Skip to content
Cloud Security Breaches

Over-permissioned identities: 13 incidents in the ledger

An over-permissioned identity is a user, role, service account or application that held more access than its job needed, so that the extra access, and not the way the identity was compromised, set how much data or infrastructure an incident reached and how long the intruder could keep moving.

Open the ledger · All eight paths

How it works

Providers state the intent plainly. Microsoft's documentation says least privilege means granting administrators exactly the permission they need to do their job, over a specific scope and for a specific time. AWS documentation describes a permissions boundary as a policy that sets the maximum permissions an identity-based policy can grant to a user or role. These are limits an owner has to apply, because identities are created with whatever access the person creating them grants.

In an incident, the identity that is compromised is often not the problem by itself. What the sources record is how far its permissions reached: a role that could read every bucket, or an account with administrative rights where a narrow one would do. AWS documentation also tells owners to find unused credentials and remove them, because access that is never used still counts when the identity is stolen.

The incidents

DisclosedOrganizationCloudHow access was gained
2025-08-26Nx (Nrwl)GitHubThe post-mortem says attackers exploited a GitHub Actions injection vulnerability to steal the npm publishing token; the repository's workflow permissions were read and write.
2025-07-23Amazon Web ServicesAWSThe bulletin says the extension had an inappropriately scoped GitHub token in its CodeBuild configuration, and with that token the threat actor committed malicious code.
2025-04-28Nova Scotia Powernot publicThe letter says an employee visited a compromised website and clicked a link, malware was installed, the threat actor moved across systems using accounts with domain administrator privileges, harvested credentials, took data between 23 and 25 April 2025, then used harvested credentials to destroy backups and deploy ransomware.
2024-08-30Verkada Inc.Verkada (cloud-based security camera provider)The FTC press release says the complaint alleges the company failed to provide appropriate security measures, including failing to require unique and complex passwords, adequately encrypt customer data and implement secure network controls. The Justice Department release says the complaint alleges failures in appropriate access management and data protection controls. Neither source states how the hacker obtained the access.
2024-05-01Dropboxnot publicThe company stated a third party gained access to an automated system configuration tool, compromised a back-end service account with privileges across the production environment, and reached the customer database; it answered how access was gained with a compromised access token.
2024-01-19MicrosoftMicrosoft 365Microsoft stated the threat actor used a password spray attack to compromise a legacy non-production test tenant account, then used that account's permissions to access corporate email accounts.
2023-09-18MicrosoftMicrosoft AzureMicrosoft stated an employee shared a URL for a blob store in a public GitHub repository, and the URL included an overly-permissive Shared Access Signature token for an internal storage account.
2023-08-16E-Commerce Enablers Pte. Ltd.AWSThe decision says an AWS access key with full administrative privileges was committed to a private GitHub repository on 4 June 2019, removed on 6 June but still viewable in the commit history, and not fully disabled when a replacement key was created. On 9 September 2020 a threat actor used it to reach the AWS environment, changed security settings to allow remote internet access to database instances and exfiltrated data. The company conjectured the key was found in the commit history.
2023-01-04CircleCICircleCIThe company stated that malware on an engineer's laptop stole a valid 2FA-backed SSO session, and that the employee held privileges to generate production access tokens.
2022-09-16UberGoogle WorkspaceThe company stated that an attacker likely bought a contractor's corporate password after malware on the contractor's personal device exposed it, then repeated two-factor login requests until the contractor accepted one.
2022-08-12Mailchimp (Intuit)MailchimpThe company stated an unauthorized actor accessed an internal tool used by customer-facing teams for customer support and account administration, after a social engineering attack on Mailchimp employees that obtained employee credentials.
2021-12-01A publicly traded wireless communications manufacturer (called Company-1 in the U.S. Attorney release)AWSThe release says an employee who was a senior developer held credentials for the company’s AWS and GitHub servers and repeatedly misused his administrative access to download the data; the company’s own account of an outside intruder was, the release says, false.
2020-07-16Blackbaud Inc.Blackbaud (software provider; the New York Attorney General calls it a cloud company)The FTC press release says that, according to the complaint, a hacker in early 2020 accessed a customer’s Blackbaud-hosted database and, once logged in, moved across multiple Blackbaud-hosted environments by using existing vulnerabilities and local administrator accounts and by creating new administrator accounts. The FTC says the company failed to segment data, adequately implement multifactor authentication and monitor for breach attempts. The company paid a ransom of 24 Bitcoin, according to the FTC.

The kind of control that addresses it

scope each identity to the actions and resources its task needs, cap delegated permissions with boundaries, and remove access that is unused

Sources

  1. Microsoft Learn: best practices for Microsoft Entra roles, read 16 Nov 2026
  2. AWS IAM User Guide: permissions boundaries, read 16 Nov 2026
  3. AWS IAM User Guide: find unused credentials, read 16 Nov 2026

Last reviewed 16 Nov 2026.