Explore our new Cybersecurity Terms Glossary, explained by our experts.

View the Glossary View the Glossary
PRIVIA

Cybersecurity Terms Glossary

What Happened in the Okta Support Breach?

In October 2023 an attacker reached Okta’s customer support system and read session tokens inside uploaded diagnostic files, with customers detecting the abuse before the vendor did.

Illustration for the cybersecurity term Okta Support Breach
Home / Blog / Cybersecurity Terms / What Happened in the Okta Support Breach?
17 November 2026

What Happened in the Okta Support Breach?

What Happened in the Okta Support Breach?

On 20 October 2023, Okta disclosed that an attacker had used stolen credentials to access its customer support case management system. The system was separate from the production identity service that customers authenticate against, and Okta was careful to say so. The distinction mattered less than it sounds, because of what customers routinely upload to support tickets.

When an administrator raises a ticket about a failing login, support engineers commonly ask for a HAR file: a recording of the browser’s network activity during the problem. A HAR file captures request and response headers verbatim, which means it captures session cookies and bearer tokens. An attacker reading those files obtains valid, live sessions for the customer’s own Okta tenant without ever touching a password or a second factor.

The Sequence

The three-week gap between the first customer report and public confirmation is the part security teams remember. Two of the world’s better-resourced security organisations detected the abuse of their own tenants before the identity provider detected the source.

The Root Cause

Okta’s own analysis identified the entry point as a service account whose credentials had been saved into an employee’s personal Google account, signed into Chrome on a managed laptop. The compromise of that personal account exposed the service credential, and the service account held permission to view support cases and their attachments.

Three ordinary failures combined: a service account used interactively, a browser profile mixing personal and corporate identity on a managed device, and a credential stored in a browser password manager rather than a controlled vault. None of them is exotic, and all three are present in most organisations.

Why Identity Providers Are a Special Case

An identity provider is a concentration of trust. Compromise of a single tenant administrator session can reach every application behind single sign-on, which is why the same vendor has featured in several significant incidents, including the Lapsus$ intrusion at a support contractor in January 2022 and the help desk social engineering that preceded the MGM Resorts attack.

The pattern recurs because attackers correctly identify that the support and administration surfaces around an identity provider are often weaker than the authentication service itself.

What Customers Should Change

  1. Sanitise diagnostic files. Strip cookie and authorisation headers from any HAR file before upload. Okta published guidance and tooling for this after the incident, and the same applies to any vendor asking for browser traces.
  2. Bind sessions where you can. Administrator session policies that tie a session to an IP address or network zone turn a stolen cookie into a useless one. Okta already supported this; the affected customers largely had not enabled it.
  3. Separate administrator identity. Use dedicated accounts on dedicated devices for identity provider administration, with phishing-resistant multi-factor authentication.
  4. Monitor the tenant’s own logs. BeyondTrust caught this because it alerted on an administrator session originating from an unexpected location. The System Log is the evidence source; treat it as a SIEM feed, not a support tool.
  5. Block personal profile sign-in on managed browsers. Enterprise browser policy can prevent the exact configuration that caused the breach.

The Supplier Assurance Lesson

For UK organisations, the incident is a clean illustration of why supplier security questionnaires are insufficient. The compromised component was not the service under contract; it was the support system attached to it. Due diligence should ask specifically what a vendor’s support staff can see, how long attachments are retained, and how quickly the vendor commits to notifying customers. Under UK GDPR Article 28, a processor must notify the controller of a personal data breach without undue delay, and a three-week investigation period is worth discussing before signing rather than afterwards.

We examine exactly these dependencies during Cyber Risk Analysis Services.

FAQ – Frequently Asked Questions

What Is a HAR File and Why Was It Dangerous?

An HTTP Archive file is a JSON recording of everything a browser sent and received during a session, produced from developer tools and commonly requested by support teams to diagnose login problems. It preserves request headers in full, including session cookies and bearer tokens. Anyone who reads the file can replay those credentials and assume the session, which is exactly what happened here.

Did the Breach Affect Okta’s Production Authentication Service?

No. The attacker accessed the customer support case management system, which is separate from the service that processes authentication. The practical impact nonetheless reached customer tenants, because the support system held uploaded diagnostic files containing live session tokens for those tenants. A breach of a supporting system can produce the same outcome as a breach of the core product.

How Did Customers Detect It Before the Vendor?

By monitoring their own identity provider logs. BeyondTrust alerted on an administrator session that appeared from an unexpected location shortly after a HAR file was uploaded, investigated within half an hour, and blocked the activity. The lesson is that identity provider system logs belong in your SIEM with alerting on administrative actions, not left as a vendor-side record you consult after an incident.

What Should We Do Before Uploading Diagnostics to Any Vendor?

Treat every diagnostic capture as a credential file until proven otherwise. Remove cookie, authorisation, and set-cookie headers from HAR files using a sanitiser, redact tokens from log bundles, and avoid capturing traces while authenticated as a privileged account where a test account would do. Record what was shared, with whom, and when, so exposure can be assessed quickly if the vendor is later breached.

Related Terms

You May Be Interested In These

What Is AS-REP Roasting?

Cybersecurity Terms

17 November 2026

What Is AS-REP Roasting?

Read More
What Is the ISA-95 Model?

Cybersecurity Terms

17 November 2026

What Is the ISA-95 Model?

Read More
What Is Deception Technology?

Cybersecurity Terms

17 November 2026

What Is Deception Technology?

Read More