for system units an important element to stress is that a as the dagger mentioned for you can use a single safe collector machine to use any number of sources so we will talk about scaling and size in the next slide but it’s a starting point a if you have multiple machines you don’t need to create a VM to collect from each one of those even if you have multiple log

types a cisco firewalls and Palo Alto firewalls you don’t need a separate collector for it you can use a single collector machine to collect from both for any source up to capacity to be discussed in us in a minute you can also use two sorry more direct advanced a configuration that will be outside of the scope of the deployment script and the central management by manual modifying configuration file mostly for the system server and that will enable you to do three additional things probably more the abilities of the system are pretty significant but free very managing first of all if you do want to use the same collector to collect those syslog SF events so sources that send self and sort of send system you’ll need to do some configuration changes on the collector box to enable that you can also filter events that might be a very important to keep really you don’t want everything to go into Sentinel mostly by using the filtering capabilities of the system server or assist of the orbs is no Kenji you can filter events and make sure that you don’t get too much events you don’t need and lastly you’ll have to go into a direct into direct a demo configuration to use TLS by following the guidelines for supporting TLS with the relevant syslog server for you our Cisco course is no Kenji the deployment procedure outlined before was about a deploying a safe collector syslog collector as I mentioned above is nearly the same the procedure outlined in our documentation is a bit more cumbersome we did not create the same deployment script that we have for set also for syslog but you can and I probably need to talk about the details of that are beyond a webinar use the same procedure for self and then modify the self connector to do see snug should be a small chain scaling sizes so our sets or seasonal collector is pretty capable we scale to as much as 10 ke PS using a very cheap hardware actually the one we tested these are specific they are VM which cost around 20 bucks per month at least in a u.s. data center the reason we can do that is because the processing itself is not done on the collector box it’s done in the cloud so the collector box is essentially just a proxy just receive data in syslog and make sure that transfers that using restful api to the cloud as i mentioned you can use the same collector for a lot of sources in 10k eps would give you a lot of flack for that one question is often asked is how do i cluster clustering would enable you to have more than take 10k eps so you know more bandwidth if you know if you need that it also will provide with a level of failover there are a few ways to do that all of them are not based on Microsoft Technology I am now suggesting things that you do using standard open source technology a simple way to do clustering will be starting distribution deciding that a certain time of events go to a specific a collector that’s of course easy but it doesn’t provide all the requirements you may have one one option that you have which I feel might be the best today is this.h a proxy it’s a very common open source a little bouncer that exists today on your Linux distribution it’s simple it’s free it provides both a clustering so it will enable you to use multiple a collectors without caring about study distribution it also enables failover so you can have 288 proxies showing the same ip address and if one fails data the other ones captured IP address and start receiving the a the event data one thing to note about HM proxy is that it supports on TCP proxies not UDP processing so you need to make sure that all your syslog sources are capable of setting system an alternate relationship proximately commercial or bouncer if you everyone its it has a cost it do roughly the same

I discussed self and syslog in length and I sometimes differentiated I sometimes just add self or say stock it’s worth noting the difference and what each one is it’s sort of important for the discussion system is the most common transport for sending for managing events originally the links were later on in networking and security systems safe common event format is a structure over syslog so it essentially standardized the payload which is commonly used in the security world it was developed by ArcSight around 20 years ago and as you can see below it provides a key value format that’s used with the safe data the main distinction is that self is by definition pre parsed it has a very structured way of presenting name fields and values when we get information yourself it is stored internally in Sentinel in that in the safe table and it’s stored using field names so it’s indexed as it gets it when we get system data on the other hand we don’t know what the star structure here and will store it with minimal parsing the few headers available with system will be parsed but the main area here between I guess fake user and then the divine would be stored as the system message it doesn’t mean that it’s just that senton allows you to parse on the fly as data is queried so the fact that it was not personal interest is not a blocker okay but if it means that any parser is required if the data is already available in the known format it’s always been is safe this is an example and since we are not getting into parcels today I’ll just presented sort of phases of how you can parse a known syslog stream into fields and make it into a a well understood and used a data not accustomed to query time parsing which is sort of a newer part I’m within the same world think about is creating views you create a function the function behaves as a view or a virtual table and when you query when you search next time use the view rather than the original table and with it you get all the parts fields as if they were real fields I did mention before that we collect remotely noticed syslog or CEF but we do roughly the same also for news event forwarding this feature is in private previous you need to apply in order to use it the reason is that this collector does not have the performance requirements that we are targeting unlike the safe connector which we feel is capable of providing the value you need so when will add you to the private preview we will discuss with you your a a speed requirements the structure is roughly the same it’s a bit similar we have the long release agent the Windows version installed on a windows VM this server is services windows event forwarding collector or web and the local newsagent knows how to forward the forward events to Azure Sentinel Logstash – we discussed a lot about the agent either by the local newsagent either as a direct agent installed on a server or an endpoint or when combined with a quick or syslog server to be a connector we are fully aware that you might be even more comfortable with logstash which is a very common way to connect events and therefore we allow you to use log stash as the collector as you can see if you are into log stash on the right side we have an outfit plug-in that can talk with other logon leagues or hunters and you know and can send the data to us so if you today have a log stash infrastructure you can add sending

events to Sentinel very very easily one advantage is it has sometimes your only choice today would be more input sources if you want to write to read files or databases or rescues such as Kafka using log stash would be the preferable way today also if you want to event processing it can be done with the agent using fluidy capabilities more people know how to use filters and rocks etc with these logs – in order to process events example will be some pre parsing or office getting a field you don’t want to get to the cloud without being hidden before lastly custom connectors you can’t by the way create a custom connector using Logstash I described it event transformation capabilities as an option here, there are other options custom connectors are used to get events from sources that we don’t support in any other way but they’re also very common yet in order to get enrichment data you want to get GIP information or customer intelligence I described last week our built-in or a supported for intelligence such as a miss four sticks and taxi but if you have your custom intelligence you want to build the connector to get it in a or asset information all those can be connected and connected using a custom connector how do you go about creating custom connector partial would be all the materials they use the API so behind the scenes they’re all API we have a partial a partial model that no that knows how to send information to Azure Sentinel it’s from your easiest way to automate or to one-time send in a file into Azure Sentinel keep in mind that since Azure Sentinel supports late parsing query time it doesn’t really matter what the format is will will support it if it sees V or JSON or XML and the parsing is built-in anything else we saw earlier on how to create a parser you can also use logic apps which is as you may recall if you’ve been to the last week webinar or automation engine when we use it as an automation engine we trigger it off an alert when you use it it to import information there’s a custom connector it may be either scheduled to run every few every few minutes or hours or whatever makes sense for the source or triggered off an HTTP request so if the source knows how to trigger sending events it can trigger the logic of automation it has the capability to read files or databases or API so it’s an alternative to Lokesh to read all those and while logic apps is a cloud creature it has an on-premise create way so you can use it to create a custom connector to collect data from long time as well and of course you can use dark api used to create your custom connector if the need arises if you’re free comfortable who’s that well if you keep in mind that direct API use and programming doesn’t mean you need the server now all those are supported Roger function so it can use server list capabilities so summarizing our a on-prem collection capabilities and I actually go to sort of this summary slide again No how do I get my slides nevermind I won’t go back I’ll summarize over this slide we discussed that the after discussing cloud collection most imaging is very easy go to the documentation I discussed in live or is and on Prem collection capabilities I mentioned installing agents on servers which with the cloud becomes more the preferred way to do things because it ensures that your servers are monitored well also when there are Auto scale you know we’re no more temporal but we also have remote collection capabilities such as using a seasonal collector which is based off the agent and a syslog server or web form or Windows Event collection as an alternative I suggested log stash which we support and which will enable you to use technology you may know better to do additional work or use

existing collection capabilities lastly I discussed custom connectors which is if nothing works otherwise your way to make sure that you get the day gets in lastly is to summarize keep in mind that not having a building connector is rarely a blocker because you don’t have to pre parse the data which in traditional seem world is a major blocker trust they get in and change to support it would be changing and on the fly crate and parser which can be done after the fact once it is already with that said I’ll move from collection focusing on from collection into our cloud architecture it’s a very different topic it’s sort of moving into another world it has some touch points but not a lot if you’re a sim guy you might need an introduction to a bathroom to understand the rest I didn’t share my background I am a same guy I’ve spent many years the dark side starting as a researcher then a product manager a focusing on products and then I was across director for amia working all these customers in Europe making sure that hopefully they’re successful in the nursing so I know a lot about see a syslog I know you know the beats since joining banks of last year I had to go through a journey learn learning ah sure because it’s cloud world and you need to understand the cloud in order to not just run sense you know in order to monitor the cloud to start with you need to learn the cloud I’ll give you an example sort of not directly related here but if you’re into same you know that the asset model is a very important part of sin and it’s about IP addresses and host names and subnets and it’s all your relevant to the cloud in the world where computing is either serverless or temporal the way you define your assets and prove them is very different while it be different between the cloud vendors resource groups will be more important than minutes so keep that in mind you need to understand Alger not just transcendence but in order to protect you and you need to understand another class in order to protect them when you try to do that for you as well so let’s start and again it’s not a comprehensive in Microsoft Azure a introduction it’s here to serve to describe how we manage a cloud state we said the first concept to understand would be the max of tenant a tenant is an authentication and any management domain it is actually a same as already a instance usually a tenant would correspond to an email suffix so everybody at have the same suffix email address because we long to same terror it doesn’t mean that all Microsoft the resources are belong to the same tenant it means that all the mailboxes so that office 365 is within the same tenant with the tenant which is you know at any management space we have something we refer to as subscriptions some of the solutions will represent SAS solutions office 365 would be a subscription within a tenant dynamics Microsoft CRM solution would be another one and in tune well not a solution also fits in within this model those will be usually one to wonder is one obvious we see five subscription within a tenant on the azure cloud platform you can have a lot of subscriptions subscriptions are the paid the payment level for a forager to have a subscription you need either in the present agreement with Microsoft or a credit card or a free subscription there are few options available around that so that’s the level that has a paying owner subscriptions include resources as you see in the bottom resources are the things you manage I just mentioned them when I talked about an asset model resources might be VMs that the simple resources demanding network elements in your fee net there might be past services whether around IT infrastructure so a firewall or you know

a pass database etc all those are resources resources are managed in resource groups which are an interim level between a subscription and resources there are softer than a subscription because they are not touched to money money is always a very hard factor but they enable grouping resources and managing them together in one room as folders of resources one area where resource groups are very useful is in a permission management permission he’s inherited in this model you assign permissions subscription level you can modify them on the resource tab group level and again on the resource level very much as you would in Windows files in step for example I am ignoring through now the managing group level it’s another level of abstraction to manage multiple subscriptions together it’s not heavily used sentinel and the discussion is not a generic other discussion we want to discuss Azure Sentinel after sentinel is a resource it’s a resource within said using Azure you can figure it off as a database or a we call it a work space where everything to do with a single sentinel deployment is consolidated but it’s managed in terms of either as a resource so it’s part of a resource group it has permissions its entirety part of a subscription which is whoever pays for it one important aspect will have to discuss when we talk about a Azure and about other Azure Sentinel will be regions and Geo’s, Azure is deployed in data centers those data centers are mostly transparent to all of you they are grouped into regions which are sets of one or a bit more one to three as I recall data centers that are very close to each other and that provide redundancy internally the map you see here is a map of other regions you can deploy essentially in any other region that supports in Logan Linux there are exceptions I probably should not have had this slide without listing the exceptions specifically because it tends to be a bit misleading we are still not available in what we call a national clouds so China or use federal one and there are a few other ones which are still missing naturally if longley is not supported their long leases if you’ve been to my last term in our or the infrastructure layer but we have with a few exceptions we support very long leases supported we at this point in time do not commit to the data staying just at the region we’re working on a plan to have better support for that but at this point in time we can we just it EU data doesn’t leave the EU and that American than a u.s. date I should say it does not leave the u.s. we do not data from the rest of the world Israel through the u.s. we understand that that’s limitation that in many countries you’ll want to have the data not living your specific country and we are working a plan to provide it but we are not there yet also the you if you select any a workspace in any a region data will still be shared with Westar which is physically Amsterdam also while I said data as you can see in the slide this is data at rest data in motion is a much bigger challenge especially since we collect data we found that it’s pretty hard to ensure that if data flows between say a UK West if you sold there and a EU West which is Amsterdam to know exactly what are the route it falls routes where their motion will ever leave the e or not it’s something we work to investigate it’s a general a other question which are we are looking into also keep in mind that we have a lot of collectors those are collecting for external services and that makes the question of word that goes before it gets who said you know even more complex so our commitments as I described it more for data at rest as I mentioned in the u.s. it stays in

the u.s. in the EU it says edu but may not stay in your country in the rest of the world today goes to the US so that was another introduction with a bit of discussion of sending around genes I mentioned that the main as your Sentinel entity or resource with your workspace the workspace is a container it includes most of what you think of as Azure Sentinel keep in mind as I mentioned before today we don’t it’s not a physical disk so it may not imply that it’s only one place it’s the resource as I said in Europe it’s part of it will be in Amsterdam even if you place it otherwise but what’s belong to that what’s part of that event database rules incidence a lot of data some other elements that we use as part of sentinel are standalone other entities or resources and can be managed separately as far as far as a permissions for example or deleting etc two examples of those will be playable switch used for information and workbooks which we use for – morning as I mentioned our workspace is the same workspace that you may have been using these long analytics in practice certainly the solution over long analytics so any sending long space is basically the log analytics workspace with the security inside solution implemented it does mean that whenever something is possible for all analytics we are upward compatible it’s possible for us you’re sensitive it’s important you summon someone sometimes you’ll go through documentation you see that it can be done with some analytics you say but what about Sentinel let’s crunch I measure the workspace and what we do find out and a core of the rest of the discussion they would be around that is that customers as many times cannot have a single workspace for reason that we’ll discuss in a moment the rest of the discussion will be around how why you will need multiple workspaces how many you need and how to manage multiple workspace is a single Sentinel entity why multiple workspaces so if you have multiple other tenants you will probably need a workspace for sentinel NH the reason is that the a our monitoring framework is very tenant centric so most sources would know how to send information only within the tenant there is a good reason for that tenant is you know education domain and cross turn authentication maybe charging and open to exploit so that’s that’s nearly mandates a creating multiple workspaces another Asian reason would be if you operate in multiple regions there are tourism that may want you to separate workspaces if you have multiple regions one is compliance you work across the ocean you want to have the data in the you thing the you in the u.s. staying in the US and therefore you have to workspaces one created in unit in an EU data center and data in a US data center it also reduces call a literally cost and latency because if your sources whether on tram or in in Azure data center or even in AWS data center are closer to some place it would imply that you have lower latency in pain less than two in cost if you have a natural setting the workspace closed by a last reason which is sour between strong and soft would be subsidiaries we have solutions and we’ll talk about later to have data for multiple subsidiaries in the single workspace sometimes an organization would prefer to have workspaces or pairs of scenery there are other reasons that would drive you to have multiple workspaces but we think that they are avoidable and this move or face may need to be consolidated one is separate billing a workspace is part of a subscription solution is the payment unit which sometimes these people to want to separate workspaces into the free subscription we do not think this is best practice it makes too many workspaces which even given the tools I’ll present later on to manage those might become just too much next there is fine-grained retention sending

and fine grained access control in both cases it’s sort of a legacy reason until probably sometime last year retention setting and access control were sort of pair workspace and therefore you would separate between workspaces to provide access to different people or set different retention settings we’ve reduced a number of capabilities around retention setting and access from within a workspace that made those reasons much less important in legacy architecture through Logan Linux which is this for a while the guidelines initially were more workspaces and with the introduction of fine-grained control the tendency is to try to conserve as workspaces so you just may come to an organization where or your organization may have a large number of orc spaces or return for different reasons so as a rule of thumb sort of how many and how much I would urge you to use one workspace for each tenant you have to monitor for each other region and for each subsidiary it’s still considerably might be a large matrix but not as large as we sometimes see today where every subscription is a workspace also if you don’t have compliance reasons if you know you may want to reduce further this would imply depending on the size of your organization and I’m going to succeed how complex is the organization anything between well one in the best case if you in a reasonably mentioned but it still can get to fifteen or twenty workspaces easily so the next step would be how to implement sensing across workspaces first step would be to conserve workspaces very getting rid of the legacy reasons next I will mention our through lighthouse which will have to implement to provide cross tenant access it’s not that if almost bases are witnessing your dinner then we’ll discuss how to work across workspace with Sentinel talking about queries workbooks configuration in a access control concerning workspaces so the important thing to keep in mind is that given how monitoring works just moving things to Center workspaces is usually cost less and easy it’s just a procedure in the organization so taking the pipe from one pool and putting it in another pool would not cost you any different and will not affect your operations in any meaningful way so it just has to be done a special case I want to mention would be a from security center historically out rescue center is many times used with default workspaces which I created fair subscription it’s not needed in any way it’s not important and you can use the data collection screen to use to to point all rescue center work a match servers to send the information to specifically selected workspace as your lighthouse archivist is a new technology that Microsoft introduced this summer and it enables cross tenant a access behind the scene it’s a it’s a user management scheme but we support it and we allow you using one to implement lighthouse and you see on the right if you know Roger that and that’s myself that I have now access to two tenants at the same time it means that every do do we search the lock screen in Sentinel and work groups and hunting can be done across workspaces as a tip the query has to take into consideration the multiple workspaces and easy way around that I mentioned before a view the cost of a view for parsing which we call functions in Sentinel those can be used also for multi workspaces so you can create a function that will represent any schema that uses Sentinel but built-in would do it across all your work spaces across tenants so any search any work group so dashboard or hunting activity can be done across workspaces within the same tenant or across tenants I did mention a cross or crosswalk Facebooks so all your a monitoring a-fib functions

can be a extend across multiple workspaces as I mentioned as of today it probably will require the our alpha box dashboards work with the current workspace and you have to go and modify them and this would be to create a function that will place any schema with your extended cross workspace source and then go to each one of the workbooks and change the queries to adjust some things that you may want to look into is cross a workspace a alert monitoring and connector status things that we are starting to provide in a single one space and have to be extended across workspaces one thing I do want to mention a and I’ll hint that the roadmap even though it’s a public discussion there’s one very important a feature within a sentinel which is not a workbook it stays in screen it’s the central screen that a analyst uses it can be replaced by a workbook but it has finality would want to use we are working to have a version of the incident screen that works across workspaces and tenants another challenge that multiple workspaces will will present would be cross a workspace a management we are in the cloud world we are very much API driven if you know either you know that arm is our way to provide DevOps abilities to scale so our templates our way to create any resource in Azure between API is an arm templates we provide you all the tools to automate a to replicate a content configuration across workspaces so get it from one a workspace the master one will you work and replicate to all the others illness for example for alerts rules and hunting queries and big things to one of our partners mortal that created a sea Sentinel which is a partial model that does that for you without you having to understand the API you you want to if you have played books and workbooks that are specific to a into a workspace you will be able to deploy them using arm and it carries also to things more granular such as permissions sum it up we don’t provide today a UI based option to manage across workspaces but we do provide all the ways to automate or the replication of coordination across workspaces which we find in the cloud world is usually the norm as it has to be part of a DevOps cycle lastly an interesting a an important aspect one of the reasons for multiple workspaces as I mentioned in the past was a access control because he wanted to the original resource owners to have access to their data it’s most notably important in the cloud world you have people that managing a Windows server farm and you take the data from them but they want to also have access to their Windows server logs it’s an alternative to using multiple workspaces we have been introduced data always ask on control and it has two variants that might be useful to you I start with the second one it’s more powerful but it has its limits resource centric RBAC resources are RBAC able you to allow access to specific IT owners just to the data collected for their own resources they don’t have access to your seem to 17o but rather through the resource search screen itself in the resource screen they have a logs tab in the logs tab they can search through their data that care is also also for resource group it doesn’t have to be resource or resources into a subscription so if they own the resource group they can search through all the data that we know that resource group it’s a very powerful feature until recently it was available only for our resources last week at Ignite we introduced max we introduced a technological Azure Arc

Azure Arc enables you to extend these management features into on-prem or higher servers in other clouds now any server can be assigned as a sentence or as a natural resource and therefore be managed from Azure as well as have access control to the data to the resource owners so now you have VMs on trim the same capability applies to them you can provide taxes to their owners to their local data without providing access to any other data or to the same itself the second option table Westerberg enables you to control access to specific types of data within sentinel a good use case would be to limit access to office activity logs office activities tend to be more sensitive in the firewall logs and you may want to say that these data you want only a fraction of the users to have access to trying to summarize the discussion around a Azure Sentinel cloud architecture Sentinel is a workspace the basic entity resource which forms that Hill is the workspace its leaves within a region in Azure and while you would want to reduce the number of walk spaces because it’s just always harder to manage more and more resources there are factors that would drive you to have multiple workspaces even in your organization those would be geographies subsidiaries or tenants you may find yourself in a situation where because you have some a few tenants a few geographies and a few subsidiaries you get to substantial number of Azure Sentinel workspaces whether 5 or 15 no training would be very specific to your organization what I provided you today was we first the best practices what – you should split workspaces because of and what you should probably avoid the news are more advanced capabilities to do using a single workspace as well as a set of steps in order to manage a large set of workspaces or a set of workspaces as a single Azure Sentinel entity I focused on a using lighthouse to enable crosswalk space a query search workbooks and hunting a I mentioned that we will develop also a crosswalks is incident screen which is not available today and then I mentioned how to centrally manage the contents for those workspaces and lastly how to manage a data access in order to provide with fear workspaces still a access to the data for resource orders before I before I move to questions I want to remind you again that’s a starting point actually it’s the second iteration we discussed the basic functionality of sent in last week and there’s more that we’ll share in upcoming webinars there’s a time until that next webinar it’s after the holidays meanwhile we do urge you to join the community week to answer it’s a very active community also follow our tech blogs we publish multiple times a week a lot of new data it’s many people who do that we have a growing community keeping the other links on this slide I talked about community you’ll see that there’s well I’ve been working with scene for a while the rate in which we’re building our community and Ryan thanks for your team for making that happen it’s very fast and you see meetups and blogs by others all the time and we’d love to support you so blogging is for you well but we’d love that if you haven’t made up if you want to people to hear about sending L don’t is it to contact us and we’ll find the right way to support you on that thank you all Ryan do we have questions. Ryan: yes, so I want to remind people here now is the chance to submit additional questions that we’ve been answering him in the I IM window all along any of the ones submitted from now on we’ll be reading out so if you have a question that’s a maybe more easily explained by a quick demo than by something we can type in a response now is a great time for that question and we can get out I need to go to the demo mode that’s where is my browser you put me on the spot for that anyways all right

so can one of the questions we had come in is how does as your sentinel monitor SAS and path services so a great question first roar demo because you asked for that so III did discussing then of how we collect from on tram because there’s a lot of subsidies there a lot of details in a way I sort of jumped through Sasson pass services because it’s easy but it’s worthwhile so for example let’s take Azure Active Directory it’s somewhere between passes us it’s very fundamental but it’s a good starting point I don’t have permissions but you see that all you need to do here is just to connect and that’s all it’s because we have the connector is behind the scenes we are in charge of everything there you have to do nothing let’s move to something a bit more complex well promise will be the wrong way let’s look into Microsoft 12 I tell you that I’m pretty fond of this one because it uses a rule set that I started developing as an open source project 15 years ago I’m not that young but let’s go to the connector page you see that this connector page is actually very typical of what you can do with most a path services Microsoft you do nothing here what you do here is you open the application get a resource and you tell it to send data to us that’s the instruction here there’s nothing you actually do here the only thing we give you here is monitoring on the left side whether we actually go to events from there or not so that will be a second one the last one I’ll show you talking about paths and SAS will be Amazon Web Services that’s not within Microsoft anymore it’s a different a cloud service but it’s not more complicated there is not a lot more the main thing here would be that you need to take our values interesting enough my AWS best practices would for Microsoft to open an account because you know we read the data for you but you have to also provide your role the one you created there and together with those three details we will collect the data for you so yes it’s a bit of a magic it’s much simpler that’s why I didn’t go into details here and it’s the same for you let us know if it’s different for you great thank you we’ve got a question here for Windows machines that are not managed by a d– how would someone deploy an agent on those prerequisite or do we have a solution for that so is told there’s no requirement for being ADA connected to install the agent there is always when something is not ADA managed there’s always the question is how to manage it software life cycle in the first place how do you install an anti-virus update for that matter so the agent itself would work very easily it doesn’t require any ad connection on any any a Microsoft Windows machine how you actually go on installing would be depending on your alternative first officer distribution even you’re not using GPO okay got it what guidance do you have for people trying to use the API do we have any public documentation on that or should people just be using the RM API how do they access that so API is a large word I mentioned API in many different aspects in in this presentation I think that the two main ones would be in a API as part of custom collection for data ingest and the second one would be API for remote for central management the ingestion API is fully documented and heavily used they a cook the configuration management API some of it is still a in public preview so might not be as documented I would like it to be for all there is a API definition files but sometimes on the level of documentation you would want between API and arm usually as your Sentinel services are controlled by API things which we use from the larger as a record system such as more groups or play books are managed using arm usually it also implies better documentation they are less new than sending itself got it thank you we’ve got a question here can we also include Intune related logs in Sentinel Anderson yes it’s a good opportunity to go into our

community blog and mentioned a blog post that I listed somewhere there and by the way talking about vlog you’ll see that we published on the 20th or the 19 from the 18th we skip the weekend run so it’s a very active blog that’s why I’m starting to show thank you all blocked contributors if I take the late the connectors a label and look just into that I did mention that we have a blog post about collecting logs from Microsoft services because many of those know how to send to us and you need the instructions on where to know how to do that in this page you’ll find things such as well ain’t you down there and that will lead you to the page in the Engine documentation tells you how to send to us information and you see in tune is not a natural service but still all of our services that you would know otherwise out to connect us as a key vault front door which is pretty new etc ok great another question here is that best practice to have different syslog and set Linux VMs on Prem instead of combined for example we have many Cisco devices sending in syslog and many Palo Alto firewalls sending incest so it’s a bit I put this way okay today it’d be easier for you to maintain TV ends because you have to manually change configuration in the collector in order to support both on the same box it’s doable it’s supported but it adds a minimal level complexity I for me it’s easy so it’s not fair therefore I’ll have to pick go is one I’ll tell you how to if your organization is very structured and very cards just what’s on the documentation goes to okay great what are the activities and data types collected by Azure Sentinel once I connect a source to Sentinel any specific keyword that I should query in the log search so I’ll try first of all even though it’s not a part it’s important to know that Sentinel does support search so you can actually use the word search in order to do a cross source Crossfield freeform search it’s not every you asked for but it’s worth knowing that you don’t need the schema to start you have an IP address you want to look for a username you want to look for you can do a search and find them across sources secondly the question is good and the question is about a document in the schema or the events that you may get we are working on that there is a link which I don’t remember by heart it’s on the github where we collect all those as usual there’s a lot that we get the events from sources so a lot of the information about what you get is incorporated in the documentation for the sources and we are doing the work of a consoling death and providing it to you I hope they’ll be ready shortly excellent thank you is there any way to hide sensitive information and the lobster events say passwords or keeper values that could catch up so I assume I’ll say two things one of them would be as a seam is you know talking about something another is same guy to do that you’ll have to use a a collector it works it with fluidly using a plugin but there’s less know how about it which locks – we have many customers as we’ve done them so by processing the events before nurse and to Sentinel that actually works only for on tram collection there is a gap there and we don’t have a way to do it for the simpler cloud-based Connect connectors so that’s that the other thing that I do want to stress is that in general it’s very hard to ensure that when you collect logs in large volumes for multiple sources there is no credentials in them if the source does not take care of that assuming that you always know where a password is is hard Ryan: got it Thank You. Can Sentinel run in one of the commercial Azure clouds but

connect to and get log and security data from one of the government clouds Ofer: I don’t know I it’s an interesting prospect it would potentially work if a domain so a tenant can cross between the government cloud and a commercial cloud I’m not sure that’s the case so I would say that for now I would use noise and answer but it’s probably worth looking into it a bit more whether there’s some way around it yeah add to that um you know a lot of the government clouds are very much their own specialty because you know that each one can be so different and it’s hard to speak about them all in general but we will be focusing more attention to resources on the government clouds that speaking to that community so watch our community closely over the next a little bit for some government related discussions and that sort of thing I’ll keep everyone posted another two that I mentioned in the two are not supporting government cloud today we do work on that so US federal cloud to be more specific perfect thank you will lighthouse eventually allow a single sent no instance to perform cross Tennant correlation and alerting so we are looking to Det we understand that’s a requirement that’s needed it’s a missing piece in supporting multiple workspaces I don’t have a timeline for that okay great thanks when see when will the Sentinel logic app trigger work on the rule templates I’m sorry well when will the Sentinel logic AK app trigger work with the rule templates so I assume that the question is and translated today you can trigger off with drool type types you can trigger off scheduled alerts but you can trigger from the other types of alerts I assume that’s the question we are working to ensure that marks of security one the fourth one which is the most important one most missing one today is supported it’s working progress if you know I provide relative information I am NOT allowed to do dates but it’s closer than others great thank you we’re about out of time here folks so I’m gonna give you a few links before we say goodbye let’s move Vic to delink stood that you can talk about him paste him in here also if we did not get to your question somehow are we missed it or you’re listening to this after the fact is it recording please go to a kms as your central community that is a great place to ask a question you’re gonna get all the same people who are answering on this call so that goes directly to our engineering team we always love to hear your feedback on the webinars whether they’re valuable to you and what we can do to improve them so you can give that at if you’re looking for recordings of any of our webinars and a list of upcoming ones you can find that at a kms security webinars you can join our community at and the difference between that and the link I gave earlier about the Sentinel community is that central one goes directly to our Sentinel group the security one at large is all of our security products so you’ll find links to all the security products there if you are interested in getting in on some of the private previews you can sign up for at I want to thank all the folks on our team who were answering questions today in the i”m window we really appreciate your help of course I want to thank over for a fantastic presentation that was very informative most of all I want to thank all of you for joining us and being a part of our community we will see you next time