Recursos · 41

Divulgação coordenada de vulnerabilidades: receber, corrigir e informar.

Criar um canal seguro para relatórios externos, avaliar o impacto, coordenar as partes afetadas e publicar orientações úteis.

Atualizado · 8 min

Laboratório cibernético fictício sem pessoas, com malas e painel de rede numa bancada escura sob luz âmbar. Ilustração editorial gerada com IA

O que este guia ajuda a alcançar

  • Tornar o reporte possível e seguro.
  • Atribuir um responsável a cada relatório.
  • Coordenar uma correção verificável.
  • Fornecer aos usuários orientações práticas.

Verificação rápida

  • É possível enviar um relatório sem uma conta de cliente?
  • O escopo e os testes permitidos estão claros?
  • Quem responde se o contato principal estiver ausente?
  • Quais produtos e fornecedores compartilham a falha?
  • Qual mitigação provisória é útil?
  • O aviso identifica as versões realmente afetadas?

Método passo a passo

  1. 1

    Publicar uma política clara

    Indicar o escopo, os testes a serem evitados, o canal de contato, detalhes úteis do relatório e o processo de resposta. Solicitar que a política seja revisada para o seu contexto.

    Entregável: política publicada e canal monitorado.

  2. 2

    Reconhecer e proteger as evidências

    Confirmar o recebimento, atribuir um identificador, limitar a circulação de detalhes confidenciais e solicitar apenas as evidências necessárias para reproduzir o problema.

    Entregável: relatório datado com responsável.

  3. 3

    Avaliar e expandir o escopo

    Reproduzir sem prejudicar os sistemas. Examinar a explorabilidade, os ativos, as versões, os dados e os componentes compartilhados com outros produtos ou fornecedores.

    Entregável: avaliação de impacto e mapeamento de stakeholders.

  4. 4

    Coordenar a remediação.

    Manter diálogo regular com o autor da denúncia e os responsáveis ​​pela manutenção. Acordar marcos de validação e adaptar a comunicação caso haja mudanças nos riscos ou na disseminação.

    Entregável: cronograma compartilhado, correções e testes de regressão.

  5. 5

    Publicar e acompanhar.

    Explicar as versões afetadas, correções ou mitigações, ações do usuário e incertezas. Verificar a implantação e atualizar o aviso conforme os fatos mudem.

    Entregável: aviso versionado e evidências de distribuição.

Exemplo prático fictício

Situação ilustrativa

Um pesquisador relata uma falha no controle de acesso em uma versão antiga da API. A equipe reconhece o problema, reproduz o caso em um ambiente autorizado e constata que um componente compartilhado afeta dois produtos.

Decisão e evidências esperadas

O plano conecta os responsáveis ​​pela manutenção, as versões, os testes de correção e uma mitigação provisória. O aviso final fornece aos usuários uma ação clara e credita o pesquisador conforme acordado.

Indicadores de gestão

IndicadorO que medePrimeira ação
Recibo rastreadoRelatórios com responsável e data da primeira respostaCorrigir canais silenciosos
Escopo estabelecidoProdutos, versões e dependências afetados confirmadosExpandir a busca antes do aviso final
Correção verificadaCasos originais e relacionados reproduzidos em versões lançadasNão fechar uma variante vulnerável
Usuários informadosOrientações chegam ao público afetadoAtualizar canais e traduções necessários

Erros comuns

  • Prometer um prazo único, independentemente do risco
  • Solicitar dados pessoais excessivos do denunciante
  • Tratar o reconhecimento como remediação
  • Publicar evidências detalhadas de exploração antes de uma mitigação viável
  • Esquecer versões incorporadas por parceiros

Perguntas frequentes

Uma política de divulgação é uma recompensa por bugs?

Não. Uma política explica como relatar e coordenar. Recompensas precisam de regras e recursos separados.

A comunicação deve sempre esperar por uma correção?

Não. O momento depende do risco, das medidas provisórias e das partes afetadas. Registre e revise a decisão.

Quando um coordenador deve ajudar?

Um coordenador pode ajudar quando vários fornecedores são afetados, um contato não responde ou um cronograma compartilhado é difícil.

Referências: dados, usos e limites

Documentação consultada em .

As notas detalhadas desta edição estão disponíveis em inglês.

Referências: dados, usos e limites
Fonte e finalidadeDado ou referência verificadaComo utilizarLimite a preservar
CVE Program

Link discussions to the same vulnerability record.

A JSON schema defines the CVE record format. An identifier and the published contents of its record are distinct things. Open the record and references; check its state, described affected versions and supplier information. Keep the identifier in action tracking.

Aprofundar o método →

A CVE identifier is neither a severity score nor proof that your installation is affected.

security.txt e divulgação

As notas detalhadas desta edição estão disponíveis em inglês.

security.txt: find a reporting channel and check its context

A researcher may find a flaw without finding the right recipient. A security.txt file makes reporting channels and disclosure policies easier to discover. Its value then depends on monitored contacts, current information and a process that connects the report to the affected product.

This reference explains the format described in RFC 9116, an IETF Informational publication dated April 2022. The eight fields below are those defined by that RFC, rather than an exhaustive list of later extensions. Examples are fictional, using .example domains reserved for documentation. This page retrieves, validates, signs and publishes no real security.txt file.

Laboratório cibernético fictício sem pessoas, com malas e painel de rede numa bancada escura sob luz âmbar. Ilustração editorial gerada com IA
A fictional empty disclosure coordination laboratory. This scene depicts neither an actual facility nor evidence of security.

Six discovery and publication checks

Six discovery and publication checks
CheckRFC 9116 reference pointUseful record
LocationFor a web service, publish at /.well-known/security.txt. If both root and well-known files exist, use the well-known version.Retrieved URL, response and any difference from /security.txt at the root.
TransportRetrieve using HTTPS with HTTP 1.0 or later; referenced web URIs also use HTTPS.Certificate check and retrieval result, separately from checks of the file’s fields.
FormatUTF-8 plain text; Content-Type text/plain with charset=utf-8. Each field occupies its own line.Received header, observed encoding and a dated copy of the text examined.
ScopeThe file applies to the domain or IP address in its retrieval URI; parent domains and subdomains are not inherited automatically.Exact host, affected product and scope explained in the linked policy.
RedirectsA redirect may be legitimate but needs further inspection. The RFC recommends keeping a record.Redirect chain, host changes and the reason for using an external provider.
MaintenanceExpires is mandatory. Contacts, links and keys need to remain accessible and controlled by the organisation.Publication owner, verification date and a local review deadline before expiry.

Eight fields defined by RFC 9116

Eight fields defined by RFC 9116
FieldPresence and repetitionUseful interpretationPitfall
ContactRequired; one or more occurrences.Contact URI, such as mailto: or an HTTPS page. Occurrences should follow the publisher’s preference order.An email address without mailto: is not a conforming Contact URI. Publishing a channel does not establish that it is monitored.
ExpiresRequired; exactly one occurrence.RFC 3339 timestamp after which the data is stale and should not be used. A date less than one year ahead is recommended.That recommendation is neither a mandatory one-year duration nor a vulnerability remediation deadline.
CanonicalOptional; may repeat.Canonical URIs where the file is published. If the retrieval URI is absent from these fields, the RFC recommends not trusting the contents.Listing several Canonical URIs does not authenticate every listed host. An additional trust mechanism remains useful.
EncryptionOptional; may repeat.URI locating an encryption key; the key itself must not appear in this field.The referenced key is not necessarily the file’s signing key. Its presence proves neither authenticity nor that a report was encrypted.
PolicyOptional; may repeat.Link to the disclosure policy: scope, reporting practices and conditions explained by the organisation.The file does not replace that policy or by itself define a reward, response deadline or testing permission.
Preferred-LanguagesOptional; at most one occurrence, containing at least one language.Comma-separated language tags with equal priority regardless of order. If absent, English may be assumed.fr, en does not prioritise French. Several Preferred-Languages lines do not replace a single list.
AcknowledgmentsOptional; may repeat.Link to a page recognising people who helped remediate vulnerabilities.Coordinate public credit; it does not authorise premature disclosure of flaw details.
HiringOptional; may repeat.Link to the organisation’s security-related vacancies.Do not confuse a hiring link with the reporting channel.

Annotated fictional example: contact, time, origin and policy

Training reading at the fictional instant 22 October 2026, 10:00 UTC. The two Contact occurrences are ordered; the two languages have equal priority. Expiry is six months later. This is not Cyber Intelligence Embassy’s security.txt file and provides no real reporting channel.

# Fictional training example, not this site's reporting policy
Contact: mailto:security@publisher.example
Contact: https://publisher.example/report
Expires: 2027-04-22T10:00:00Z
Canonical: https://publisher.example/.well-known/security.txt
Policy: https://publisher.example/disclosure-policy
Preferred-Languages: fr, en

This example remains unsigned. Canonical identifies the expected URI but does not replace authenticity evidence. Before sending a sensitive report, verify the recipient, policy and appropriate transmission protections.

Eight situations separating format, freshness and trust

Each excerpt isolates one question; no conclusion is a complete file validation. Observations and proposed next steps are editorial guidance. Addresses shown in excerpts are not links to contact.

Essential fields are present

Fictional retrieval URI : https://publisher.example/.well-known/security.txt

Contact: mailto:security@publisher.example
Expires: 2027-04-22T10:00:00Z
Canonical: https://publisher.example/.well-known/security.txt

A contact, future expiry and the expected canonical URI appear in this excerpt.

Continue checking transport, recipient identity and policy. These fields prove neither mailbox operation nor testing permission.

Contact contains a bare email address
Contact: security@publisher.example
Expires: 2027-04-22T10:00:00Z

The mailto: scheme is missing; the value lacks the URI form required for Contact.

Correct the source publication. A tool’s possible tolerance does not establish format conformity.

Expires occurs twice
Contact: mailto:security@publisher.example
Expires: 2027-01-22T10:00:00Z
Expires: 2027-04-22T10:00:00Z

Two occurrences conflict with the Expires uniqueness rule.

Publish one expiry after correcting the file. Silently choosing the more convenient date conceals the publication defect.

The expiry date has passed
Contact: mailto:security@publisher.example
Expires: 2026-10-21T10:00:00Z

At the fictional evaluation instant the file expired twenty-four hours earlier.

Treat its data as stale and seek a verified current channel. Retain the old copy as history without presenting it as a current contact.

Canonical omits the retrieval URI

Fictional retrieval URI : https://api.publisher.example/.well-known/security.txt

Contact: mailto:security@publisher.example
Expires: 2027-04-22T10:00:00Z
Canonical: https://publisher.example/.well-known/security.txt

The file retrieved for api.publisher.example lists only publisher.example in Canonical.

Apply the recommendation not to trust these contents in this context and request publication verification. Similar hostnames do not resolve the mismatch.

The report concerns a subdomain

Fictional retrieval URI : https://publisher.example/.well-known/security.txt

Host concerned by the report : https://api.publisher.example/

Contact: mailto:security@publisher.example
Expires: 2027-04-22T10:00:00Z

The file belongs to publisher.example; the reported asset is api.publisher.example.

Do not infer subdomain coverage. Check its own discovery location and the product scope explicitly described in the policy.

The file redirects to a provider

Fictional redirect chain : https://publisher.example/.well-known/security.txt → https://intake.partner.example/security.txt

Contact: https://intake.partner.example/report
Expires: 2027-04-22T10:00:00Z

The redirect chain changes host and ends at intake.partner.example.

Document each hop and verify the channel delegation. Using a provider proves neither fraud nor authenticity.

An encryption key is referenced, but not authenticated
Contact: mailto:security@publisher.example
Expires: 2027-04-22T10:00:00Z
Encryption: https://publisher.example/report-key.asc

Encryption supplies an HTTPS key URI without local evidence of the key’s identity.

Authenticate the key through an appropriate means before use. Separate report encryption, file signing and actual protection of the submitted message.

From discovered channel to disclosure case: six useful records

The CERT/CC guide distinguishes finding a contact, providing a useful report and coordinating disclosure. The following outline is a local working proposal, imposing neither a universal fix deadline nor automatic publication.

From discovered channel to disclosure case: six useful records
RecordHow it helpsLimit to state
Product and versionConnect the report to a team and affected releases.A website and an embedded component may have different owners.
Observed behaviourExplain expected and actual results with appropriate reproduction steps.Separate reproduced facts, impact hypotheses and untested conditions.
Channel and recipientsKeep the discovery URL, chosen contact and consulted policy.Send sensitive evidence only after verifying the recipient and transmission protection.
Owner and follow-upRecord acknowledgement, case identifier and agreed next contact.Acknowledgement confirms neither impact nor remediation.
Affected partiesIdentify maintainers, suppliers and products sharing the component.A supplier’s fix does not establish deployment by every consumer.
Fix and communicationConnect fixed versions, remaining limits and actions expected from users.Record the agreed timetable and revisit it as facts change; Expires does not dictate a public disclosure date.

Three limits to keep visible

  • The presence or absence of security.txt alone neither grants nor denies testing permission; applicable conditions need separate verification.
  • RFC 9116 recommends an OpenPGP signature for the file; verifying a signature and establishing trust in its key remain distinct questions.
  • A compromised site can publish a false contact. Intrusions or suspected file tampering may call for an independent channel; this mechanism primarily addresses vulnerability response.

Primary sources and scope

Sources checked on

Referências oficiais

As referências apoiam o método. Adapte as verificações ao seu contexto; elas não são uma certificação. Os títulos das referências originais e os documentos de origem podem estar em outro idioma.

Referências consultadas em .