<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0"><channel><title><![CDATA[CapMinds]]></title><description><![CDATA[CapMinds]]></description><link>https://capminds.hashnode.dev</link><generator>RSS for Node</generator><lastBuildDate>Fri, 20 Nov 2026 19:54:12 GMT</lastBuildDate><atom:link href="https://capminds.hashnode.dev/rss.xml" rel="self" type="application/rss+xml"/><language><![CDATA[en]]></language><ttl>60</ttl><item><title><![CDATA[Avoiding Common S3 Misconfigurations That Lead to HIPAA Breaches]]></title><description><![CDATA[Learn how to prevent common Amazon S3 misconfigurations, secure PHI, manage access permissions, and reduce HIPAA compliance risks in healthcare.]]></description><link>https://capminds.hashnode.dev/avoiding-common-s3-misconfigurations-that-lead-to-hipaa-breaches</link><guid isPermaLink="true">https://capminds.hashnode.dev/avoiding-common-s3-misconfigurations-that-lead-to-hipaa-breaches</guid><category><![CDATA[HIPAA compliance services]]></category><category><![CDATA[AWS]]></category><category><![CDATA[Cloud Computing]]></category><category><![CDATA[cybersecurity]]></category><dc:creator><![CDATA[CapMinds Tech]]></dc:creator><pubDate>Thu, 19 Nov 2026 12:17:48 GMT</pubDate><content:encoded><![CDATA[<h2>1. Disabling Block Public Access and overlooking broad bucket policies</h2>
<p>Teams often disable S3 Block Public Access for testing and forget to restore it. Buckets created since April 2023 start with BPA enabled, so exposure usually comes from an explicit change or an older bucket.</p>
<p>With BPA off, a policy with "Principal": "*" and s3:GetObject makes every object readable by anyone. Enable all four settings (BlockPublicAcls, IgnorePublicAcls, BlockPublicPolicy, RestrictPublicBuckets) at both levels:</p>
<p><code>aws s3control put-public-access-block --account-id 111122223333 \</code></p>
<p><code>--public-access-block-configuration \</code></p>
<p><code>BlockPublicAcls=true,IgnorePublicAcls=true,BlockPublicPolicy=true,RestrictPublicBuckets=true</code></p>
<p><code>aws s3api put-public-access-block --bucket healthcare-phi-bucket \</code></p>
<p><code>--public-access-block-configuration \</code></p>
<p><code>BlockPublicAcls=true,IgnorePublicAcls=true,BlockPublicPolicy=true,RestrictPublicBuckets=true</code></p>
<p>BPA only addresses public access, so also review bucket policies, access points, and cross-account grants. For an analytics vendor or backup service, name the specific role, the minimum action, and the prefix:</p>
<p><code>{</code></p>
<p><code>"Sid": "AnalyticsVendorReadExports",</code></p>
<p><code>"Effect": "Allow",</code></p>
<p><code>"Principal": { "AWS": "arn:aws:iam::444455556666:role/vendor-analytics-reader" },</code></p>
<p><code>"Action": "s3:GetObject",</code></p>
<p><code>"Resource": "arn:aws:s3:::healthcare-phi-bucket/analytics-exports/*"</code></p>
<p><code>}</code></p>
<p>Require sts: ExternalId when a vendor assumes a role in your account. Run IAM Access Analyzer for S3 to flag buckets shared outside your organization.</p>
<h2>2. Overly permissive IAM roles and legacy ACLs</h2>
<p>An application that reads patient documents does not need s3:*. That wildcard includes s3:DeleteObject and bucket-management actions such as s3:PutBucketPolicy, so a compromised role could exfiltrate PHI, delete it, or re-expose the bucket.</p>
<p>Grant only what the function needs. Object actions attach to the object ARN, while s3:ListBucket attaches to the bucket ARN:</p>
<p><code>[</code></p>
<p><code>{ "Effect": "Allow", "Action": "s3:GetObject",</code></p>
<p><code>"Resource": "arn:aws:s3:::healthcare-phi-bucket/patient-documents/*" },</code></p>
<p><code>{ "Effect": "Allow", "Action": "s3:ListBucket",</code></p>
<p><code>"Resource": "arn:aws:s3:::healthcare-phi-bucket",</code></p>
<p><code>"Condition": { "StringLike": { "s3:prefix": ["patient-documents/*"] } } }</code></p>
<p><code>]</code></p>
<p>If the bucket uses SSE-KMS, the role also needs kms: Decrypt. Use roles with short-lived credentials, and use IAM Access Analyzer policy generation to trim permissions to actual CloudTrail activity.</p>
<p>Legacy ACLs add risk. The AuthenticatedUsers group means any AWS account holder worldwide, not your organization’s users. Where compatibility allows, disable ACLs:</p>
<p><code>aws s3api put-bucket-ownership-controls --bucket healthcare-phi-bucket \</code></p>
<p><code>--ownership-controls 'Rules=[{ObjectOwnership=BucketOwnerEnforced}]'</code></p>
<p>Before enabling it, find ACL dependencies in server access logs and CloudTrail data events using the aclRequired field, and list object ACLs with S3 Inventory. The call fails with InvalidBucketAclWithObjectOwnership until any bucket ACL grants to other accounts are migrated to policies.</p>
<h2>3. Treating encryption as the only HIPAA security control:</h2>
<p>SSE-S3 and SSE-KMS protect stored data, but authorized callers still receive decrypted objects. A customer-managed KMS key adds a second check. Callers need both S3 permission and kms: Decrypt, and every decrypt is logged in CloudTrail. Enable automatic key rotation and S3 Bucket Keys to cut KMS request volume.</p>
<p>Enforce HTTPS with a bucket policy deny. S3 endpoints already require TLS 1.2 or higher, so the secure-transport condition is what blocks plain HTTP:</p>
<p><code>{</code></p>
<p><code>"Sid": "DenyInsecureTransport",</code></p>
<p><code>"Effect": "Deny",</code></p>
<p><code>"Principal": "*",</code></p>
<p><code>"Action": "s3:*",</code></p>
<p><code>"Resource": ["arn:aws:s3:::healthcare-phi-bucket", "arn:aws:s3:::healthcare-phi-bucket/*"],</code></p>
<p><code>"Condition": { "Bool": { "aws:SecureTransport": "false" } }</code></p>
<p><code>}</code></p>
<p>Pre-signed URLs are bearer tokens. Anyone holding one has the signer’s permissions on that object. A URL signed with role credentials expires when the role session ends, even if its own expiry is later. Keep expiry to minutes, sign with a narrow dedicated role, and cap validity centrally:</p>
<p><code>{</code></p>
<p><code>"Sid": "LimitPresignedUrlValidity",</code></p>
<p><code>"Effect": "Deny",</code></p>
<p><code>"Principal": "*",</code></p>
<p><code>"Action": "s3:*",</code></p>
<p><code>"Resource": "arn:aws:s3:::healthcare-phi-bucket/*",</code></p>
<p><code>"Condition": { "NumericGreaterThan": { "s3:signatureAge": 600000 } }</code></p>
<p><code>}</code></p>
<p>If a URL leaks, shortening future expiry does not invalidate it. A pre-signed URL stops working when the credentials that signed it are revoked, deleted, or deactivated. Deactivate the signer’s access key or use Revoke active sessions on the role, then investigate.</p>
<h2>4. Missing audit logs and infrastructure deployment guardrails:</h2>
<p>CloudTrail management events record changes like PutBucketPolicy, but not GetObject, PutObject, or DeleteObject. Enable S3 data events for PHI buckets only, since they are billed per event. Keep the trailing slash in the ARN:</p>
<p><code>aws cloudtrail put-event-selectors --trail-name phi-audit-trail \</code></p>
<p><code>--advanced-event-selectors '[{</code></p>
<p><code>"Name": "PHI object-level events",</code></p>
<p><code>"FieldSelectors": [</code></p>
<p><code>{ "Field": "eventCategory", "Equals": ["Data"] },</code></p>
<p><code>{ "Field": "resources.type", "Equals": ["AWS::S3::Object"] },</code></p>
<p><code>{ "Field": "resources.ARN", "StartsWith": ["arn:aws:s3:::healthcare-phi-bucket/"] }</code></p>
<p><code>]}]'</code></p>
<p>Without data events, you cannot reconstruct which objects were read during an exposure. Use a multi-Region trail with log file validation, delivered to a separate log-archive account with SSE-KMS and Object Lock. Server access logging adds request detail but is delivered best-effort, so treat CloudTrail as the record.</p>
<p>Detect drift with AWS Config rules (s3-bucket-public-read-prohibited, s3-bucket-public-write-prohibited, s3-bucket-ssl-requests-only) and Security Hub controls S3.22 and S3.23, which check object-level logging. Use EventBridge alerts on PutBucketPolicy, PutBucketAcl, PutBucketPublicAccessBlock, and DeleteBucketPublicAccessBlock. Add GuardDuty S3 Protection and Amazon Macie for discovery. Review Macie findings rather than treating them as a complete PHI inventory.</p>
<p>Manage buckets as code with Terraform, CloudFormation, or CDK, and scan templates in CI. Then block the risky actions with an SCP:</p>
<p><code>{</code></p>
<p><code>"Sid": "ProtectS3PublicAccessBlock",</code></p>
<p><code>"Effect": "Deny",</code></p>
<p><code>"Action": ["s3:PutAccountPublicAccessBlock", "s3:PutBucketPublicAccessBlock"],</code></p>
<p><code>"Resource": "*",</code></p>
<p><code>"Condition": { "StringNotLike": { "aws: PrincipalArn": "arn:aws:iam::*:role/SecurityBreakGlass" } }</code></p>
<p><code>}</code></p>
<p>SCPs do not apply to the management account, so keep PHI workloads out of it. Enable versioning and Object Lock for backups.</p>
<h2>Can Your S3 Bucket Policies, IAM Roles, and CloudTrail Logs Withstand a HIPAA Audit?</h2>
<p>We audit your S3 access controls, least-privilege IAM, KMS encryption, and CloudTrail audit logging for PHI as part of HIPAA-compliant <a href="https://www.capminds.com/aws-healthcare-cloud-solutions/">AWS cloud security</a>, then add infrastructure-as-code guardrails so your fixes don’t drift back.</p>
<h2>\</h2>
]]></content:encoded></item><item><title><![CDATA[Securing OAuth2 Scopes in Epic Showroom Deployments: Architecture Pitfalls That Stall Production Cutovers]]></title><description><![CDATA[Learn how to secure OAuth2 scopes in Epic Showroom deployments, avoid authorization pitfalls, and resolve integration issues before production cutover]]></description><link>https://capminds.hashnode.dev/securing-oauth2-scopes-in-epic-showroom-deployments-architecture-pitfalls-that-stall-production-cutovers</link><guid isPermaLink="true">https://capminds.hashnode.dev/securing-oauth2-scopes-in-epic-showroom-deployments-architecture-pitfalls-that-stall-production-cutovers</guid><category><![CDATA[epic]]></category><category><![CDATA[epic integration]]></category><category><![CDATA[healthcare]]></category><dc:creator><![CDATA[CapMinds Tech]]></dc:creator><pubDate>Wed, 18 Nov 2026 13:15:05 GMT</pubDate><content:encoded><![CDATA[<p>Most Epic integration failures do not happen because an engineer misunderstood basic OAuth 2.0. They happen during a health system’s Technical Security Review (TSR) when a hospital Systems Architect realizes your application requests coarse scopes that conflict with internal Chronicles security models.</p>
<p>Having architected SMART on FHIR deployments across Epic Hyperspace, Hyperdrive, and MyChart environments, I see the same structural mistakes delay production cutovers by months.</p>
<p>Here is what actually breaks in Epic’s OAuth2 layer and how to architect around it before a health system’s Technical Integration Engineer rejects your build.</p>
<h2>1. Requesting Coarse SMART v1 Scopes Under ONC HTI-1 and SMART v2</h2>
<p>If your Epic Showroom (formerly App Orchard) manifest still requests patient/Observation.read, expect immediate pushback from hospital InfoSec. In Epic’s data model, the Observation resource maps across laboratory results (LWS), vital signs, social history, and sensitive SDOH questionnaires.</p>
<p>Under the HL7 SMART App Launch v2.0.0 specification and ONC HTI-1 mandates [Citation: HL7 SMART App Launch IG v2.0.0; 45 CFR § 170.315(g)(10)], Epic enforces granular CRUD/S syntax (.c, .r, .u, .d, .s) and sub-resource category filters. If your clinical decision support module only evaluates lab panels, requesting blanket read access violates HIPAA’s Minimum Necessary standard</p>
<p>[SS: Epic Interconnect Token Exchange Wire Capture, Granular SMART v2 Scope Enforcement]</p>
<p><code>POST /interconnect-fhir-oauth/oauth2/token HTTP/1.1</code></p>
<p><code>Host:</code> <a href="http://fhir.epic-healthsystem.org"><code>fhir.epic-healthsystem.org</code></a></p>
<p><code>Content-Type: application/x-www-form-urlencoded</code></p>
<p><code>grant_type=authorization_code&amp;code=eyJ0eXAi...&amp;redirect_uri=https%3A%2F%</code><a href="http://2Fapp.vendor.io"><code>2Fapp.vendor.io</code></a><code>%2Fcb</code></p>
<p><code>HTTP/1.1 200 OK</code></p>
<p><code>{</code></p>
<p><code>"access_token": "eyJhbGciOiJSUzI1NiIs...",</code></p>
<p><code>"token_type": "Bearer",</code></p>
<p><code>"expires_in": 3600,</code></p>
<p><code>"scope": "launch patient/</code><a href="http://Observation.rs"><code>Observation.rs</code></a><code>?category=</code><a href="http://terminology.hl7.org/CodeSystem/observation-category%7Claboratory"><code>http://terminology.hl7.org/CodeSystem/observation-category|laboratory</code></a><code>",</code></p>
<p><code>"patient": "erXuFYUfucBZaryVksYEcMg3"</code></p>
<p><code>}</code></p>
<p>Drop legacy .read wildcards. Declare category-constrained .rs scopes so Epic Interconnect restricts token privileges at the gateway before querying Chronicles.</p>
<h2>2. The offline_access Trap in Provider Hyperdrive Launches</h2>
<p>Junior guides routinely tell developers to append offline_access whenever background syncing is needed. In a provider-facing EHR launch inside Epic Hyperdrive, that mistake will get your app blocked during security review.</p>
<p>When a physician launches your app from an active chart, the OAuth token is bound to that clinician’s EMP (User) and SER (Provider) Chronicles records. Hospital CISOs almost universally block offline_access on clinician-launched apps because they will not allow an external server to hold a perpetual refresh token acting on behalf of a logged-out physician. Furthermore, Epic access tokens default to a 3,600-second (1-hour) TTL—not 8 hours, as generic blogs often misstate.</p>
<p>The Architectural Fix: Keep your Hyperdrive EHR-launch session strictly interactive (launch + user/<em>.rs or patient/</em>.rs). If your platform requires asynchronous data ingestion after the clinician closes the chart, decouple the workflows: register a separate SMART Backend Services client ID using client_credentials and system/*.rs scopes authenticated via asymmetric JWT (private_key_jwt)</p>
<h2>3. Ignoring Epic’s Dual-Gate Authorization (OAuth Scope Chronicles Access)</h2>
<p>Passing OAuth2 scope validation at the Epic Interconnect gateway does not guarantee delivery of FHIR resources. Epic enforces a two-tier authorization check:</p>
<p>Outer Gate (OAuth2): Does the Bearer token contain user/DiagnosticReport.rs? Inner Gate (Chronicles Security): Does the authenticated EMP record possess the Epic Security Points and ORL (Department) access required to view that specific encounter, especially for "Break-the-Glass" (BTG) or restricted behavioral health charts?</p>
<p>[SS: Epic Interconnect 403 OperationOutcome When OAuth Passes but EMP Security Fails]</p>
<p><code>{</code></p>
<p><code>"resourceType": "OperationOutcome",</code></p>
<p><code>"issue": [{</code></p>
<p><code>"severity": "error",</code></p>
<p><code>"code": "forbidden",</code></p>
<p><code>"details": {</code></p>
<p><code>"coding": [{</code></p>
<p><code>"system": "urn:oid:1.2.840.114350.1.13.0.1.7.2.657369",</code></p>
<p><code>"code": "4101",</code></p>
<p><code>"display": "User lacks security point to access restricted encounter."</code></p>
<p><code>}]</code></p>
<p><code>}</code></p>
<p><code>}]</code></p>
<p><code>}</code></p>
<p>Never assume a granted OAuth scope guarantees payload access.</p>
<p>Your ingestion pipeline must gracefully handle Epic OID 1.2.840.114350 OperationOutcome exceptions without crashing the clinician's Hyperdrive frame when their EMP profile hits a restricted record.</p>
<h2>4. jwks_uri Caching Collisions and iss Discovery Failures</h2>
<p>Confidential apps authenticating to Epic cannot use symmetric client secrets in production; Epic requires RS384/RS256 signatures verified against a public JWK Set URL (jwks_uri) registered in Vendor Services.</p>
<p>Two operational oversights routinely break production cutovers:</p>
<p>Non-Overlapping Key Rotation: Epic Interconnect caches your jwks_uri payload. If you rotate your private signing key and immediately purge the prior public key from your JWK endpoint, token exchanges fail until Epic’s cache TTL expires. Publish the new kid (Key ID) in your JWK Set at least 48 hours before signing client_assertion JWTs with the new private key. Hardcoding Sandbox Endpoints: Every health system hosts its own Interconnect gateway. During an EHR launch, your app must dynamically extract the iss parameter, fetch {iss}/.well-known/smart-configuration, and validate the iss URI against an explicit allowlist of onboarded customer tenant endpoints before redirecting to /authorize.</p>
<h2>Pre-Cutover Verification Checklist for Your Customer’s Epic TIE</h2>
<p>Scope-to-API Parity: Every requested scope maps 1:1 to a listed API in your Showroom App Record; unused resources are stripped. Client Auth Alignment: Public SPAs/mobile clients enforce PKCE (code_challenge_method=S256); confidential backends pass RFC 7523 client_assertion JWTs with a unique jti nonce and exp 5 minutes. Context Separation: MyChart standalone launches exclusively request patient/* scopes; Hyperdrive EHR launches request launch and respect EMP session lifetimes. Audit Correlation: Outbound logs capture the Epic patient FHIR ID, encounter token claim, and OAuth sub user identifier for HIPAA security rules.</p>
<h2>Need an Architect to Vet Your Epic Showroom Manifest Before Security Review?</h2>
<p>Stop losing months to avoidable TSR rejections. Schedule a <a href="https://www.capminds.com/blog/5-vendors-to-evaluate-for-epic-and-oracle-health-fhir-integration/">Technical Architecture</a> Review with Our Epic FHIR Engineering Team</p>
]]></content:encoded></item><item><title><![CDATA[Securing OAuth2 Scopes in Epic Showroom Deployments: Architecture Pitfalls That Stall Production Cutovers]]></title><description><![CDATA[Learn how to secure OAuth2 scopes in Epic Showroom deployments, avoid authorization pitfalls, and resolve integration issues before production cutover]]></description><link>https://capminds.hashnode.dev/securing-oauth2-scopes-in-epic-showroom-deployments-architecture-pitfalls-that-stall-production-cutovers</link><guid isPermaLink="true">https://capminds.hashnode.dev/securing-oauth2-scopes-in-epic-showroom-deployments-architecture-pitfalls-that-stall-production-cutovers</guid><category><![CDATA[epic]]></category><category><![CDATA[epic integration]]></category><category><![CDATA[healthcare]]></category><dc:creator><![CDATA[CapMinds Tech]]></dc:creator><pubDate>Wed, 18 Nov 2026 13:15:04 GMT</pubDate><content:encoded><![CDATA[<p>Most Epic integration failures do not happen because an engineer misunderstood basic OAuth 2.0. They happen during a health system’s Technical Security Review (TSR) when a hospital Systems Architect realizes your application requests coarse scopes that conflict with internal Chronicles security models.</p>
<p>Having architected SMART on FHIR deployments across Epic Hyperspace, Hyperdrive, and MyChart environments, I see the same structural mistakes delay production cutovers by months.</p>
<p>Here is what actually breaks in Epic’s OAuth2 layer and how to architect around it before a health system’s Technical Integration Engineer rejects your build.</p>
<h2>1. Requesting Coarse SMART v1 Scopes Under ONC HTI-1 and SMART v2</h2>
<p>If your Epic Showroom (formerly App Orchard) manifest still requests patient/Observation.read, expect immediate pushback from hospital InfoSec. In Epic’s data model, the Observation resource maps across laboratory results (LWS), vital signs, social history, and sensitive SDOH questionnaires.</p>
<p>Under the HL7 SMART App Launch v2.0.0 specification and ONC HTI-1 mandates [Citation: HL7 SMART App Launch IG v2.0.0; 45 CFR § 170.315(g)(10)], Epic enforces granular CRUD/S syntax (.c, .r, .u, .d, .s) and sub-resource category filters. If your clinical decision support module only evaluates lab panels, requesting blanket read access violates HIPAA’s Minimum Necessary standard</p>
<p>[SS: Epic Interconnect Token Exchange Wire Capture, Granular SMART v2 Scope Enforcement]</p>
<p><code>POST /interconnect-fhir-oauth/oauth2/token HTTP/1.1</code></p>
<p><code>Host:</code> <a href="http://fhir.epic-healthsystem.org"><code>fhir.epic-healthsystem.org</code></a></p>
<p><code>Content-Type: application/x-www-form-urlencoded</code></p>
<p><code>grant_type=authorization_code&amp;code=eyJ0eXAi...&amp;redirect_uri=https%3A%2F%</code><a href="http://2Fapp.vendor.io"><code>2Fapp.vendor.io</code></a><code>%2Fcb</code></p>
<p><code>HTTP/1.1 200 OK</code></p>
<p><code>{</code></p>
<p>  <code>"access_token": "eyJhbGciOiJSUzI1NiIs...",</code></p>
<p>  <code>"token_type": "Bearer",</code></p>
<p>  <code>"expires_in": 3600,</code></p>
<p>  <code>"scope": "launch patient/</code><a href="http://Observation.rs"><code>Observation.rs</code></a><code>?category=</code><a href="http://terminology.hl7.org/CodeSystem/observation-category%7Claboratory"><code>http://terminology.hl7.org/CodeSystem/observation-category|laboratory</code></a><code>",</code></p>
<p>  <code>"patient": "erXuFYUfucBZaryVksYEcMg3"</code></p>
<p><code>}</code></p>
<p>Drop legacy .read wildcards. Declare category-constrained .rs scopes so Epic Interconnect restricts token privileges at the gateway before querying Chronicles.</p>
<h2>2. The offline_access Trap in Provider Hyperdrive Launches</h2>
<p>Junior guides routinely tell developers to append offline_access whenever background syncing is needed. In a provider-facing EHR launch inside Epic Hyperdrive, that mistake will get your app blocked during security review.</p>
<p>When a physician launches your app from an active chart, the OAuth token is bound to that clinician’s EMP (User) and SER (Provider) Chronicles records. Hospital CISOs almost universally block offline_access on clinician-launched apps because they will not allow an external server to hold a perpetual refresh token acting on behalf of a logged-out physician. Furthermore, Epic access tokens default to a 3,600-second (1-hour) TTL—not 8 hours, as generic blogs often misstate.</p>
<p>The Architectural Fix: Keep your Hyperdrive EHR-launch session strictly interactive (launch + user/<em>.rs or patient/</em>.rs). If your platform requires asynchronous data ingestion after the clinician closes the chart, decouple the workflows: register a separate SMART Backend Services client ID using client_credentials and system/*.rs scopes authenticated via asymmetric JWT (private_key_jwt)</p>
<h2>3. Ignoring Epic’s Dual-Gate Authorization (OAuth Scope Chronicles Access)</h2>
<p>Passing OAuth2 scope validation at the Epic Interconnect gateway does not guarantee delivery of FHIR resources. Epic enforces a two-tier authorization check:</p>
<p>Outer Gate (OAuth2): Does the Bearer token contain user/DiagnosticReport.rs? Inner Gate (Chronicles Security): Does the authenticated EMP record possess the Epic Security Points and ORL (Department) access required to view that specific encounter, especially for "Break-the-Glass" (BTG) or restricted behavioral health charts?</p>
<p>[SS: Epic Interconnect 403 OperationOutcome When OAuth Passes but EMP Security Fails]</p>
<p><code>{</code></p>
<p>  <code>"resourceType": "OperationOutcome",</code></p>
<p>  <code>"issue": [{</code></p>
<p>    <code>"severity": "error",</code></p>
<p>    <code>"code": "forbidden",</code></p>
<p>    <code>"details": {</code></p>
<p>      <code>"coding": [{</code></p>
<p>        <code>"system": "urn:oid:1.2.840.114350.1.13.0.1.7.2.657369",</code></p>
<p>        <code>"code": "4101",</code></p>
<p>        <code>"display": "User lacks security point to access restricted encounter."</code></p>
<p>      <code>}]</code></p>
<p>    <code>}</code></p>
<p>  <code>}]</code></p>
<p><code>}</code></p>
<p>Never assume a granted OAuth scope guarantees payload access.</p>
<p>Your ingestion pipeline must gracefully handle Epic OID 1.2.840.114350 OperationOutcome exceptions without crashing the clinician's Hyperdrive frame when their EMP profile hits a restricted record.</p>
<h2>4. jwks_uri Caching Collisions and iss Discovery Failures</h2>
<p>Confidential apps authenticating to Epic cannot use symmetric client secrets in production; Epic requires RS384/RS256 signatures verified against a public JWK Set URL (jwks_uri) registered in Vendor Services.</p>
<p>Two operational oversights routinely break production cutovers:</p>
<p>Non-Overlapping Key Rotation: Epic Interconnect caches your jwks_uri payload. If you rotate your private signing key and immediately purge the prior public key from your JWK endpoint, token exchanges fail until Epic’s cache TTL expires. Publish the new kid (Key ID) in your JWK Set at least 48 hours before signing client_assertion JWTs with the new private key. Hardcoding Sandbox Endpoints: Every health system hosts its own Interconnect gateway. During an EHR launch, your app must dynamically extract the iss parameter, fetch {iss}/.well-known/smart-configuration, and validate the iss URI against an explicit allowlist of onboarded customer tenant endpoints before redirecting to /authorize.</p>
<h2>Pre-Cutover Verification Checklist for Your Customer’s Epic TIE</h2>
<p>Scope-to-API Parity: Every requested scope maps 1:1 to a listed API in your Showroom App Record; unused resources are stripped. Client Auth Alignment: Public SPAs/mobile clients enforce PKCE (code_challenge_method=S256); confidential backends pass RFC 7523 client_assertion JWTs with a unique jti nonce and exp 5 minutes. Context Separation: MyChart standalone launches exclusively request patient/* scopes; Hyperdrive EHR launches request launch and respect EMP session lifetimes. Audit Correlation: Outbound logs capture the Epic patient FHIR ID, encounter token claim, and OAuth sub user identifier for HIPAA security rules.</p>
<h2>Need an Architect to Vet Your Epic Showroom Manifest Before Security Review?</h2>
<p>Stop losing months to avoidable TSR rejections. Schedule a <a href="https://www.capminds.com/blog/5-vendors-to-evaluate-for-epic-and-oracle-health-fhir-integration/">Technical Architecture</a> Review with Our Epic FHIR Engineering Team</p>
]]></content:encoded></item></channel></rss>