Everyone. So I will be talking about AI driven identity verification and how we can balance security and privacy. So as digital transactions, they proliferate and increase the risk of security breaches. It is also increasing and traditional identity verification methods, they often fall short in addressing those sophisticated things. And there are also the regulations like GDPR and hipaa, and there are robust compliance measures that we need to adhere to in order to stay compliant for these AI driven identity verifications.
So we will explore the limitations of conventional methods and discuss the pressing need for advanced AI driven solutions to safeguard the identity verification processes. So we will talk about the context of your identity verification, so in the context of your identity identity verification. So we will be using the C back model, the context of our compliance based access control model that will address the complexities of modern security needs.
And it'll interpret the real time con context analysis. So this is actually a model that I'm working on.
And in this model we will be using the context for access control. And we can also use it for the identity verification, like if you are going to have access to some resources for the access control thing, and you can set a context like if someone is accessing some resource during a set time of day, or they are, they are approaching it from some set location. So how we can use that context in order to determine that that resource it has been accessed from the same location by the same person, or his permission should stay same or it should not stay same.
So in our, if the context engine, it is going to map the location and it is going to map the timing and other context.
So it'll actually determine that whether that person, he should be given the access to that resource or the access should not be granted. So we can also apply it to the identity verification. And for identity verification, when we are going to use the facial recognition, so we can actually take different, we can divide the face into the different regions, we can derive different features and we can then generate the binderies from those features.
And then we can use those binderies to create our context. And then we can merge other context like location and device and other things with that, the context that we have generated from that financial, that facial recognition. And then we, we can combine it and we can generate mapping through a decision view engine that that person who is going to have an access or who is going to prove his identity so he is the same person or not.
So the kickback model, it will be consisting of the three components, like the context analysis engine, the context analysis engine.
It'll be gathering and analyzing all the contextual data that we will be creating. And then the compliance verification module and then the, the C fusion engine. So the compliance verification module will ensure that all identity verification processes adhere to the relevant regulations and the decision fusion engine. It'll actually integrate the data from previous components to make the real time decision and it'll be based on the machine learning and all the data that will be coming from the other modules.
So the decision fusion engine will run its own mapping and it'll see that if the person who is going to get an access or who is trying to prove his identity, so that is the same person or not. And it'll actually be mapping the several context that I talked about, like if we are using it like for identity verification, so it it'll decompile all those boundaries and it'll select the features, different features, and then on the basis of those features, it'll generate the, the other contacts and it'll use that context like location and the device and all those things.
So then it'll map all of those things and it'll generate its own decision on the basis of that. So the context analysis engine will leverage the AI together and scrutinize the contextual data and significantly enhancing the accuracy of the identity verification. This engine will evaluate the factors such as the device type, user location, behavioral patterns, and we can also use it the behavioral patterns. We can use like the typing speed and facial recognition and location device, ev, anything.
We can use it as a context by con considering these contextual elements, the engine can detect the anomalies and potential security threats. So in case if we are going to use this kickback for access control, so what it can do is if someone who is trying to access some resource out of some business hour or something like that, or from some location from where he's not normally accessing that resource, so it will automatically, it'll automatically degrade the level of their access.
So if they have the privileged access to some resource, so it'll degrade it to some basic level access or something like that.
And the compliance verification with ai, the, the compliance verification module will dynamically adjust the verification methods to meet the regulatory requirements and it'll continuously check the requirements with the GDPR and hipaa, for instance, in regions where the strict data privacy laws are existing.
So the module will ensure that only the anonymized facial data and the facial feature vectors that we have extracted from various facial regions and we have converted them into the binaries using the local binary multis select feature learning. So it is being used and this AI driven compliance check will not only maintain regulatory adherence, but it'll also adopt to the new regulations and offering a new foolproof solution.
So the, the CN fian for realtime identity verification will play a pivotal role into this method. It'll be integrating the context and the compliance data to make real time decisions and it'll be using the AI and we'll be evaluating the risk and legitimacy of identity verification requests and determining whether to approve, deny, or require the further verification.
And this capability allows the system to respond swiftly. So here in this picture you can say that we have divided the phase into various different regions and from these different regions.
So we are extracting the features and after extracting those features, so we are actually converting them into the binderies using the local bindery feature learning method. And we can also use the multi-select feature learning for this method. And then once those images, those have been extracted, so we can convert it into the low dimensional image, and from that low dimensional image we can convert it into the binderies and from Binderies then we can decompose and we can select the features from that face, facial image or some other biometric image.
And we can also then use it as a context together with the other contextual element. So the data flow will be, the user will request the I and the user request will be initiated and anonymized feature vector is generated along with the context data and it'll, the context analysis engine will analyze the facial feature and facial feature vector and compare it to stored vectors.
Then the compliance verification module will ensure that the data processing is GDPR compliant and we'll check the consent and data minimization and all the requirements that are there.
And then the decision fuel and engine will combine the results from the context analysis and compliance verification to make an access decision. So the benefits of the CAC back model, the imple, the implementation of this back identity verification will bring the numerous benefits. So it'll enhance the security in achieving through this sophisticated context of error analysis. While dynamic compliance checks ensure adherence to evolving regulations and users will enjoy a seamless experience with real time decision making.
So we can use this C back model for both the biometric identification and also on top of the PBA and R back and attribute based access control. We can also use this kickback method and we can set our compliance policies in a way that the context analysis engine is, it is mapping all the context and then it is mapping the context with those compliance policies that the organization has. And even if the organization don't have any compliance policies that have been set in the, for their access control.
So it'll automatically be gathering the data, like the user behavior and the access patterns and all those things. And it'll be making the automated decision through the DC fusion engine. So this is actually something that I'm researching on and I have published my first draft of my paper on based on this idea. So I'm exploring it and making it more comprehensive and robust.
So thank you.
Yes, thank you very much for bringing this through. Thank you. What are some of the, the takeaways that organizations could have? You said you're, you're working on a, a first draft of this, you're working on, on really building up the concept. If an organization was interested in learning more about this on, on pursuing this idea, where could they go?
What, how could they get involved?
Yeah, so they can get involved in a way that my paper is publicly available on SSRN and ResearchGate and also I'm building up prototype around it.
So yeah, and for the organizations it'll be very beneficial because actually right now though Microsoft and other companies, they are providing the solutions. Like you can set your compliance policies and there are things, many things through which you can set your compliance policies and if something go out of compliance. So your access will be, you will not be granted the access, but actually the need for this solution is like there should be some robust solution that can always be mapping things and all the data that is being generated for, especially for the access control.
So there the engine is gathering all that data and it is using it to make the automated decisions because all those compliance policies that are right now there, so they cannot fully comprehend and they cannot fully address all the requirements of the organizations. Like if I am traveling to somewhere and my policy is that I should not be able to access the resource, so I can use many things to access the resources from different locations. So it'll not be using some context or something to generate a decision based on the c fuel and engine.
Yes. So
Yeah.
Thank you very much.
We appreciate it. Thank you.
