0:00 AIML which is that whenever, whenever SASE Gateway is not sure about the risk rating of a particular file or something, then that file is uploaded to Advanced Security Cloud. 0:17 And then it goes through all of these various services such as static file, static analysis and multi AV AIML and multi multi sandbox. 0:28 Since 2017, we have put a lot of investment in AIML and for malware detection based on AIML. 0:39 So now we support a lot more files like Olay PDF. 0:46 So Olay is basically all of the docx, PPTXXLS, JavaScript, PDF, all of them we support now also EXE files. 0:57 And then we use different models for different file types and we are constantly creating our models and for performance and accuracy and we are tweaking them for better efficacy. 1:12 For anomaly detection, we are now using definitely we do anomaly detection for both networking as well as security event. 1:19 And we are we also using in our graph database. 1:23 So graph database helps us to find new anomalies and and do something called as community detection as well as blast radius detection. 1:31 So if John Doe he downloaded a file bad file then who else are affected? 1:36 Who is John interacting with? 1:39 So we can get all of this information because now we have this graph database as well. 1:46 Then here this event correlation, intelligent alerting and troubleshooting again. 1:51 So whether it's for networking as well as security events, you know, there are tons of logs been generated, which log is important, which log the admin should be looking at to make it more easy. 2:05 So again, you know we have made more improvements to our event correlation intelligent alerting and troubleshooting so that the user can zoom into the most important incidences and alert rather than be worried about rather than be lost in the clutter. 2:22 So that is again you know done by our AIML engine for data loss prevention. 2:26 We are again adding our models can do deal DLP for text, PDF, doc images. 2:34 We are using AIML for that. 2:36 Then I talked about models which can detect pictures of structured and unstructured images and we and and then we also have now LLM firewall so that user can safely access Gen. 2:53 AI platforms so that they don't unknowingly exfiltrate some information which is confidential to the company and they don't even download something which is confidential to do some other company. 3:05 And then again for for troubleshooting. 3:08 So the Verbo copilot, which we have again, more improvements we have made so that so the user can be told why his voice quality at a branch is bad. 3:21 And and we have improved in our Versa GPT using our Versa GPT, which now ingests all of our support tickets and documentation. 3:31 And as a result, we can we can give any answer relating to Versa. 3:36 So the the knowledge base and it has been improved. 3:39 And then in terms of performance measurement and capacity planning also we have made a good improvement and we have we have been evaluating GP US from vendors like NVIDIA and DP US from them as well. 3:55 So we plan to incorporate them into a Warsaw Cloud Gateways and also use some other tensor codes and the EMS instructions which are part of Intel CPUs. 4:04 So one last few slides here is that this so until now what happens is when we do something like UEPA at this WhatsApp Cloud Gateway. 4:15 So we have definitely tons of visibility as a networking element. 4:20 And we also have visibility using EIP or HIP which is the endpoint information profile or the host information profile from the from the laptop. 4:32 It sends a lot of information to the WhatsApp Cloud Gateway. 4:35 But now we are adding more, we are adding an agent onto this endpoint. 4:40 So we have a lot more information about the endpoint like similar to what EDRS would have and XDRS would have so that we can much make a much more informed decision. 4:51 So here what happens is that like let's say that this person goes to normally goes to US sites. 4:59 And so the worst analytics will give this information towards VANI, which is baselining this person Alice. 5:07 And now if, if she if, if her laptop, some process within a laptop, a bot which is already maybe it's already breached. 5:15 And if the if the words I am if the if this bot, which in the endpoint, it reaches out to a command and control and say Russia. 5:26 Then again a log is generated and and the UE BA makes a decision based on that. 5:32 But now we are, we want to have a lot more decision which is on the end point itself, which today we do have from the EIP perspective, but today we don't have relating to some of these things like system logs and application logs and security logs. 5:50 What changes are made to files and directories? 5:52 What are the running processes, failed login attempts and malware detection. 5:57 So all of these, the Versa agent will report all of this to the Versa VANI to make a much more proactive and informed decision. 6:06 So that is what is being added as part of the Versa solution. 6:13 And so this is without depending on any kind of EDR, we'll be able to have a lot more information which is what is happening on the Versa on the on the endpoint devices. 6:25 And then like anytime, once we know that our endpoint is acting funky suspicious, then we automatically inform everybody in near real time. 6:37 And they can, they can do all kinds of enforcement like re authenticate the user disconnected from SAS application and re and force them to reconnect, mirror the users traffic and all of that. 6:52 So, so, so this is some of the things which we have done for use of graph databases and new for identification of new anomalies, community detection, blast radius, determining blast radius and again more information from the endpoints for more proactive detection.