27 Organisations, No MFA Prompt: Inside the Stolen Entra ID Credential Campaign
When non-human identities are breached, MFA will not save you. Here is how attackers pivoted through trusted app credentials to reach Microsoft cloud resources and MSP customer tenants and how to lock down your tenant.
Entra ID Workload Identity Security concerns the identities used by applications, scripts, services and deployment pipelines. They do not sign in like employees, yet they can read files, retrieve secrets and change cloud resources. Their permissions often outlive the project that created them. Non-human identity management must track those changes, not just the application’s handover.
Todyl’s investigation describes a November 2026 campaign in which stolen Microsoft Entra ID application credentials were used to reach 27 organisations, including MSP customers. The app-only sign-ins did not ask an employee to complete MFA because no user was involved. There was no Microsoft vulnerability to patch; the applications were accepted because the credentials were valid.
Todyl first saw the activity on 1 November: about 45 app registrations across 18 organisations were targeted, and more than 33 authenticated successfully. One multi-tenant MSP application reached 15 customer tenants. Microsoft’s threat-intelligence report places successful service-principal sign-ins from the main address as early as 28 October. These are observed figures, not a complete victim count. Entra ID Workload Identity Security matters because one application’s access extended beyond its owner.
How the attackers got in and what they accessed
The actor signed in as applications, not employees. Todyl observed stolen app credentials, an Azure DevOps federated workload identity and replayed user refresh tokens. With app-only authentication, no interactive user is present to complete MFA. A policy written for employee sign-ins will not automatically cover that route.
Once authenticated, the applications requested tokens for Azure Resource Manager, Key Vault, Microsoft Graph and Storage. The actor searched Microsoft Graph and downloaded SharePoint files that could contain passwords, private keys, VPN profiles, backup credentials and remote-access details. A compromised MSP application made the problem bigger: its permissions provided a way into customer environments as well as the provider’s own tenant.
Microsoft also reported that, since August 2026, agentic tooling had been searching public GitHub repositories for exposed credentials and AI-service keys. In one case, an agent moved from finding a secret to taking data from Azure Key Vault in five minutes. Microsoft linked the later Entra activity to the same broader operation; Todyl described it as an automated credential crawler in its 6 November update.
Why service principals matter to Entra ID Workload Identity Security
Start with the logs. Microsoft’s main indicator for the service-principal activity was 199.19.77.84. It also reported 199.19.77.18, but warned that this Tor exit node may be linked to unrelated activity. Use the indicators alongside the behaviour in the advisories and search back to 28 October, where log retention allows. These checks are Entra ID Workload Identity Security in practice: connect the sign-in to the activity that followed it.
Review application sign-ins. Check service-principal and non-interactive sign-in logs, including failures. Pay attention to unfamiliar source addresses, dormant applications that suddenly become active and a failed credential attempt followed shortly by a successful one.
Follow the token trail. Look for one application requesting access to Azure Resource Manager, Key Vault, Microsoft Graph and Storage within a short period. Then check Key Vault reads, SharePoint or Storage downloads, permission changes and access to resources the application does not normally use.
Audit Azure DevOps. Review service connections, federated credentials, PAT creation and first use, pipeline edits, repository changes and unusual build activity. Preserve the logs before making changes that could remove evidence.
Contain and rotate. Remove compromised secrets, certificates and federated credentials from affected applications; disable a service principal where necessary; and revoke affected user sessions, PATs and SSH keys. Rotate every credential the application could read. Deleting a file from a repository or SharePoint does not make the exposed credential unusable.
Microsoft also recommends blocking the exact 199.19.77.84/32 address through an eligible workload-identity policy, rather than blocking a wider hosting network. MSPs should review Partner Center access, GDAP relationships and each downstream customer tenant.
What security teams should check
Start with sign-in and audit logs, searching as far back as retention allows. Microsoft identified 199.19.77.84 as the main source address. It also mentioned 199.19.77.18, a Tor exit node that may be unrelated. Treat that second address with care, and do not block an entire hosting network based on one indicator.
A service principal security review should connect each sign-in to what the application accessed afterwards. Check four areas:
- Application sign-ins: Review successful and failed sign-ins. Look for unfamiliar source addresses, dormant apps that suddenly become active, and failed credential attempts followed by successful sign-ins.
- Cloud activity: Look for one application requesting tokens for Azure Resource Manager, Key Vault, Graph and Storage in a short period. Check secret reads, downloads, permission changes and access outside its normal pattern.
- Azure DevOps security: Review service connections, federated credentials, pipeline edits, personal access token creation, repository changes and unfamiliar build activity. Check build logs and command-line arguments for exposed secrets. Preserve evidence before disabling resources.
- Credential response: Remove compromised secrets, certificates and federated credentials. Disable affected service principals where necessary, revoke relevant sessions and tokens, and rotate every credential the application could read. Deleting an exposed file does not invalidate it.
Conditional Access for workload identities can block sign-ins from untrusted locations or risk conditions, but it does not cover every machine identity. Microsoft requires Workload ID Premium, and policies must be assigned directly to eligible single-tenant service principals. Managed identities and multitenant applications are outside the documented scope. Read Microsoft’s guidance before relying on this control. It is one part of Entra ID Workload Identity Security, not a replacement for permissions reviews or credential protection.
Use managed identities for Azure-hosted workloads where supported, and federation for external workloads where available. If a client secret is still required, keep it in an approved store, restrict retrieval, record its owner and expiry, and test the replacement procedure. That is application credential security: know where a credential is used and replace it without breaking the service.
Where CyberArk fits
CyberArk Secrets Manager can give applications and automation a controlled way to retrieve secrets at runtime, rather than leaving values in source code, pipeline files or shared documents. For application secrets management, the benefit is practical: fewer copies of a credential and tighter control over which workload can retrieve it.
Teams can automate client secret rotation where the application and integration support it. CyberArk Central Policy Manager (CPM) rotates credentials for supported target types. For an Entra client secret or custom application, confirm the integration and how the new value will reach every consumer. The job is finished only when the application uses the replacement and the old credential is retired. Some environments will need a custom plugin or API workflow, tested before an incident occurs.
CyberArk Privileged Session Manager (PSM) can also isolate and record supported administrator sessions. That helps protect the people managing the environment and gives investigators a record of privileged activity. It is a separate control from service-principal sign-ins, so it should not be presented as a direct fix for app-only authentication.
Inventory Owners, App IDs, Service Principals And Trusts
Least Privilege, Narrow Scopes, Short-Lived Elevation
Vault Retrieval, Credential Rotation And Safe Delivery
Protect Repos, Service Connections, Approvals And Build Agents
Sign-In Telemetry, API Activity, Alerts And Response Runbooks
Where CyberArk fits into Entra ID Workload Identity Security
CyberArk Secrets Manager can give applications and automation a controlled way to retrieve secrets at runtime, rather than leaving values in source code, pipeline files or shared documents. For application secrets management, the benefit is practical: fewer copies of a credential and tighter control over which workload can retrieve it.
Teams can automate client secret rotation where the application and integration support it. CyberArk Central Policy Manager (CPM) rotates credentials for supported target types. For an Entra client secret or custom application, confirm the integration and how the new value will reach every consumer. The job is finished only when the application uses the replacement and the old credential is retired. Some environments will need a custom plugin or API workflow, tested before an incident occurs.
CyberArk Privileged Session Manager (PSM) can also isolate and record supported administrator sessions. That helps protect the people managing the environment and gives investigators a record of privileged activity. It is a separate control from service-principal sign-ins, so it should not be presented as a direct fix for app-only authentication
How Hassium Cyber Solutions can help
Hassium Cyber Solutions provides CyberArk implementation services and ongoing IAM/PAM engineering for Vault, PVWA, CPM and PSM deployments, upgrades and migrations. The team has 15+ security specialists, 20+ PAM certifications and engineers with 6 to 16 years of PAM experience, including CyberArk Certified Delivery Engineers. Projects can include application onboarding, custom CPM integrations, reconciliation and platform health checks.
For organisations reviewing privileged accounts and cloud access, see our Privileged Access Management services. Where additional engineering capacity is needed, our IAM staff augmentation team can support implementation and operations.
Hassium’s Vaulty AI is also being developed for environments where AI agents interact with privileged tools through MCP servers. It places policy and approval checks between an agent’s proposed action and execution; credentials are not passed to the model.
Could one exposed credential put your cloud environment at risk?
Frequently Asked Questions (FAQs)
Why did the stolen application credentials not trigger MFA?
The applications used app-only authentication. No employee was signing in, so there was no user to complete an MFA challenge. User-focused Conditional Access policies do not automatically cover service-principal sign-ins.
Does Conditional Access cover every workload identity?
No. Microsoft’s workload-identity policies have licensing and scope limits. They cover eligible single-tenant service principals, not managed identities or third-party multitenant applications. Check the current Microsoft documentation for the exact requirements.
Can CyberArk automatically rotate every Entra client secret?
Not without checking the target and integration. CPM supports defined target types; an Entra client secret may need a supported integration, custom plugin or API workflow. Test the replacement process before relying on it.
What should an MSP review first?
Review service principals, delegated relationships, Partner Center and GDAP access across customer tenants. Preserve logs and rotate credentials the application could access.
An application does not need an administrator’s password to cause serious damage; it needs a valid identity with more access than its job requires. Entra ID Workload Identity Security starts with knowing which applications have that access, who owns them and how quickly it can be removed. Good service principal security depends on clear ownership and a tested response. For help reviewing service principals, application secrets or a CyberArk environment, contact Hassium Cyber Solutions at admin@hassiumsolutions.com or +91 86889 12225.
