标签：
With the default Apache HBase configuration, everyone is allowed to read from and write to all tables available in the system. For many enterprise setups, this kind of policy is unacceptable.
Administrators can set up firewalls that decide which machines are allowed to communicate with HBase. However, machines that can pass the firewall are still allowed to read from and write to all tables. This kind of mechanism is effective but insufficient because HBase still cannot differentiate between multiple users that use the same client machines, and there is still no granularity with regard to HBase table, column family, or column qualifier access.
In this post, we will discuss how Kerberos is used with Hadoop and HBase to provide User Authentication, and how HBase implements User Authorization to grant users permissions for particular actions on a specified set of data.
A secure HBase aims to protect against sniffers, unauthenticated/unauthorized users and network-based attacks. It does not protect against authorized users who accidentally delete all the data.
HBase can be configured to provide User Authentication, which ensures that only authorized users can communicate with HBase. The authorization system is implemented at the RPC level, and is based on the Simple Authentication and Security Layer (SASL), which supports (among other authentication mechanisms) Kerberos. SASL allows authentication, encryption negotiation and/or message integrity verification on a per connection basis ( “hbase.rpc.protection” configuration property).
The next step after enabling User Authentication is to give an admin the ability to define a series of User Authorization rules that allow or deny particular actions. The Authorization system, also known as the Access Controller Coprocessor or Access Control List (ACL), is available from HBase 0.92 (CDH4) onward and gives the ability to define authorization policy (Read/Write/Create/Admin), with table/family/qualifier granularity, for a specified user.
Kerberos is a networked authentication protocol. It is designed to provide strong authentication for client/server applications by using secret-key cryptography. The Kerberos protocol uses strong cryptography (AES, 3DES, …) so that a client can prove its identity to a server (and vice versa) across an insecure network connection. After a client and server have used Kerberos to prove their identities, they can also encrypt all of their communications to assure privacy and data integrity as they go about their business.
Ticket exchange protocol
At a high level, to access a service using Kerberos, each client must follow three steps:
Since HBase depends on HDFS and ZooKeeper, secure HBase relies on a secure HDFS and a secure ZooKeeper. This means that the HBase servers need to create a secure service session, as described above, to communicate with HDFS and ZooKeeper.
All the files written by HBase are stored in HDFS. As in Unix filesystems, the access control provided by HDFS is based on users, groups and permissions. All the files created by HBase have “hbase” as user, but this access control is based on the username provided by the system, and everyone that can access the machine is potentially able to “sudo” as the user “hbase”. Secure HDFS adds the authentication steps that guarantee that the “hbase” user is trusted.
ZooKeeper has an Access Control List (ACL) on each znode that allows read/write access to the users based on user information in a similar manner to HDFS.
Now that our users are authenticated via Kerberos, we are sure that the username that we received is one of our trusted users. Sometimes this is not enough granularity – we want to control that a specified user is able to read or write a table. To do that, HBase provides an Authorization mechanism that allows restricted access for specified users.
To enable this feature, you must enable the Access Controller coprocessor, by adding it to hbase-site.xml under the master and region server coprocessor classes. (See how to setup the HBase security configuration here.)
A coprocessor is code that runs inside each HBase Region Server and/or Master. It is able to intercept most operations (put, get, delete, …), and run arbitrary code before and/or after the operation is executed.
Using this ability to execute some code before each operation, the Access Controller coprocessor can check the user rights and decide if the user can or cannot execute the operation.
Rights management and _acl_ table
The HBase shell has a couple of commands that allows an admin to manage the user rights:
grant [table] [family] [qualifier]revoke [table] [family] [qualifier]As you see, an admin has the ability to restrict user access based on the table schema:
grant ‘User-W‘, ‘R‘, ‘Table-X‘, ‘Family-Y‘)grant ‘User-W‘, ‘RW‘, ‘Table-X‘, ‘Family-Y‘, ‘Qualifier-Z‘)An admin also has the ability to grant global rights, which operate at the cluster level, such as creating tables, balancing regions, shutting down the cluster and so on:
grant ‘User-W‘, ‘C‘)grant ‘User-W‘, ‘A‘)All the permissions are stored in a table created by the Access Controller coprocessor, called _acl_. The primary key of this table is the table name that you specify in the grant command. The _acl_ table has just one column family and each qualifier describes the granularity of rights for a particular table/user. The value contains the actual rights granted.
As you can see, the HBase shell commands are tightly related to how the data is stored. The grant command adds or updates one row, and the revoke command removes one row from the _acl_ table.
Access Controller under the hood
As mentioned previously, the Access Controller coprocessor uses the ability to intercept each user request, and check if the user has the rights to execute the operations.
For each operation, the Access Controller needs to query the _acl_ table to see if the user has the rights to execute the operation.
However, this operation can have a negative impact on performance. The solution to fix this problem is using the _acl_ table for persistence and ZooKeeper to speed up the rights lookup. Each region server loads the _acl_ table in memory and get notified of changes by the ZkPermissionWatcher. In this way, every region server has the updated value every time and each permission check is performed by using an in-memory map.
While Kerberos is a stable, well-tested and proven authentication system, the HBase ACL feature is still very basic and its semantics are still evolving. HBASE-6096 is the umbrella JIRA as reference for all the improvements to ship in a v2 of the ACL feature.
Another open topic on authorization and access control is implementing a per-KeyValue security system (HBASE-6222) that will give the ability to have different values on the same cell associated with a security tag. That would allow to showing a particular piece of information based on the user’s permissions.
HBase Security adds two extra features that allow you to protect your data against sniffers or other network attacks (by using Kerberos to authenticate users and encrypt communications between services), and allow you to define User Authorization policies, restrict operations, and limit data visibility for particular users.
原文地址：http://blog.cloudera.com/blog/2012/09/understanding-user-authentication-and-authorization-in-apache-hbase/
How-to: Enable User Authentication and Authorization in Apache HBase
标签：
原文地址：http://www.cnblogs.com/yueweimian/p/4260444.html
