You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@jena.apache.org by Andy Seaborne <an...@apache.org> on 2018/11/15 10:33:14 UTC

Fuseki & Security

Hi all - the work of Fuseki and security is nearly ready to merge to 
master so here is a brief overview of what has become a large PR (sorry 
about that).

Jena 3.9.0 introduced a library to for graph-level security (module 
jena-fuseki-access). This work adds user authentication, https, and 
support for access control on the server as a whole and on databases.

JENA-1594 : Provide per-graph access control in Fuseki
JENA-1623 : Access control of server and services
JENA-1627 : HTTPS support
PR#486: https://github.com/apache/jena/pull/486

Most of this is relying on Jetty features and packaging them up to make 
them easier to use in the context of Fuseki.

* https support.

Because data transfer over unencrypted HTTP is not a good thing if you 
wanted to restrict access to the data in any way.

* Users and passwords.

Basic and Digest auth supported; the server defaults to digest auth. 
that might have to change because in my experience simple scripts work 
with basic auth - it is the default with curl and wget.

The code provides password file support - for more complex integrations 
into a deployment environment, e.g. SSO, LDAP, access to the Jetty 
features is easily available. If there are common patterns, better 
support can be added later but for now it's standalone user/passwd or 
extension in Java code.

* Access control for the server

The server can be set to require an authenticated user.

* Access control for the services

This can also be applied to individual services.

These two are probably the most useful part of the new features for 
general use.

A deployed server can have different access control lists on the server 
itself, and on various databases (a user has to satisfy all restrictions).

There is no per-endpoint security ATM - something to add later. The same 
dataset can be shared between different services with different access 
control lists.

* Graph-level access control.

This only applies to a read-only database. There are per-user access 
control lists; the union graph is all named graphs the user has access to.

When used with TDB, the implementation uses TDB tuple filtering so it is 
efficient. In the general case, it is a specially created view of the 
dataset which is functionally correct but will call back to general 
query execution, and for expensive queries may well be observable less 
efficient.

Updates can be done under different access control via a different set 
of service endpoints over the same shared TDB database.

     Andy

Unrelated small things : these are small fixes of things I came across 
or have been mentioned recently during the lifetime of this PR which got 
fixed:

JENA-1625 : Update DynamicDatasets.
JENA-1626 : Txn improvements.
JENA-1631 : Use DatasetGraphMapLink transactions.
JENA-1629 : Print CONSTRUCT Quads executions.
JENA-1632 : https support in SOH scripts.

Re: Fuseki & Security

Posted by Andy Seaborne <an...@apache.org>.

On 17/11/2018 18:06, Andy Seaborne wrote:
> Merged - along with all the bits and pieces that accumulated.
> 
> First draft, not finished, documentation
> 
> http://jena.staging.apache.org/documentation/fuseki2/data-access-control.html 
> 
> 
> Outstanding, but not blocking a release:
> 
> Endpoint access control (it can be done by defining different services 
> but it will be neater if it were more consistently handled).

Done now.

> 
>      Andy

I am very grateful to Mulesoft for funding this work.  It isn't a small 
thing to do in small pieces, so having blocks of time to be able to do 
it has been essential.

Re: Fuseki & Security

Posted by Andy Seaborne <an...@apache.org>.
Merged - along with all the bits and pieces that accumulated.

First draft, not finished, documentation

http://jena.staging.apache.org/documentation/fuseki2/data-access-control.html

Outstanding, but not blocking a release:

Endpoint access control (it can be done by defining different services 
but it will be neater if it were more consistently handled).

     Andy

On 15/11/2018 10:33, Andy Seaborne wrote:
> Hi all - the work of Fuseki and security is nearly ready to merge to 
> master so here is a brief overview of what has become a large PR (sorry 
> about that).
> 
> Jena 3.9.0 introduced a library to for graph-level security (module 
> jena-fuseki-access). This work adds user authentication, https, and 
> support for access control on the server as a whole and on databases.
> 
> JENA-1594 : Provide per-graph access control in Fuseki
> JENA-1623 : Access control of server and services
> JENA-1627 : HTTPS support
> PR#486: https://github.com/apache/jena/pull/486
> 
> Most of this is relying on Jetty features and packaging them up to make 
> them easier to use in the context of Fuseki.
> 
> * https support.
> 
> Because data transfer over unencrypted HTTP is not a good thing if you 
> wanted to restrict access to the data in any way.
> 
> * Users and passwords.
> 
> Basic and Digest auth supported; the server defaults to digest auth. 
> that might have to change because in my experience simple scripts work 
> with basic auth - it is the default with curl and wget.
> 
> The code provides password file support - for more complex integrations 
> into a deployment environment, e.g. SSO, LDAP, access to the Jetty 
> features is easily available. If there are common patterns, better 
> support can be added later but for now it's standalone user/passwd or 
> extension in Java code.
> 
> * Access control for the server
> 
> The server can be set to require an authenticated user.
> 
> * Access control for the services
> 
> This can also be applied to individual services.
> 
> These two are probably the most useful part of the new features for 
> general use.
> 
> A deployed server can have different access control lists on the server 
> itself, and on various databases (a user has to satisfy all restrictions).
> 
> There is no per-endpoint security ATM - something to add later. The same 
> dataset can be shared between different services with different access 
> control lists.
> 
> * Graph-level access control.
> 
> This only applies to a read-only database. There are per-user access 
> control lists; the union graph is all named graphs the user has access to.
> 
> When used with TDB, the implementation uses TDB tuple filtering so it is 
> efficient. In the general case, it is a specially created view of the 
> dataset which is functionally correct but will call back to general 
> query execution, and for expensive queries may well be observable less 
> efficient.
> 
> Updates can be done under different access control via a different set 
> of service endpoints over the same shared TDB database.
> 
>      Andy
> 
> Unrelated small things : these are small fixes of things I came across 
> or have been mentioned recently during the lifetime of this PR which got 
> fixed:
> 
> JENA-1625 : Update DynamicDatasets.
> JENA-1626 : Txn improvements.
> JENA-1631 : Use DatasetGraphMapLink transactions.
> JENA-1629 : Print CONSTRUCT Quads executions.
> JENA-1632 : https support in SOH scripts.