MRG Grid 2.2: multiple vulnerabilities

From the Red Hat advisory:

A number of unprotected resources (web pages, export functionality, image viewing) were found in Cumin. An unauthenticated user could bypass intended access restrictions, resulting in information disclosure. (CVE-2012-2680)

Cumin could generate weak session keys, potentially allowing remote attackers to predict session keys and obtain unauthorized access to Cumin. (CVE-2012-2681)

Multiple cross-site scripting flaws in Cumin could allow remote attackers to inject arbitrary web script on a web page displayed by Cumin. (CVE-2012-2683)

An SQL injection flaw in Cumin could allow remote attackers to manipulate the contents of the back-end database via a specially-crafted URL. (CVE-2012-2684)

When Cumin handled image requests, clients could request images of arbitrary sizes. This could result in large memory allocations on the Cumin server, leading to an out-of-memory condition. (CVE-2012-2685)

Cumin did not protect against Cross-Site Request Forgery attacks. If an attacker could trick a user, who was logged into the Cumin web interface, into visiting a specially-crafted web page, it could lead to unauthorized command execution in the Cumin web interface with the privileges of the logged-in user. (CVE-2012-2734)

A session fixation flaw was found in Cumin. An authenticated user able to pre-set the Cumin session cookie in a victim's browser could possibly use this flaw to steal the victim's session after they log into Cumin. (CVE-2012-2735)

It was found that authenticated users could send a specially-crafted HTTP POST request to Cumin that would cause it to submit a job attribute change to Condor. This could be used to change internal Condor attributes, including the Owner attribute, which could allow Cumin users to elevate their privileges. (CVE-2012-3459)

It was discovered that Condor's file system authentication challenge accepted directories with weak permissions (for example, world readable, writable and executable permissions). If a user created a directory with such permissions, a local attacker could rename it, allowing them to execute jobs with the privileges of the victim user. (CVE-2012-3492)

It was discovered that Condor exposed private information in the data in the ClassAds format served by condor_startd. An unauthenticated user able to connect to condor_startd's port could request a ClassAd for a running job, provided they could guess or brute-force the PID of the job. This could expose the ClaimId which, if obtained, could be used to control the job as well as start new jobs on the system. (CVE-2012-3493)

It was discovered that the ability to abort a job in Condor only required WRITE authorization, instead of a combination of WRITE authorization and job ownership. This could allow an authenticated attacker to bypass intended restrictions and abort any idle job on the system. (CVE-2012-3491)