Summary
{field}.isFilterable access control can be bypassed in update and delete mutations by adding additional unique filters. These filters can be used as an oracle to probe the existence or value of otherwise unreadable fields.
Specifically, when a mutation includes a where clause with multiple unique filters (e.g. id and email), Keystone will attempt to match records even if filtering by the latter fields would normally be rejected by field.isFilterable or list.defaultIsFilterable. This can allow malicious actors to infer the presence of a particular field value when a filter is successful in returning a result.
Impact
This affects any project relying on the default or dynamic isFilterable behaviour (at the list or field level) to prevent external users from using the filtering of fields as a discovery mechanism. While this access control is respected during findMany operations, it was not completely enforced during update and delete mutations when accepting more than one unique where values in filters.
This has no impact on projects using isFilterable: false or defaultIsFilterable: false for sensitive fields, or if you have otherwise omitted filtering by these fields from your GraphQL schema. (See workarounds)
Patches
This issue has been patched in @keystone-6/core version 6.5.0.
Workarounds
To mitigate this issue in older versions where patching is not a viable pathway.
- Set
isFilterable: false statically for relevant fields to prevent filtering by them earlier in the access control pipeline (that is, don't use functions)
- Set
{field}.graphql.omit.read: true for relevant fields, which implicitly removes filtering by these fields your GraphQL schema
- Deny
update and delete operations for the relevant lists completely (e.g list({ access: { operation: { update: false, delete: false } }, ... }))
References
Summary
{field}.isFilterableaccess control can be bypassed inupdateanddeletemutations by adding additional unique filters. These filters can be used as an oracle to probe the existence or value of otherwise unreadable fields.Specifically, when a mutation includes a
whereclause with multiple unique filters (e.g.idandemail), Keystone will attempt to match records even if filtering by the latter fields would normally be rejected byfield.isFilterableorlist.defaultIsFilterable. This can allow malicious actors to infer the presence of a particular field value when a filter is successful in returning a result.Impact
This affects any project relying on the default or dynamic
isFilterablebehaviour (at the list or field level) to prevent external users from using the filtering of fields as a discovery mechanism. While this access control is respected duringfindManyoperations, it was not completely enforced duringupdateanddeletemutations when accepting more than one uniquewherevalues in filters.This has no impact on projects using
isFilterable: falseordefaultIsFilterable: falsefor sensitive fields, or if you have otherwise omitted filtering by these fields from your GraphQL schema. (See workarounds)Patches
This issue has been patched in
@keystone-6/coreversion 6.5.0.Workarounds
To mitigate this issue in older versions where patching is not a viable pathway.
isFilterable: falsestatically for relevant fields to prevent filtering by them earlier in the access control pipeline (that is, don't use functions){field}.graphql.omit.read: truefor relevant fields, which implicitly removes filtering by these fields your GraphQL schemaupdateanddeleteoperations for the relevant lists completely (e.glist({ access: { operation: { update: false, delete: false } }, ... }))References