Logo
Cybersecurity

Your Cloud Vendor Has Access. What Happens When Its Credentials Are Compromised?

An attack involving Qubole's AWS integration credentials exposed a weakness in how organisations manage trusted third-party access. The incident shows why cloud security reviews must include the identities and permissions connecting external platforms to internal environments.

ITSEC AsiaITSEC Asia
|
Nov 11, 2026
Your Cloud Vendor Has Access. What Happens When Its Credentials Are Compromised?

On 9 November 2026, security researchers published findings about a supply-chain attack involving Qubole, a cloud data-processing platform. Attackers used compromised IAM credentials associated with the platform's AWS account to access multiple customer AWS environments. The investigation found that the activity took place between 14 and 26 October. [1]

The attackers exploited cross-account IAM roles that customers had configured to allow Qubole's platform to access their cloud resources. Once inside, they enumerated cloud infrastructure, searched for credentials and accessed data stored in Amazon S3 buckets. Some buckets were reportedly downloaded in their entirety. [1][2]

The activity went beyond one cloud service. Investigators found attempts to retrieve secrets from AWS Secrets Manager, Systems Manager Parameter Store and Lambda environment variables. Discovered credentials were then tested against AWS, Microsoft Entra ID and Okta services. This created opportunities to reach additional resources beyond the original cloud integration. [1]

The most revealing detail was that some customer environments still trusted Qubole's AWS account despite no longer actively using the integration. The trust relationships remained in place, allowing the attackers' role-assumption requests to succeed.

The incident illustrates a common cloud security problem: an integration that was approved for a legitimate business purpose can become an entry point when the identity operating it is compromised.

The Risk Behind Cross-Account Trust

Cross-account IAM access is a standard way to let an external platform work with customer resources without sharing long-term AWS access keys.

A customer creates an IAM role and allows a specified external AWS account to assume it. The vendor then uses the AWS Security Token Service (STS) AssumeRole operation to obtain temporary credentials with the permissions defined by that role.

The arrangement is practical, but its security depends on how the trust relationship is configured and maintained.

An External ID can help address the confused deputy problem, but AWS explicitly states that it should not be treated as a secret. It also cannot compensate for an overly broad trust policy or excessive permissions. [3]

If an external identity is compromised, the attacker may inherit whatever access the customer has granted to it.

This makes third-party access a business risk as much as an identity-management issue. A vendor may need permission to read a specific dataset, but that should not automatically mean it can retrieve unrelated application secrets or access every storage bucket.

What Organisations Should Review

The incident offers several practical checks for cloud and security teams:

  • Inventory external trust relationships. Identify every IAM role that permits access from an external account, including integrations that are no longer in use.

  • Apply least privilege. Restrict each role to the resources and actions required by its business purpose.

  • Narrow trusted identities. Where supported, trust specific vendor principals rather than an entire external account. Apply appropriate source-network conditions.

  • Monitor role assumptions. Review CloudTrail for unexpected AssumeRole activity, unfamiliar source networks and access from dormant integrations.

  • Protect downstream credentials. Limit permissions to Secrets Manager, Lambda configuration, Parameter Store and sensitive S3 data.

  • Review session exposure. Revoking an access key does not necessarily invalidate temporary credentials already issued. Investigate and revoke active sessions where necessary.

AWS recommends temporary credentials for workloads, least-privilege permissions and regular removal of unused roles and credentials. Its IAM Access Analyzer can also help identify unexpected public and cross-account access. [3][4]

Testing the Digital Attack Surface

Cloud security reviews need to consider how exposed applications, APIs, identities and network services interact. A vulnerability in one digital asset can provide a route toward resources that were never intended to be publicly reachable.

Authorised penetration testing can help organisations assess supported digital environments, identify reachable weaknesses and validate whether exposed services create additional attack paths.

Bronyx AI, ITSEC's AI-assisted autonomous penetration testing platform, supports authorised assessment of web applications, APIs and defined network environments. It can help security teams examine exposed digital assets and validate technical weaknesses within an approved testing scope.

It does not replace AWS IAM policy reviews, CloudTrail monitoring or cloud-native access analysis. Those controls remain necessary to identify excessive permissions and suspicious cross-account activity.

The question organisations should ask is straightforward: if a trusted vendor identity were compromised today, how much of the company's cloud environment could it reach?

Learn more about Bronyx AI at https://bronyx.ai/.

References

  1. Wiz Research, Trusted by Default: Qubole Cross-Account Access Abused in the Wild, 9 November 2026. https://www.wiz.io/blog/qubole-cross-account-access-abused-in-the-wild

  2. The Register, Attackers backdoor trusted AWS integrations to steal secrets, 9 November 2026. https://www.theregister.com/security/2026/11/09/attackers_backdoor_trusted_aws_integrations_to_steal_secrets/

  3. Amazon Web Services, Access to AWS accounts owned by third parties, AWS IAM documentation. https://docs.aws.amazon.com/IAM/latest/UserGuide/id_roles_common-scenarios_third-party.html

  4. Amazon Web Services, Security best practices in IAM, AWS IAM documentation. https://docs.aws.amazon.com/IAM/latest/UserGuide/best-practices.html

Share this post

You may also like

Protecting Critical Infrastructure Requires a Different Kind of Cybersecurity Training
Cybersecurity

Protecting Critical Infrastructure Requires a Different Kind of Cybersecurity Training

A suspicious connection appears in a corporate network. The security team investigates, isolates the affected system and starts remediation. In a conventional IT environment, that may be a reasonable first response. Now imagine the affected system controls part of an electricity substation, a water treatment facility or an industrial production line. Isolating it without understanding the operational consequences could disrupt a process that needs to remain stable and safe. This is one of the defining challenges of operational technology (OT) security. NIST's Guide to Operational Technology Security explains that OT systems interact with the physical environment and operate under specific performance, reliability and safety requirements. Security decisions must account for those requirements alongside confidentiality and integrity. [https://www.google.com/s2/favicons?domain=https://www.nist.gov&sz=32] NIST +1 For cybersecurity workforce development, that changes the training equation. KNOWING THE NETWORK ISN'T ENOUGH OT environments include industrial control systems, supervisory control and data acquisition (SCADA), building automation and other systems that monitor or control physical processes. Their operating requirements can differ considerably from those of ordinary business IT. [https://www.google.com/s2/favicons?domain=https://www.nist.gov&sz=32] NIST A security practitioner working in this environment needs

ITSEC AsiaITSEC Asia
|
Okt 13, 2026 — 4 minutes read
Cybersecurity Training Works Best When Employees Know What to Do Next
Cybersecurity

Cybersecurity Training Works Best When Employees Know What to Do Next

Most employees have seen a phishing warning. They know to look for unusual links, unexpected attachments and messages that create urgency. Yet when a convincing message arrives during a busy workday, remembering the advice and acting on it are two different things. A message may look routine, appear to come from a senior colleague or arrive just as someone is rushing to meet a deadline. That gap between awareness and action is worth addressing. In its 30 September 2026 release, the European Commission reported that 74% of EU employees had encountered suspicious emails, messages or links at work in the previous six months. Phishing was the most commonly reported threat, affecting 39% of respondents. The survey also found that only 48% felt able to recognize an AI-generated fake video. ENISA's summary of the findings adds another important detail: only 45% of employees said their organization provided regular cybersecurity awareness updates, while 85% expressed interest in improving their skills. Lack of time was the most frequently reported barrier to training, cited

ITSEC AsiaITSEC Asia
|
Okt 27, 2026 — 4 minutes read
Teaching People to Use AI Without Security Is Half a Lesson
Cybersecurity

Teaching People to Use AI Without Security Is Half a Lesson

Seven hundred and fifty government employees across Kalimantan started AI training this week. The program, run by BLSDM Komdigi Banjarmasin with ASEAN Foundation, covers AI fundamentals, ethics, implementation and something that deserves to sit comfortably beside all three: data security. That combination makes sense. Organizations are teaching more people how to use AI because AI is becoming part of ordinary work. Employees can summarize documents, analyse information, draft material and automate routine tasks without writing a line of code. Every new capability also creates a new question: what exactly are we giving the AI? THE AI SKILL GAP HAS A SECURITY SIDE An employee can be perfectly competent at prompting an AI system while making poor security decisions around it. Imagine someone needs to summarize a lengthy internal report. Uploading the document may produce an excellent summary. Whether the document should have been uploaded in the first place is an entirely different test. As AI becomes commonplace, workers need enough security literacy to ask questions such as: * What information am I

ITSEC AsiaITSEC Asia
|
Sep 08, 2026 — 3 minutes read

Receive weekly
updates on new posts

Subscribe