Hi,
I am looking to use SELinux to secure a process that is made up of a number of discrete, sequential stages. One stage communicates to the next by writing results to a file and then an external process modifies the SELinux context of the file to allow the next stage to read the file and so on until the final stage is reached and the processing stops.
The problem I have is that the number of stages is variable and can change with each invocation of the process, i.e. when I create the process I know the number of stages that will be required in it, but the number of stages could change with each invocation. I think therefore, that I need a means of creating new contexts on the fly and assigning them to the processes. Is it possible with SELinux to create a new security context (domain for the output file, and user/role for the stage process) on the fly and execute a process within that context such that it could poll a directory for input files and, if it is permitted to read the file perform its operation?
Many Thanks,
Tim Sheppard
This email and any attachments to it may be confidential and are intended solely for the use of the individual to whom it is addressed. If you are not the intended recipient of this email, you must neither take any action based upon its contents, nor copy or show it to anyone. Please contact the sender if you believe you have received this email in error. QinetiQ may monitor email traffic data and also the content of email for the purposes of security. QinetiQ Limited (Registered in England & Wales: Company Number: 3796233) Registered office: Cody Technology Park, Ively Road, Farnborough, Hampshire, GU14 0LX http://www.qinetiq.com.
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
On 04/24/2012 01:16 PM, Tim Sheppard wrote:
Hi,
I am looking to use SELinux to secure a process that is made up of a number of discrete, sequential stages. One stage communicates to the next by writing results to a file and then an external process modifies the SELinux context of the file to allow the next stage to read the file and so on until the final stage is reached and the processing stops.
The problem I have is that the number of stages is variable and can change with each invocation of the process, i.e. when I create the process I know the number of stages that will be required in it, but the number of stages could change with each invocation. I think therefore, that I need a means of creating new contexts on the fly and assigning them to the processes. Is it possible with SELinux to create a new security context (domain for the output file, and user/role for the stage process) on the fly and execute a process within that context such that it could poll a directory for input files and, if it is permitted to read the file perform its operation?
Many Thanks,
Tim Sheppard
If each process is going to require the same approximate access. I would use MCS labeling for separation.
Basically define a sandbox type for your processes to run within and then have your controlling app pick an MCS label for the process and the content.
chcon -t confined_file_t -l s0:c1 passing.file runcon -t confined_proc_t -l s0:c1 proc1 When done chcon -l s0:c2 passing.file runcon -t confined_proc_t -l s0:c2 proc2 ...
On 4/24/2012 1:37 PM, Daniel J Walsh wrote:
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
On 04/24/2012 01:16 PM, Tim Sheppard wrote:
Hi,
I am looking to use SELinux to secure a process that is made up of a number of discrete, sequential stages. One stage communicates to the next by writing results to a file and then an external process modifies the SELinux context of the file to allow the next stage to read the file and so on until the final stage is reached and the processing stops.
The problem I have is that the number of stages is variable and can change with each invocation of the process, i.e. when I create the process I know the number of stages that will be required in it, but the number of stages could change with each invocation. I think therefore, that I need a means of creating new contexts on the fly and assigning them to the processes. Is it possible with SELinux to create a new security context (domain for the output file, and user/role for the stage process) on the fly and execute a process within that context such that it could poll a directory for input files and, if it is permitted to read the file perform its operation?
Many Thanks,
Tim Sheppard
If each process is going to require the same approximate access. I would use MCS labeling for separation.
Basically define a sandbox type for your processes to run within and then have your controlling app pick an MCS label for the process and the content.
chcon -t confined_file_t -l s0:c1 passing.file runcon -t confined_proc_t -l s0:c1 proc1 When done chcon -l s0:c2 passing.file runcon -t confined_proc_t -l s0:c2 proc2 ... -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.12 (GNU/Linux) Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/
iEYEARECAAYFAk+W5O4ACgkQrlYvE4MpobNfcgCdHfJl2HRNfsi3YsAou1lA4MY9 gfIAoNzG8u5W3o9f7LUAixBSxvaLyXyE =iHPC
-----END PGP SIGNATURE-----
selinux mailing list selinux@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/selinux
It sounds like you want an assured pipeline for your process. You should be able to search for that term on the selinuxproject.org wiki and find a demonstration of that concept on there. If its not there email Stephen Smalley and see what the status of that code is. I don't remember if we ever released it.
Dave
On Tue, 2012-04-24 at 21:25 -0400, David P. Quigley wrote:
It sounds like you want an assured pipeline for your process. You should be able to search for that term on the selinuxproject.org wiki and find a demonstration of that concept on there. If its not there email Stephen Smalley and see what the status of that code is. I don't remember if we ever released it.
http://selinuxproject.org/page/PipelineDemo
On 24/04/12 18:37, Daniel J Walsh wrote:
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
On 04/24/2012 01:16 PM, Tim Sheppard wrote:
Hi,
I am looking to use SELinux to secure a process that is made up of a number of discrete, sequential stages. One stage communicates to the next by writing results to a file and then an external process modifies the SELinux context of the file to allow the next stage to read the file and so on until the final stage is reached and the processing stops.
The problem I have is that the number of stages is variable and can change with each invocation of the process, i.e. when I create the process I know the number of stages that will be required in it, but the number of stages could change with each invocation. I think therefore, that I need a means of creating new contexts on the fly and assigning them to the processes. Is it possible with SELinux to create a new security context (domain for the output file, and user/role for the stage process) on the fly and execute a process within that context such that it could poll a directory for input files and, if it is permitted to read the file perform its operation?
Many Thanks,
Tim Sheppard
If each process is going to require the same approximate access. I would use MCS labeling for separation.
Basically define a sandbox type for your processes to run within and then have your controlling app pick an MCS label for the process and the content.
chcon -t confined_file_t -l s0:c1 passing.file runcon -t confined_proc_t -l s0:c1 proc1 When done chcon -l s0:c2 passing.file runcon -t confined_proc_t -l s0:c2 proc2 ... -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.12 (GNU/Linux) Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/
iEYEARECAAYFAk+W5O4ACgkQrlYvE4MpobNfcgCdHfJl2HRNfsi3YsAou1lA4MY9 gfIAoNzG8u5W3o9f7LUAixBSxvaLyXyE =iHPC -----END PGP SIGNATURE-----
Daniel,
Many Thanks, I'll look into MCS a bit more then. The other requirement I have to to be able to do this all programatically with C/C++ rather than via bash. I assume that as there is a bash command there are comparable function calls?
Cheers,
Tim
This email and any attachments to it may be confidential and are intended solely for the use of the individual to whom it is addressed. If you are not the intended recipient of this email, you must neither take any action based upon its contents, nor copy or show it to anyone. Please contact the sender if you believe you have received this email in error. QinetiQ may monitor email traffic data and also the content of email for the purposes of security. QinetiQ Limited (Registered in England & Wales: Company Number: 3796233) Registered office: Cody Technology Park, Ively Road, Farnborough, Hampshire, GU14 0LX http://www.qinetiq.com.
On 24/04/12 18:37, Daniel J Walsh wrote:
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
On 04/24/2012 01:16 PM, Tim Sheppard wrote:
Hi,
I am looking to use SELinux to secure a process that is made up of a number of discrete, sequential stages. One stage communicates to the next by writing results to a file and then an external process modifies the SELinux context of the file to allow the next stage to read the file and so on until the final stage is reached and the processing stops.
The problem I have is that the number of stages is variable and can change with each invocation of the process, i.e. when I create the process I know the number of stages that will be required in it, but the number of stages could change with each invocation. I think therefore, that I need a means of creating new contexts on the fly and assigning them to the processes. Is it possible with SELinux to create a new security context (domain for the output file, and user/role for the stage process) on the fly and execute a process within that context such that it could poll a directory for input files and, if it is permitted to read the file perform its operation?
Many Thanks,
Tim Sheppard
If each process is going to require the same approximate access. I would use MCS labeling for separation.
Basically define a sandbox type for your processes to run within and then have your controlling app pick an MCS label for the process and the content.
chcon -t confined_file_t -l s0:c1 passing.file runcon -t confined_proc_t -l s0:c1 proc1 When done chcon -l s0:c2 passing.file runcon -t confined_proc_t -l s0:c2 proc2 ... -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.12 (GNU/Linux) Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/
iEYEARECAAYFAk+W5O4ACgkQrlYvE4MpobNfcgCdHfJl2HRNfsi3YsAou1lA4MY9 gfIAoNzG8u5W3o9f7LUAixBSxvaLyXyE =iHPC -----END PGP SIGNATURE-----
The other thing I need to be able to do is constrain a process to only permit it to change a context from c1 -> c2 (or label1 -> label2 in the general case). Is this possible to manage in a policy at all and if so is there any runtime flexibility over the rules that can be created to allow me to specify the only legal changes are c1 -> c2 -> c3 -> end for process1-1 and c1 -> c2 -> c3 -> c4-> c5 -> end for process1-2. Essentially, I need to be able to ensure that a process cannot change the security labelling to bypass a stage in the chain.
Many Thanks,
Tim
This email and any attachments to it may be confidential and are intended solely for the use of the individual to whom it is addressed. If you are not the intended recipient of this email, you must neither take any action based upon its contents, nor copy or show it to anyone. Please contact the sender if you believe you have received this email in error. QinetiQ may monitor email traffic data and also the content of email for the purposes of security. QinetiQ Limited (Registered in England & Wales: Company Number: 3796233) Registered office: Cody Technology Park, Ively Road, Farnborough, Hampshire, GU14 0LX http://www.qinetiq.com.
On Wed, 2012-04-25 at 10:37 +0100, Tim Sheppard wrote:
The other thing I need to be able to do is constrain a process to only permit it to change a context from c1 -> c2 (or label1 -> label2 in the general case). Is this possible to manage in a policy at all and if so is there any runtime flexibility over the rules that can be created to allow me to specify the only legal changes are c1 -> c2 -> c3 -> end for process1-1 and c1 -> c2 -> c3 -> c4-> c5 -> end for process1-2. Essentially, I need to be able to ensure that a process cannot change the security labelling to bypass a stage in the chain.
That's more easily expressed via Type Enforcement (which supports intransitive relationships) than MLS. Type Enforcement is what is normally used for assured pipelines. Some people have instead modeled assured pipelines using MLS categories by using incomparable category sets on the input and output sides of each stage and clever configuration, but that's more complicated and less precise in its control of the information flow.
With regard to runtime flexibility, there isn't presently a way to programmatically instantiate new domains and types; your application can only use ones that are already defined in the policy. So options might include: - Ship a policy module with your application that defines the maximal set of domains/types ever required for one of these pipelines, and then your application can just use the relevant subset for any given instance, or
- Implement some kind of userspace daemon that runs with privilege (and thus can load policy) and accepts requests to instantiate new pipeline stage domains and types, generates new policy modules to define them, and loads them (I think maybe some folks from Tresys have done this in the past?).
Or you could always extend SELinux to directly support such functionality, but that would be a larger project (and require a custom kernel).
Many Thanks,
I will look into the Secure Pipelines and policy delivery / update options.
Regards,
Tim
On 25/04/12 13:40, Stephen Smalley wrote:
On Wed, 2012-04-25 at 10:37 +0100, Tim Sheppard wrote:
The other thing I need to be able to do is constrain a process to only permit it to change a context from c1 -> c2 (or label1 -> label2 in the general case). Is this possible to manage in a policy at all and if so is there any runtime flexibility over the rules that can be created to allow me to specify the only legal changes are c1 -> c2 -> c3 -> end for process1-1 and c1 -> c2 -> c3 -> c4-> c5 -> end for process1-2. Essentially, I need to be able to ensure that a process cannot change the security labelling to bypass a stage in the chain.
That's more easily expressed via Type Enforcement (which supports intransitive relationships) than MLS. Type Enforcement is what is normally used for assured pipelines. Some people have instead modeled assured pipelines using MLS categories by using incomparable category sets on the input and output sides of each stage and clever configuration, but that's more complicated and less precise in its control of the information flow.
With regard to runtime flexibility, there isn't presently a way to programmatically instantiate new domains and types; your application can only use ones that are already defined in the policy. So options might include:
- Ship a policy module with your application that defines the maximal
set of domains/types ever required for one of these pipelines, and then your application can just use the relevant subset for any given instance, or
- Implement some kind of userspace daemon that runs with privilege (and
thus can load policy) and accepts requests to instantiate new pipeline stage domains and types, generates new policy modules to define them, and loads them (I think maybe some folks from Tresys have done this in the past?).
Or you could always extend SELinux to directly support such functionality, but that would be a larger project (and require a custom kernel).
This email and any attachments to it may be confidential and are intended solely for the use of the individual to whom it is addressed. If you are not the intended recipient of this email, you must neither take any action based upon its contents, nor copy or show it to anyone. Please contact the sender if you believe you have received this email in error. QinetiQ may monitor email traffic data and also the content of email for the purposes of security. QinetiQ Limited (Registered in England & Wales: Company Number: 3796233) Registered office: Cody Technology Park, Ively Road, Farnborough, Hampshire, GU14 0LX http://www.qinetiq.com.
selinux@lists.fedoraproject.org
-
Daniel J Walsh -
David Quigley -
Stephen Smalley -
Tim Sheppard