Data Leakage Detection for AWS Bedrock

As artificial intelligence (AI) becomes deeply integrated into enterprise workflows, organizations are increasingly leveraging cloud-based AI services to enhance efficiency and decision-making.

‍

In 2024, 56% of organizations adopted AI to develop custom applications, with 39% of Azure users leveraging Azure OpenAI services. However, with rapid AI adoption in cloud environments, security risks are escalating. As AI continues to shape business operations, the security and privacy risks associated with cloud-based AI services must not be overlooked. Understanding these risks (and how to mitigate them) is essential for organizations looking to protect their proprietary models and sensitive data.

‍

When discussing AI services in cloud environments, there are two primary types of services that introduce different types of security and privacy risks. This article dives into these risks and explores best practices to mitigate them, ensuring organizations can leverage AI securely and effectively.

1. Leading Generative AI Platforms & Their Business Applications

Examples include OpenAI, Google, Meta, and Microsoft, which develop large-scale AI models and provide AI-related services, such as Azure OpenAI, Amazon Bedrock, Google’s Bard, Microsoft Copilot Studio. These services allow organizations to build AI Agents and GenAI services that  are designed to help users perform tasks more efficiently by integrating with existing tools and platforms. For instance, Microsoft Copilot can provide writing suggestions, summarize documents, or offer insights within platforms like Word or Excel.

What is RAG (Retrieval-Augmented Generation)?

Many AI systems use Retrieval-Augmented Generation (RAG) to improve accuracy. Instead of solely relying on a model’s pre-trained knowledge, RAG allows the system to fetch relevant data from external sources, such as a vector database, using algorithms like k-nearest neighbor. This retrieved information is then incorporated into the model’s response.

‍

When used in enterprise AI applications, RAG enables AI agents to provide contextually relevant responses. However, it also introduces a risk - if access controls are too broad, users may inadvertently gain access to sensitive corporate data.

How Does RAG (Retrieval-Augmented Generation) Apply to AI Agents?

In AI agents, RAG is typically used to enhance responses by retrieving relevant information from a predefined knowledge base.

‍

Example: In AWS Bedrock, you can define a serverless vector database in OpenSearch as a knowledge base for a custom AI agent. This setup allows the agent to retrieve and incorporate relevant context dynamically, effectively implementing RAG.

Security Risks of Generative AI Platforms

Custom generative AI applications, such as AI agents or enterprise-built copilots, are often integrated with organizational knowledge bases like Amazon S3, SharePoint, Google Drive, and other data sources. While these models are typically not directly trained on sensitive corporate data, the fact that they can access these sources creates significant security risks.

One potential risk is data exposure through prompts, but this only arises under certain conditions. If access controls aren’t properly configured, users interacting with AI agents might unintentionally or maliciously - prompt the model to retrieve confidential or private information.This isn’t limited to cleverly crafted prompts; it reflects a broader issue of improper access control and governance.

Configuration and Access Control Risks

The configuration of the AI agent is a critical factor. If an agent is granted overly broad access to enterprise data without proper role-based restrictions, it can return sensitive information to users who lack the necessary permissions. For instance, a model connected to an S3 bucket with sensitive customer data could expose that data if permissions aren’t tightly controlled.

‍

A common scenario might involve an AI agent designed for Sales that has access to personally identifiable information (PII) or customer records. If the agent is not properly restricted, it could be queried by employees outside of Sales, such as developers - who should not have access to that data.

Example Risk Scenario

An employee asks a Copilot-like agent to summarize company-wide sales data. The AI returns not just high-level figures, but also sensitive customer or financial details that were unintentionally exposed due to lax access controls.

Challenges in Mitigating These Risks

The core challenge, particularly relevant to platforms like Sentra, is enforcing governance to ensure only appropriate data is used and accessible by AI services.

‍

This includes:

• Defining and enforcing granular data access controls.
• Preventing misconfigurations or overly permissive settings.
• Maintaining real-time visibility into which data sources are connected to AI models.
• Continuously auditing data flows and access patterns to prevent leaks.

Without rigorous governance and monitoring, even well-intentioned GenAI implementations can lead to serious data security incidents.

2. ML and AI Studios for Building New Models

Many companies, such as large financial institutions, build their own AI and ML models to make better business decisions, or to improve their user experiences. Unlike large foundational models from major tech companies, these custom AI models are trained by the organization itself on their applications or corporate data.
‍

Security Risks of Custom AI Models

1. Weak Data Governance Policies - If data governance policies are inadequate, sensitive information, such as customers' Personally Identifiable Information (PII), could be improperly accessed or shared during the training process. This can lead to data breaches, privacy compliance violations, and unethical AI usage. The growing recognition of AI-related risks has driven the development of more AI compliance frameworks.
   ‍
2. Excessive Access to Training Data and AI Models - Granting unrestricted access to training datasets and machine learning (ML)/AI models increases the risk of data leaks and misuse. Without proper access controls, sensitive data used in training can be exposed to unauthorized individuals, leading to compliance and security concerns.
   ‍
3. AI Agents Exposing Sensitive Data -  AI agents that do not have proper safeguards can inadvertently expose sensitive information to a broad audience within an organization. For example, an employee could retrieve confidential data such as the CEO’s salary or employment contracts if access controls are not properly enforced.
   ‍
4. Insecure Model Storage – Once a model is trained, it is typically stored in the same environment (e.g., in Amazon SageMaker, the training job stores the trained model in S3). If not properly secured, proprietary models could be exposed to unauthorized access, leading to risks such as model theft.
   ‍
5. Deployment Vulnerabilities – A lack of proper access controls can result in unauthorized use of AI models. Organizations need to assess who has access: Is the model public? Can external entities interact with or exploit it?

Shadow AI and Forgotten Assets – AI models or artifacts that are not actively monitored or properly decommissioned can become a security risk. These overlooked assets can serve as attack vectors if discovered by malicious actors.

Example Risk Scenario

A bank develops an AI-powered feature that predicts a customer’s likelihood of repaying a loan based on inputs like financial history, employment status, and other behavioral indicators. While this feature is designed to enhance decision-making and customer experience, it introduces significant risk if not properly governed.
‍

During development and training, the model may be exposed to personally identifiable information (PII), such as names, addresses, social security numbers, or account details, which is not necessary for the model’s predictive purpose.
‍

⚠️ Best practice: Models should be trained only on the minimum necessary data required for performance, excluding direct identifiers unless absolutely essential. This reduces both privacy risk and regulatory exposure.

If the training pipeline fails to properly separate or mask this PII, the model could unintentionally leak sensitive information. For example, when responding to an end-user query, the AI might reference or infer details from another individual’s record - disclosing sensitive customer data without authorization.

‍

This kind of data leakage, caused by poor data handling or weak governance during training, can lead to serious regulatory non-compliance, including violations of GDPR, CCPA, or other privacy frameworks.

Common Risk Mitigation Strategies and Their Limitations

Many organizations attempt to manage AI-related risks through employee training and awareness programs. Employees are taught best practices for handling sensitive data and using AI tools responsibly.
While valuable, this approach has clear limitations:
‍

• Training Alone Is Insufficient:
  Human error remains a major risk factor, even with proper training. Employees may unintentionally connect sensitive data sources to AI models or misuse AI-generated outputs.
  
  
• Lack of Automated Oversight:
  Most organizations lack robust, automated systems to continuously monitor how AI models use data and to enforce real-time security policies. Manual review processes are often too slow and incomplete to catch complex data access risks in dynamic, cloud-based AI environments.
  ‍
• ‍Policy Gaps and Visibility Challenges:
  Organizations often operate with multiple overlapping data layers and services. Without clear, enforceable policies, especially automated ones - certain data assets may remain unscanned or unprotected, creating blind spots and increasing risk.

Reducing AI Risks with Sentra’s Comprehensive Data Security Platform

Managing AI risks in the cloud requires more than employee training.
Organizations need to adopt robust data governance frameworks and data security platforms (like Sentra’s) that address the unique challenges of AI.

This includes:

• Discovering AI Assets: Automatically identify AI agents, knowledge bases, datasets, and models across the environment.
• Classifying Sensitive Data: Use automated classification and tagging to detect and label sensitive information accurately.
  Monitoring AI Data Access: Detect which AI agents and models are accessing sensitive data, or using it for training - in real time.
• Enforcing Access Governance: Govern AI integrations with knowledge bases by role, data sensitivity, location, and usage to ensure only authorized users can access training data, models, and artifacts.
• Automating Data Protection: Apply masking, encryption, access controls, and other protection methods automatically across data and AI artifacts used in training and inference processes.

By combining strong technical controls with ongoing employee training, organizations can significantly reduce the risks associated with AI services and ensure compliance with evolving data privacy regulations.

<blogcta-big>

‍

Imagine the fallout of a single misstep—a phishing scam tricking an employee into sharing sensitive data. The breach doesn’t just compromise information; it shakes trust, tarnishes reputations, and invites compliance penalties. With data breaches on the rise, safeguarding your organization’s Microsoft 365 environment has never been more critical.

‍

Data classification helps prevent such disasters. This article provides a clear roadmap for protecting and classifying Microsoft 365 data. It explores how data is saved and classified, discusses built-in tools for protection, and covers best practices for maintaining  Microsoft 365 data protection.

‍

How Is Data Saved and Classified in Microsoft 365? 

‍

Microsoft 365 stores data across tools and services. For example, emails are stored in Exchange Online, while documents and data for collaboration are found in Sharepoint and Teams, and documents or files for individual users are stored in OneDrive. This data is primarily unstructured—a format ideal for documents and images but challenging for identifying sensitive information.

‍

All of this data is largely stored in an unstructured format typically used for documents and images. This format not only allows organizations to store large volumes of data efficiently; it also enables seamless collaboration across teams and departments. However, as unstructured data cannot be neatly categorized into tables or columns, it becomes cumbersome to discern what data is sensitive and where it is stored. 

‍

To address this, Microsoft 365 offers a data classification dashboard that helps classify data of varying levels of sensitivity and data governed by different regulatory compliance frameworks. But how does Microsoft identify sensitive information with unstructured data? 

‍

Microsoft employs advanced technologies such as RegEx scans, trainable classifiers, Bloom filters, and data classification graphs to identify and classify data as public, internal, or confidential. Once classified, data protection and governance policies are applied based on sensitivity and retention labels.

‍

Data classification is vital for understanding, protecting, and governing data. With your ​​Microsoft 365 data classified appropriately, you can ensure seamless collaboration without risking data exposure.

Microsoft 365 Data Protection and Classification Tools

Microsoft 365 includes several key tools and frameworks for classifying and securing data. Here are a few. 

Microsoft Purview 

Microsoft Purview is a cornerstone of data classification and protection within Microsoft 365.

‍

Key Features: 

• Over 200+ prebuilt classifiers and the ability to create custom classifiers tailored to specific business needs.
• Purview auto-classifies data across Microsoft 365 and other supported apps, such as Adobe Photoshop and Adobe PDF, while users work on them.
• Sensitivity labels that apply encryption, watermarks, and access restrictions to secure sensitive data.‍
• Double Key Encryption to ensure that sensitivity labels persist even when file formats change.
  ‍

‍

Purview autonomously applies sensitivity labels like "confidential" or "highly confidential" based on preconfigured policies, ensuring optimal access control. These labels persist even when files are shared or converted to other formats, such as from Word to PDF.

‍

Additionally, Purview’s data loss prevention (DLP) policies prevent unauthorized sharing or deletion of sensitive data by flagging and reporting violations in real time. For example, if a sensitive file is shared externally, Purview can immediately block the transfer and alert your security team.

Microsoft Defender 

Microsoft Defender for Cloud Apps strengthens security by providing a cloud app discovery window to identify applications accessing data. Once identified, it classifies files within these applications based on sensitivity, applying appropriate protections as per preconfigured policies.

‍

Key Features:

• Data Sensitivity Classification: Defender identifies sensitive files and assigns protection based on sensitivity levels, ensuring compliance and reducing risk. For example, it labels files containing credit card numbers, personal identifiers, or confidential business information with sensitivity classifications like "Highly Confidential."
• Threat Detection and Response: Defender detects known threats targeted at sensitive data in emails, collaboration tools (like SharePoint and Teams), URLs, file attachments, and OneDrive. If an admin account is compromised, Microsoft Defender immediately spots the threat, disables the account, and notifies your IT team to prevent significant damage.‍
• Automation: Defender automates incident response, ensuring that malicious activities are flagged and remediated promptly.

Intune 

Microsoft Intune provides comprehensive device management and data protection, enabling organizations to enforce policies that safeguard sensitive information on both managed and unmanaged smartphones, computers, and other devices.

‍

Key Features:

• Customizable Compliance Policies: Intune allows organizations to enforce device compliance policies that align with internal and regulatory standards. For example, it can block non-compliant devices from accessing sensitive data until issues are resolved.
• Data Access Control: Intune disallows employees from accessing corporate data on compromised devices or through insecure apps, such as those not using encryption for emails.‍
• Endpoint Security Management: By integrating with Microsoft Defender, Intune provides endpoint protection and automated responses to detected threats, ensuring only secure devices can access your organization’s network.

‍

Intune supports organizations by enabling the creation and enforcement of device compliance policies tailored to both internal and regulatory standards. These policies detect non-compliant devices, issue alerts, and restrict access to sensitive data until compliance is restored. Conditional access ensures that only secure and compliant devices connect to your network.

‍

Microsoft 365-managed apps like Outlook, Word, and Excel. These policies define which apps can access specific data, such as emails, and regulate permissible actions, including copying, pasting, forwarding, and taking screenshots. This layered security approach safeguards critical information while maintaining seamless app functionality.

Does Microsoft have a DLP Solution?

Microsoft 365’s data loss prevention (DLP) policies represent the implementation of the zero-trust framework. These policies aim to prevent oversharing, accidental deletion, and data leaks across Microsoft 365 services, including Exchange Online, SharePoint, Teams, and OneDrive, as well as Windows and macOS devices.

‍

Retention policies, deployed via retention labels, help organizations manage the data lifecycle effectively.These labels ensure that data is retained only as long as necessary to meet compliance requirements, reducing the risks associated with prolonged data storage.

What is the Microsoft 365 Compliance Center?

The Microsoft 365 compliance center offers tools to manage policies and monitor data access, ensuring adherence to regulations. For example, DLP policies allow organizations to define specific automated responses when certain regulatory requirements—like GDPR or HIPAA—are violated.

‍

Microsoft Purview Compliance Portal: This portal ensures sensitive data is classified, stored, retained, and used in adherence to relevant compliance regulations. Meanwhile, Microsoft 365’s MPIP ensures that only authorized users can access sensitive information, whether collaborating on Teams or sharing files in SharePoint. Together, these tools enable secure collaboration while keeping regulatory compliance at the forefront.

‍

12 Best Practices for Microsoft 365 Data Protection and Classification

To achieve effective Microsoft 365 data protection and classification, organizations should follow these steps:

1. Create precise labels, tags, and classification policies; don’t rely solely on prebuilt labels and policies, as definitions of sensitive data may vary by context.
2. Automate labeling to minimize errors and quickly capture new datasets.
3. Establish and enforce data use policies and guardrails automatically to reduce risks of data breaches, compliance failures, and insider threat risks. 
4. Regularly review and update data classification and usage policies to reflect evolving threats, new data storage, and changing compliance laws.o policies must stay up to date to remain effective.
5. Define context-appropriate DLP policies based on your business needs; factoring in remote work, ease of collaboration, regional compliance standards, etc.
6. Apply encryption to safeguard data inside and outside your organization.
7. Enforce role-based access controls (RBAC) and least privilege principles to ensure users only have access to data and can perform actions within the scope of their roles. This limits the risk of accidental data exposure, deletion, and cyberattacks.
8. Create audit trails of user activity around data and maintain version histories to prevent and track data loss.
9. Follow the 3-2-1 backup rule: keep three copies of your data, store two on different media, and one offsite.
10. Leverage the full suite of Microsoft 365 tools to monitor sensitive data, detect real-time threats, and secure information effectively.
11. Promptly resolve detected risks to mitigate attacks early.
12. Ensure data protection and classification policies do not impede collaboration to prevent teams from creating shadow data, which puts your organization at risk of data breaches.

For example, consider #3. If a disgruntled employee starts transferring sensitive intellectual property to external devices in preparation for a ransomware attack, having the right data use policies in place will allow your organization to stop the threat before it escalates. 

Microsoft 365 Data Protection and Classification Limitations

Despite Microsoft 365’s array of tools, there are some key gaps. AI/ML-powered data security posture management (DSPM) and data detection and response (DDR) solutions fill these easily.

‍

The top limitations of Microsoft 365 data protection and classification are the following:

• Limitations Handling Large Volumes of Unstructured Data: Purview struggles to automatically classify and apply sensitivity labels to diverse and vast datasets, particularly in Azure services or non-Microsoft clouds. 
• Contextless Data Classification: Without considering context, Microsoft Purview’s MPIP can lead to false positives (over-labeling non-sensitive data) or false negatives (missing sensitive data). 
• Inconsistent Labeling Across Providers: Microsoft tools are limited to its ecosystem, making it difficult for enterprises using multi-cloud environments to enforce consistent organization-wide labeling.
• Minimal Threat Response Capabilities: Microsoft Defender relies heavily on IT teams for remediation and lacks robust autonomous responses.
• Sporadic Interruption of User Activity: Inaccurate DLP classifications can disrupt legitimate data transfers in collaboration channels, frustrating employees and increasing the risk of shadow IT workarounds.

Sentra Fills the Gap: Protection Measures to Address Microsoft 365 Data Risks

Today’s businesses must get ahead of data risks by instituting Microsoft 365 data protection and classification best practices such as least privilege access and encryption. Otherwise, they risk data exposure, damaging cyberattacks, and hefty compliance fines. However, implementing these best practices depends on accurate and context-sensitive data classification in Microsoft 365. 

‍

Sentra’s Cloud-native Data Security Platform enables secure collaboration and file sharing across all Microsoft 365 services including SharePoint, OneDrive, Teams, OneNote, Office, Word, Excel, and more. Sentra provides data access governance, shadow data detection, and privacy audit automation for M365 data. It also evaluates risks and alerts for policy or regulatory violations.

‍

Specifically, Sentra complements Purview in the following ways:

1. Sentra Data Detection & Response (DDR): Continuously monitors for threats such as data exfiltration, weakening of data security posture, and other suspicious activities in real time. While Purview Insider Risk Management focuses on M365 applications, Sentra DDR extends these capabilities to Azure and non-Microsoft applications.
2. Data Perimeter Protection: Sentra automatically detects and identifies an organization’s data perimeters across M365, Azure, and non-Microsoft clouds. It alerts “organizations when sensitive data leaves its boundaries, regardless of how it is copied or exported.
3. Shadow Data Reduction: Using context-based analysis powered by Sentra’s DataTreks™, the platform identifies unnecessary shadow data, reducing the attack surface and improving data governance.
4. Training Data Monitoring: Sentra monitors training datasets continuously, identifying privacy violations of sensitive PII or real-time threats like training data poisoning or suspicious access.
5. Data Access Governance: Sentra adds to Purview’s data catalog by including metadata on users and applications with data access permissions, ensuring better governance.
6. Automated Privacy Assessments: Sentra automates privacy evaluations aligned with frameworks like GDPR and CCPA, seamlessly integrating them into Purview’s data catalog.
7. Rich Contextual Insights: Sentra delivers detailed data context to understand usage, sensitivity, movement, and unique data types. These insights enable precise risk evaluation, threat prioritization, and remediation, and they can be consumed via an API by DLP systems, SIEMs, and other tools.

‍

By addressing these gaps, Sentra empowers organizations to enhance their Microsoft 365 data protection and classification strategies. Request a demo to experience Sentra’s innovative solutions firsthand.

<blogcta-big>

‍

This RFP Guide is designed to help organizations create their own RFP for selection of Cloud-native Data Security Platform (DSP) & Data Security Posture Management (DSPM) solutions. The purpose is to identify key essential requirements  that will enable effective discovery, classification, and protection of sensitive data across complex environments, including in public cloud infrastructures and in on-premises environments.

Instructions for Vendors

Each section provides essential and recommended requirements to achieve a best practice capability. These have been accumulated over dozens of customer implementations.  Customers may also wish to include their own unique requirements specific to their industry or data environment.

1. Data Discovery & Classification

‍

‍

2. Data Access Governance

‍

3. Posture, Risk Assessment & Threat Monitoring

‍

‍

4. Incident Response & Remediation

‍

5. Infrastructure & Deployment

‍

‍

6. Operations & Support

‍

‍

7. Pricing & Licensing

‍

‍

Conclusion

This RFP template is designed to facilitate a structured and efficient evaluation of DSP and DSPM solutions. Vendors are encouraged to provide comprehensive and transparent responses to ensure an accurate assessment of their solution’s capabilities.

‍

Sentra’s cloud-native design combines powerful Data Discovery and Classification, DSPM, DAG, and DDR capabilities into a complete Data Security Platform (DSP). With this, Sentra customers achieve enterprise-scale data protection and do so very efficiently - without creating undue burdens on the personnel who must manage it.

‍

To learn more about Sentra’s DSP, request a demo here and choose a time for a meeting with our data security experts. You can also choose to download the RFP as a pdf.