Hello,

Users of Apache Hadoop should be aware of a security vulnerability recently
discovered, as described by the following CVE. In particular, please note
the "Users affected", "Versions affected", and "Mitigation" sections.

Best,
Aaron

--
Aaron T. Myers
Software Engineer, Cloudera

CVE-2012-1574: Apache Hadoop user impersonation vulnerability

Severity: Critical

Vendor: The Apache Software Foundation

Versions Affected:
Hadoop 0.20.203.0, 0.20.204.0, and 0.20.205.0
Hadoop 1.0.0 to 1.0.1
Hadoop 0.23.0 to 0.23.1.

Users affected: Users who have enabled Hadoop's Kerberos/MapReduce security
features.

Impact: Vulnerability allows an authenticated malicious user to impersonate
any other user on the cluster.

Mitigation:
0.20.20x.x and 1.0.x users should upgrade to 1.0.2
0.23.x users should upgrade to 0.23.2 when it becomes available

Credit:
This issue was discovered by Aaron T. Myers of Cloudera.

This is not a helpful disclosure.

Now we know our "secure" deployment is vulnerable, but have no idea how to mitigate. Claiming an upgrade to a nonexistent version with an, apparently, uncommitted fix as a mitigation is not viable. Where is the JIRA for this?

Best regards,
    - Andy

Problems worthy of attack prove their worth by hitting back. - Piet Hein (via Tom White)

----- Original Message -----
> From: Aaron T. Myers <[EMAIL PROTECTED]>
> To: [EMAIL PROTECTED]; [EMAIL PROTECTED]; [EMAIL PROTECTED]; [EMAIL PROTECTED]
> Cc:
> Sent: Thursday, April 5, 2012 7:31 PM
> Subject: [CVE-2012-1574] Apache Hadoop user impersonation vulnerability
>
> Hello,
>
> Users of Apache Hadoop should be aware of a security vulnerability recently
> discovered, as described by the following CVE. In particular, please note
> the "Users affected", "Versions affected", and
> "Mitigation" sections.
>
> Best,
> Aaron
>
> --
> Aaron T. Myers
> Software Engineer, Cloudera
>
> CVE-2012-1574: Apache Hadoop user impersonation vulnerability
>
> Severity: Critical
>
> Vendor: The Apache Software Foundation
>
> Versions Affected:
> Hadoop 0.20.203.0, 0.20.204.0, and 0.20.205.0
> Hadoop 1.0.0 to 1.0.1
> Hadoop 0.23.0 to 0.23.1.
>
> Users affected: Users who have enabled Hadoop's Kerberos/MapReduce security
> features.
>
> Impact: Vulnerability allows an authenticated malicious user to impersonate
> any other user on the cluster.
>
> Mitigation:
> 0.20.20x.x and 1.0.x users should upgrade to 1.0.2
> 0.23.x users should upgrade to 0.23.2 when it becomes available
>
> Credit:
> This issue was discovered by Aaron T. Myers of Cloudera.
>

On Apr 6, 2012, at 10:02 AM, Andrew Purtell wrote:

> This is not a helpful disclosure.
>
> Now we know our "secure" deployment is vulnerable, but have no idea how to mitigate. Claiming an upgrade to a nonexistent version with an, apparently, uncommitted fix as a mitigation is not viable. Where is the JIRA for this?

*SIGH* You're right, we messed up. We waited for the stable line to be fixed with Hadoop 1.0.2, but we should have waited for the 0.23.2 vote to pass too. The bug is fixed in 0.23.2 rc 0.

-- Owen

I received off list communication that the fix is here: https://github.com/apache/hadoop-common/commit/fda454
Thank you, this is the missing disclosure we were looking for.
I did not go so far back in time as >~ 21 days because the announcement was made today, so missed it.
So there is additional mitigation possible, for example, a user can patch task-controller quite readily and roll out an emergency upgrade.
Best regards,
    - Andy

Problems worthy of attack prove their worth by hitting back. - Piet Hein (via Tom White)

----- Original Message -----
> From: Andrew Purtell <[EMAIL PROTECTED]>
> To: "[EMAIL PROTECTED]" <[EMAIL PROTECTED]>; "[EMAIL PROTECTED]" <[EMAIL PROTECTED]>
> Cc:
> Sent: Friday, April 6, 2012 10:02 AM
> Subject: Re: [CVE-2012-1574] Apache Hadoop user impersonation vulnerability
>
>T his is not a helpful disclosure.
>
> Now we know our "secure" deployment is vulnerable, but have no idea
> how to mitigate. Claiming an upgrade to a nonexistent version with an,
> apparently, uncommitted fix as a mitigation is not viable. Where is the JIRA for
> this?
>
> Best regards,
>
>
>     - Andy
>
> Problems worthy of attack prove their worth by hitting back. - Piet Hein (via
> Tom White)
>
>
>
> ----- Original Message -----
>>  From: Aaron T. Myers <[EMAIL PROTECTED]>
>>  To: [EMAIL PROTECTED]; [EMAIL PROTECTED];
> [EMAIL PROTECTED]; [EMAIL PROTECTED]
>>  Cc:
>>  Sent: Thursday, April 5, 2012 7:31 PM
>>  Subject: [CVE-2012-1574] Apache Hadoop user impersonation vulnerability
>>
>>  Hello,
>>
>>  Users of Apache Hadoop should be aware of a security vulnerability recently
>>  discovered, as described by the following CVE. In particular, please note
>>  the "Users affected", "Versions affected", and
>>  "Mitigation" sections.
>>
>>  Best,
>>  Aaron
>>
>>  --
>>  Aaron T. Myers
>>  Software Engineer, Cloudera
>>
>>  CVE-2012-1574: Apache Hadoop user impersonation vulnerability
>>
>>  Severity: Critical
>>
>>  Vendor: The Apache Software Foundation
>>
>>  Versions Affected:
>>  Hadoop 0.20.203.0, 0.20.204.0, and 0.20.205.0
>>  Hadoop 1.0.0 to 1.0.1
>>  Hadoop 0.23.0 to 0.23.1.
>>
>>  Users affected: Users who have enabled Hadoop's Kerberos/MapReduce
> security
>>  features.
>>
>>  Impact: Vulnerability allows an authenticated malicious user to impersonate
>>  any other user on the cluster.
>>
>>  Mitigation:
>>  0.20.20x.x and 1.0.x users should upgrade to 1.0.2
>>  0.23.x users should upgrade to 0.23.2 when it becomes available
>>
>>  Credit:
>>  This issue was discovered by Aaron T. Myers of Cloudera.
>>
>

Hi Andrew,

On Fri, Apr 6, 2012 at 10:02 AM, Andrew Purtell <[EMAIL PROTECTED]> wrote:

> This is not a helpful disclosure.
>

It's certainly helpful for users of 0.20.20x. and 1.0.x, who can
immediately upgrade to 1.0.2, which was released yesterday. I agree it's
not very helpful for users of 0.23.x, but the assumption is that there are
far fewer of those than users of 0.20.20x and 1.0.x.

Now we know our "secure" deployment is vulnerable, but have no idea how to
> mitigate. Claiming an upgrade to a nonexistent version with an, apparently,
> uncommitted fix as a mitigation is not viable. Where is the JIRA for this?
>

Per the Apache security guidelines (
http://www.apache.org/security/committers.html), there is no up-stream JIRA.

I trust you understand the sensitivity of this issue, and the need to
balance a desire to disclose the issue fully to all users with a desire to
not publish exploits of the issue.

 --
Aaron T. Myers
Software Engineer, Cloudera

> I trust you understand the sensitivity of this issue, and the need to balance a desire to disclose the issue fully to all users with a desire to not publish exploits of the issue.

I can understand that point of view. However,

1) This is open source, not binary only distribution. The patch for this particular issue as I understand it is already in the public change history of the project, just not clearly called out. So what are you actually hiding here?

2) The CVE was itself 404 when I sent the earlier email, so the only available detail was the announcement to security@, a Cloudera web page not referenced, and project change history. I went back 14 days, not far enough, but how was I lnow? Therefore in the absence of information the language of the disclosure implies that the Hadoop implementation of Kerberos authentication is worthless.

Therefore I submit that next time more context is available in the disclosure announcement.

Best regards,

    - Andy
On Apr 6, 2012, at 10:20 AM, "Aaron T. Myers" <[EMAIL PROTECTED]> wrote:

> I trust you understand the sensitivity of this issue, and the need to
> balance a desire to disclose the issue fully to all users with a desire to
> not publish exploits of the issue.

Furthermore, I expect vendors were fully in the loop on some private mailing list. But here users get rather poor disclosure. Need I remind everyone that in open source, users are your peers? If one of your peers is running a customized version of your open source product in production, you must admit there was no actionable information in that disclosure.

Best regards,

    - Andy
On Apr 6, 2012, at 11:43 AM, Andrew Purtell <[EMAIL PROTECTED]> wrote:

>> I trust you understand the sensitivity of this issue, and the need to balance a desire to disclose the issue fully to all users with a desire to not publish exploits of the issue.
>
> I can understand that point of view. However,
>
> 1) This is open source, not binary only distribution. The patch for this particular issue as I understand it is already in the public change history of the project, just not clearly called out. So what are you actually hiding here?
>
> 2) The CVE was itself 404 when I sent the earlier email, so the only available detail was the announcement to security@, a Cloudera web page not referenced, and project change history. I went back 14 days, not far enough, but how was I lnow? Therefore in the absence of information the language of the disclosure implies that the Hadoop implementation of Kerberos authentication is worthless.
>
> Therefore I submit that next time more context is available in the disclosure announcement.
>
> Best regards,
>
>    - Andy
>
>
> On Apr 6, 2012, at 10:20 AM, "Aaron T. Myers" <[EMAIL PROTECTED]> wrote:
>
>> I trust you understand the sensitivity of this issue, and the need to
>> balance a desire to disclose the issue fully to all users with a desire to
>> not publish exploits of the issue.

Hey Andrew,

The project member were in the loop on the private Hadoop security
mailing list. This wasn't a vendor discussion.

We had a discussion  about how much to disclose before sending out
this notification, and there were differing opinions. Agree that we
should disclose more information next time around, I'll push hard for
that next time.

Thanks,
Eli

On Fri, Apr 6, 2012 at 12:08 PM, Andrew Purtell <[EMAIL PROTECTED]> wrote:
> Furthermore, I expect vendors were fully in the loop on some private mailing list. But here users get rather poor disclosure. Need I remind everyone that in open source, users are your peers? If one of your peers is running a customized version of your open source product in production, you must admit there was no actionable information in that disclosure.
>
> Best regards,
>
>    - Andy
>
>
> On Apr 6, 2012, at 11:43 AM, Andrew Purtell <[EMAIL PROTECTED]> wrote:
>
>>> I trust you understand the sensitivity of this issue, and the need to balance a desire to disclose the issue fully to all users with a desire to not publish exploits of the issue.
>>
>> I can understand that point of view. However,
>>
>> 1) This is open source, not binary only distribution. The patch for this particular issue as I understand it is already in the public change history of the project, just not clearly called out. So what are you actually hiding here?
>>
>> 2) The CVE was itself 404 when I sent the earlier email, so the only available detail was the announcement to security@, a Cloudera web page not referenced, and project change history. I went back 14 days, not far enough, but how was I lnow? Therefore in the absence of information the language of the disclosure implies that the Hadoop implementation of Kerberos authentication is worthless.
>>
>> Therefore I submit that next time more context is available in the disclosure announcement.
>>
>> Best regards,
>>
>>    - Andy
>>
>>
>> On Apr 6, 2012, at 10:20 AM, "Aaron T. Myers" <[EMAIL PROTECTED]> wrote:
>>
>>> I trust you understand the sensitivity of this issue, and the need to
>>> balance a desire to disclose the issue fully to all users with a desire to
>>> not publish exploits of the issue.