Skip to content
Cloud Security Breaches

30 questions, answered from the data

Answers to common questions about the Cloud Breach Ledger: which sources it accepts, what not public means, how the eight access paths are assigned, how counts are computed and how to report an error. Each answer is short, names its source where one applies and gives the read date.

Open the ledger ยท Read the method

What is the Cloud Breach Ledger?
The Cloud Breach Ledger is a record of public cloud incidents from 2019 to 16 November 2026, with each entry read by the access path that was used. Every field comes from a company statement, regulator filing, court document or provider post-mortem. A field with no such source reads "not public".
Which sources does the ledger accept?
Company statements (press releases, blog posts, customer notices, SEC filings by the company such as an 8-K Item 1.05), regulator filings and decisions, court documents and provider post-mortems. News articles, vendor research blogs and Wikipedia are not accepted as sources. They may point to a primary source, nothing more.
What does "not public" mean in an entry?
It means no primary source states that fact as of the read date, 16 November 2026. It does not mean the fact is secret or unknown to the company. If a company later publishes it, the field changes and the correction is logged.
How many access paths does the ledger use?
Eight: leaked access keys, stolen user credentials, over-permissioned identities, public resources, cross-tenant flaws, third-party access, exposed network services and retained access. Each entry has one primary path and, only where the sources describe a second route, one secondary path.
How is an incident assigned to an access path?
The path follows how access was gained, in the words of the primary source. Where the source does not say how, the entry is assigned by what the source does say, and the access field reads "not public" for the rest. No path is inferred from rumor or from a news account.
Does the ledger rank companies or providers?
No. Entries are listed by date and filtered by path, provider or year. Counts by provider reflect how many incidents have a primary source that names that provider, so a higher count says nothing about which provider is safer.
Why does the ledger not name attackers or employees?
The ledger records organizations only. Where a company's own statement uses a group name, the ledger quotes it as the company's label. Otherwise it writes "an individual" or "a former employee", and it makes no claim about motive or blame.
Can an incident be in the ledger without a named cloud provider?
Yes, if the primary source describes cloud-hosted data or a cloud service but the provider field has no public statement behind it. The provider then reads "not public". The entry still qualifies when the source places the incident in a cloud or software-as-a-service tenant.
How are counts on the ledger computed?
Each count is the number of entries in the ledger that match a filter, for example one access path or one year of publication, on the edition date. They are counts of sourced incidents, not estimates of how many incidents happened. Entries added later change the next edition's counts.
What is the difference between a ledger count and a quoted statistic?
A ledger count is built from the entries on this site. A quoted statistic is a figure published by a third party, such as an annual breach report, and is shown with its publisher, year and link. The two are never added together or compared as if they measured the same thing.
How do I report an error in an entry?
Every entry has a corrections link to editor@cloudsecuritybreaches.com with the entry name in the subject. Send the primary source that shows the correct fact. A correction is accepted when a company statement, regulator filing, court document or provider post-mortem supports it.
How often is the ledger updated?
The ledger is published in dated editions, and this is the edition of 16 November 2026. The news page lists primary-source items dated 18 August 2026 onward. Each page shows the date its sources were read.
What is a leaked access key?
A leaked access key is a long-term key or token that ended up in a place others could read, such as a repository or a log, and was used. AWS's documentation says access keys are long-term credentials, and recommends temporary credentials in their place (read 16 Nov 2026).
What is the safer alternative to a long-term access key?
Temporary credentials that expire. AWS's IAM best practices page recommends that workloads use temporary credentials through IAM roles, and Google Cloud's documentation recommends choosing a more secure alternative to service account keys wherever possible (read 16 Nov 2026).
Why does a stolen password matter if the account has multi-factor authentication?
Because some second factors can be bypassed, and a stolen session token needs no second factor at all. Google Cloud's M-Trends 2026 post says threat actors harvest long-lived OAuth tokens and session cookies (read 16 Nov 2026). The control that applies is phishing-resistant sign-in and short session lifetimes.
What is an over-permissioned identity in an incident?
It is an identity whose access was wider than its job, so the extra access set how far the incident reached. The ledger uses this path when the primary source ties the scale of exposure to what the identity could reach, not only to how it was compromised.
How does a public resource differ from a leaked key?
A public resource needs no credential at all: anyone who finds the address can read it. A leaked key is a credential that someone used. Amazon S3's Block Public Access documentation says new buckets do not allow public access by default (read 16 Nov 2026).
What does cross-tenant mean in cloud incidents?
It means access crossed the boundary between two customers of the same provider. The cause can be a flaw in the provider's service or a setting in a customer's application. Microsoft's tenant documentation explains how single and multitenant apps are configured (read 16 Nov 2026).
What counts as third-party access?
Access that arrived through a vendor, supplier, integration or contractor that the company had trusted. Examples in the ledger include incidents that began with tokens held by an integration. AWS's confused deputy documentation covers how third-party roles should be limited (read 16 Nov 2026).
What is an exposed network service?
A reachable service or request path that gave a caller a way into a workload, such as an open management port or a forged request to a metadata endpoint. AWS's security blog describes metadata service enhancements against SSRF (read 16 Nov 2026).
What is retained access?
Access that a former employee, contractor or application kept after the reason for it ended. The ledger uses this path only when a primary source, such as a court document, says that a person kept access. It names no individual.
Does the ledger include vulnerabilities in cloud workloads?
Only when a primary source ties an incident to cloud data or a cloud tenant. A vulnerability alone is not an incident, and scanning for them is outside this ledger.
Where do the research techniques come from?
From vendors' and researchers' own publications, each logged with the publisher's name, the publication date and the link. A technique entry describes disclosed research. It is not an incident and it is not counted in the ledger.
What does the ledger say about Act's ReplicaHeist research?
It is logged as one technique entry like any other: published 3 November 2026 by act.security research, read 16 November 2026, with the link to the publication. The ledger summarizes it in two neutral sentences and does not rate it.
Are the quoted statistics the ledger's own?
No. They are third-party figures such as those in IBM's Cost of a Data Breach reports, shown as stated with year and link. The ledger does not recompute or adjust them.
Why do third-party statistics and ledger counts differ?
Surveys count what respondents report or what a vendor's customers experienced. The ledger counts only incidents with a primary source. A report may include thousands of cases the ledger cannot source, so the two measure different things.
Does a regulator action prove how an incident happened?
Only to the extent the regulator's document says so. The ledger quotes the filing for what it states, for example a consent order's findings, and records the rest as not public. A penalty amount says nothing about the access path.
How is a company's 8-K treated?
As a company statement. A Form 8-K under Item 1.05 describes a material cybersecurity incident as the company understood it on the filing date. Later amendments may change facts, so the ledger reads the amendments and shows each source's date.
Can a company ask for an entry to be removed?
An entry stays while its primary sources stay public. A company can send a correction with a primary source to editor@cloudsecuritybreaches.com. The ledger changes a fact when a primary source supports the change, and it does not remove a sourced incident on request.
Can I cite the ledger?
Yes. Use the line on the methodology page: Cloud Security Breaches, the Cloud Breach Ledger, edition 16 Nov 2026, https://cloudsecuritybreaches.com/methodology/. Cite the primary source as well when you rely on one entry.

Last reviewed 16 Nov 2026.