Hello everyone. This is Shruti. Today we'll be covering into how AI is transforming security architectures and cloud-based CRM systems. We'll be looking into real world enterprise implementations across multiple industries, and we'll examine the practical intersection of how machine learning capabilities can help robust security frameworks through detailed case studies. From Starbucks, capital One, Mayo Clinic, JP Morgan Chase, we'll uncover both successes. And critical failures that provide valuable lessons for our own implementations. In today's agenda, we'll be covering about the core concepts of machine learning, and we'll be going through case study. One by one, which is Starbucks, followed by Capital One breach, and then how in the healthcare machine learning security is transforming the industry. And last but not least, the JP Morgan Chase multi-cloud machine learning security and how the security is orchestration, how the security is orchestrating across. Multiple platforms. So our session will move from theoretical foundations to practical implementation strategies with each case study. Building upon the lessons of previous examples. In the current security architecture, the challenges that we often face are manual threat detection, which is sometimes awfully long and painful. We take reactive security posters, which is. What we do after the security incident happens, rather than protecting it proactively, we have stat static access controls, which needs maintenance over time. And have to adjust based on the industry requirements. We have limited visibility across platforms where the security doesn't really talk and it works in silos in each platform with machine learning, we have automated analy detection, which is continuously learning from data patterns. And it detects the anomalies and flags to the system. We have predictive threat modeling, which predicts the threats even before they occur. By behavioral analysis, we have adaptive access management, which switches the accesses dynamically based on the role of the accessor. The platform security correlation helps in orchestrating the security across multiple platforms. And this is what we'll see in the multi-cloud architecture by JP Morgan J. Machine learning fundamentally transforms cloud security by enabling systems to learn automatically from the data patterns, predict potential vulnerabilities, and take defenses in real time. This shift from reactive to proactive security postures is particularly crucial in customer relationship management industry, where the customer data sensitivity is often. It's often accompanied with complex access requirements. Let's dive into machine learning security architecture components. There are four building blocks to it. One is threat detection, where the machine learning models identify enamels behavioral patterns and potential security breaches. Second one is access intelligence, adaptive authentication systems that adjust the security requirements based on the risk profiles. Data protection, which automatically classifies the data into highly sensitive information and enforces the encryption security orchestration, which helps us in orchestrating security across multiple systems. It coordinates security response. These core components form the foundation of modern machine learning, enhanced security architectures. When properly implemented, they create a very synergistic system that continuously improves itself and it adapt its defensive capabilities through learning while maintaining compliance with regulatory frameworks. Now, let's take a look into Starbucks. Starbucks has implemented Salesforce's CRM security architecture. What are the core concepts that Starbucks implement in their security enhancement? One is personalized experience, which has advanced machine learning algorithms that deliver tailored customer recommendations while enforcing robust security protocols. Second one is behavioral monitoring, sophisticated pattern recognition. We'll identify and flag enamel analyst transactions in real time. Third one is going to be data classification, which is com compartmentalization, where it automatically categorizes the data during data ingestion itself and enforces granular access controls and encrypts the data as well. So this Starbucks has revolutionized their security framework by embedding advanced machine learning algorithms within their Salesforce CRM environment, creating a seamless fusion. Of customer experiences and also robust security measures. This implementation demonstrates how intelligent data processing can significantly reduce any potential attacks while maintaining performance at global enterprise scale. The breakthrough innovation lies in their security first approach to personalization, where protection mechanisms are woven into the customer engine. This inter integration ensures that data safeguards automatically scale in proportion to customer interactions establishing a new paradigm for secure CRM implementation. Let's take a look at the technical details. There are three main components to the security architecture that Starbucks have implemented. Which is first one is Einstein's Einstein Analytics integration, which is a core feature of Salesforce. This in this feature, there are custom machine learning models for correlating security events with customer behavior patterns across 30,000 plus locations globally. There's real time encryption pipeline where automatic field level encryption is enforced based on data categorization, and this is being processed over 87 million plus customer records while maintaining quick response time. Adaptive permission framework, which is again, the role-based access we talk about, we talked about as one of the core building components of machine learning architecture, the context of our access controls, adjust the employee data access based on behavioral baselines that are set up by the company and ly detection across two 50,000 plus user accounts. The technical implementation leverages Salesforce shield with some customer extensions, creating a security fabric that doesn't compromise the customer experience. This approach determines that security and functionality can be complementary rather than competing priorities. Let's move on to Capital One, AWS breach analysis. So the backstory of it in short would be there is a mis misconfigured web application firewall inside AWS environment. The attacker was able to use SSRF vulnerability to access the metadata service, which is server site forgery request. How was the attacker able to use it? The attacker was able to gain access to IAM credential access temporarily, which has access permission. And was successfully able to data exfiltrate a hundred million plus customer records without detection. So this 2019 Capital One breach serves as a critical case study in cloud security failures analysis reveals that machine learning could have prevented this. Let's do dive into how it could have been implemented. The first component that comes into picture for machine learning is behavioral analysis. So the machine learning would detect abnormal IAM credential usage patterns, misconfiguration scanning. So whenever there is any insecure configuration or invalid configuration built machine learning can automatically scan the system based on the regulatory checks that we enforce, and it automatically identifies any insecure wire firewall settings, access intelligence, role-based access control. It evaluates the access based on the context for resource request. Right? Modern Machine Learning security solutions could have prevented the speech through the about three, which is continuous monitoring of credential usage patterns, automatic detection of configuration, drift, and alerting the system that there is. Of Misconfigured Firewall by applying zero trust principles in role-based access controls which is enhanced with machine learning decision making. The system would have required additional verification for unusual access patterns, which potentially blocks the attack before data exposure. Now let's dive into Mayo Clinic Healthcare Machine Learning security. The three main components that Mayo Clinic uses in machine learning security are data classification. So there is machine learning powered automatic data classification of protected health information with 99.3% accuracy across 23 million patient records ensuring appropriate security controls are applied contextually. Advanced role-based access control intelligence, the predictive analysis, the predictive access patterns analysis that identifies potential unauthorized access attempt before they occur with 87% reduction in false positives compared to traditional rule-based systems. Automated compliance monitor, which is scanning the system continuously for HIPAA compliance verification, using machine learning models, trained on proper audit findings. Maintaining compliance while processing 8.4 TB of daily healthcare data. This implementation is in Microsoft Cloud for healthcare, and this demonstrates how machine learning transforms static security controls into dynamic learning systems that adapt to complex workflows of healthcare industry without compromising. The compliance requirements. Let's dive into the Mayo Clinic's machine learning security architecture. The first one would be data ingestion, which is categorizing the data automatically while intaking. Okay with classification and encryption enforced unclassified data. The second component would be machine learning process, where all the behavioral and access pattern analysis and anomaly detection happens. Third component is compliance verification, which is scanning the system continuously with our automated regulatory checks and validations that we put in place. Fourth one would be contextual access, which is switching the permissions dynamically based on context or role of whoever is trying to access the resources. Mayo Clinic's architecture creates a continuous learning system that enhances protection while streamlining clinical workflows. Their machine learning models evolved from 76% accuracy to over 98% through operational feedback loops, demonstrating how healthcare organizations can leverage AI to improve security posture while meeting specialized industry requirements. Let's dive into our last use case. JP Morgan Chase, multi-cloud machine learning security. So JP Morgan Chase has used different clouds to for different security aspects. They've used AWS for AWS security, which has machine learning powered ly detection engine for transaction processing and database access, which is handling over 5 trillion daily data points. Second one would be GCP Google Cloud Platform, which has AI model training environment with specialized container security and federated learning across jurisdictions. Third one is Microsoft Azure, which takes care of customer facing services with adaptive authentication and real time threat intelligence integration. This multi-cloud strategy represents the cutting edge of machine learning powered security orchestration. Their implementation creates a unified security fabric across A-W-S-G-C-P and Azure while maintaining consistent policy enforcement through machine learning governance models. This approach enables them to leverage the strengths of each cloud provider while maintaining a coherent security posture that meets stringent requirements of financial services regulation. The advantage that they have seen with this cross cloud machine learning security architecture, which 99.7% fraud detection, so the accuracy of machine learning driven transaction system processing over $7.4 trillion annually. They have 5,000 plus security models deployed using machine learning algorithms across clouds for specific security functions. 89% incident reduction, decrease in security incidents following machine learning implementation. The average time for any identified threats. Is three minutes, which is very quick. JP Morgan's approach demonstrates how enterprises can create a coherent security strategy across multi-cloud platforms. Their central machine learning orchestration platforms normalizes security telemetry from different providers, enabling comprehensive threat detection regardless of where the workloads are deployed. For any implementation of machine learning, enhanced security, there are four. Building blocks that we will have to start with. The first one is security assessment, connecting a comprehensive evaluation of existing security controls, identifying gaps and opportunities for machine learning enhancement. We'll have to map the current data flows to understand where machine learning can provide the highest security value if we plug it in. Then we start with developing the models, which is building and training the machine learning models for specific security functions like anomaly detection, role-based access controls, compliance monitoring, scanning the system, establishing the baseline for all of the behavioral profiles and validation methodologies Also. Come under Model Development integration architecture. Now comes designing the technical architecture for integrating this machine learning security components with existing cloud infrastructure, ensuring appropriate data access is given and the processing capabilities are granted. Last one would be governance framework, where you'll be establishing an oversight process for machine learning security models, including guardrails, performance monitoring, bias detection, continuous improvement methodologies. So this framework provides a structured approach to implementing machine learning within your security architecture, focusing on high value use cases while maintaining appropriate governance and compliance awareness. So what are the key takeaways and next steps for us? Machine learning fundamentally transforms cloud security from static rules to adaptive reactive to proactive intelligent systems that continuously improve secure successful implementations like Starbucks, may clinic demonstrate that security and functionality can be complimentary rather than competing priorities. As you begin your implementation journey, focus on identifying high value use cases where machine learning can plug in and address specific security challenges in your environment. Start with well-defined problems. Establish clear success metrics, build towards a comprehensive machine learning security framework through iterative improvement. Thank you.