Re: [PATCH v10 8/8] selinux: include a consumer of the new IMA critical data hook

[Mimi Zohar]

On Wed, 2021-01-13 at 17:10 -0500, Paul Moore wrote:
> On Wed, Jan 13, 2021 at 4:11 PM Mimi Zohar <zohar@xxxxxxxxxxxxx> wrote:
> > On Wed, 2021-01-13 at 14:19 -0500, Paul Moore wrote:
> > > On Wed, Jan 13, 2021 at 2:13 PM Mimi Zohar <zohar@xxxxxxxxxxxxx> wrote:
> > > > On Tue, 2021-01-12 at 11:27 -0500, Paul Moore wrote:
> > > > > On Thu, Jan 7, 2021 at 11:07 PM Tushar Sugandhi
> > > > > <tusharsu@xxxxxxxxxxxxxxxxxxx> wrote:
> > > > > > From: Lakshmi Ramasubramanian <nramas@xxxxxxxxxxxxxxxxxxx>
> > > > > >
> > > > > > SELinux stores the active policy in memory, so the changes to this data
> > > > > > at runtime would have an impact on the security guarantees provided
> > > > > > by SELinux.  Measuring in-memory SELinux policy through IMA subsystem
> > > > > > provides a secure way for the attestation service to remotely validate
> > > > > > the policy contents at runtime.
> > > > > >
> > > > > > Measure the hash of the loaded policy by calling the IMA hook
> > > > > > ima_measure_critical_data().  Since the size of the loaded policy
> > > > > > can be large (several MB), measure the hash of the policy instead of
> > > > > > the entire policy to avoid bloating the IMA log entry.
> > > > > >
> > > > > > To enable SELinux data measurement, the following steps are required:
> > > > > >
> > > > > > 1, Add "ima_policy=critical_data" to the kernel command line arguments
> > > > > >    to enable measuring SELinux data at boot time.
> > > > > > For example,
> > > > > >   BOOT_IMAGE=/boot/vmlinuz-5.10.0-rc1+ root=UUID=fd643309-a5d2-4ed3-b10d-3c579a5fab2f ro nomodeset security=selinux ima_policy=critical_data
> > > > > >
> > > > > > 2, Add the following rule to /etc/ima/ima-policy
> > > > > >    measure func=CRITICAL_DATA label=selinux
> > > > > >
> > > > > > Sample measurement of the hash of SELinux policy:
> > > > > >
> > > > > > To verify the measured data with the current SELinux policy run
> > > > > > the following commands and verify the output hash values match.
> > > > > >
> > > > > >   sha256sum /sys/fs/selinux/policy | cut -d' ' -f 1
> > > > > >
> > > > > >   grep "selinux-policy-hash" /sys/kernel/security/integrity/ima/ascii_runtime_measurements | tail -1 | cut -d' ' -f 6
> > > > > >
> > > > > > Note that the actual verification of SELinux policy would require loading
> > > > > > the expected policy into an identical kernel on a pristine/known-safe
> > > > > > system and run the sha256sum /sys/kernel/selinux/policy there to get
> > > > > > the expected hash.
> > > > > >
> > > > > > Signed-off-by: Lakshmi Ramasubramanian <nramas@xxxxxxxxxxxxxxxxxxx>
> > > > > > Suggested-by: Stephen Smalley <stephen.smalley.work@xxxxxxxxx>
> > > > > > Reviewed-by: Tyler Hicks <tyhicks@xxxxxxxxxxxxxxxxxxx>
> > > > > > ---
> > > > > >  Documentation/ABI/testing/ima_policy |  3 +-
> > > > > >  security/selinux/Makefile            |  2 +
> > > > > >  security/selinux/ima.c               | 64 ++++++++++++++++++++++++++++
> > > > > >  security/selinux/include/ima.h       | 24 +++++++++++
> > > > > >  security/selinux/include/security.h  |  3 +-
> > > > > >  security/selinux/ss/services.c       | 64 ++++++++++++++++++++++++----
> > > > > >  6 files changed, 149 insertions(+), 11 deletions(-)
> > > > > >  create mode 100644 security/selinux/ima.c
> > > > > >  create mode 100644 security/selinux/include/ima.h
> > > > >
> > > > > I remain concerned about the possibility of bypassing a measurement by
> > > > > tampering with the time, but I appear to be the only one who is
> > > > > worried about this so I'm not going to block this patch on those
> > > > > grounds.
> > > > >
> > > > > Acked-by: Paul Moore <paul@xxxxxxxxxxxxxx>
> > > >
> > > > Thanks, Paul.
> > > >
> > > > Including any unique string would cause the buffer hash to change,
> > > > forcing a new measurement.  Perhaps they were concerned with
> > > > overflowing a counter.
> > >
> > > My understanding is that Lakshmi wanted to force a new measurement
> > > each time and felt using a timestamp would be the best way to do that.
> > > A counter, even if it wraps, would have a different value each time
> > > whereas a timestamp is vulnerable to time adjustments.  While a
> > > properly controlled and audited system could be configured and
> > > monitored to detect such an event (I *think*), why rely on that if it
> > > isn't necessary?
> >
> > Why are you saying that even if the counter wraps a new measurement is
> > guaranteed.   I agree with the rest of what you said.
> 
> I was assuming that the IMA code simply compares the passed
> "policy_event_name" value to the previous value, if they are different
> a new measurement is taken, if they are the same the measurement
> request is ignored.  If this is the case the counter value is only
> important in as much as that it is different from the previous value,
> even simply toggling a single bit back and forth would suffice in this
> case.  IMA doesn't keep a record of every previous "policy_event_name"
> value does it?  Am I misunderstanding how
> ima_measure_critical_data(...) works?

Originally, there was quite a bit of discussion as to how much or how
little should be measured for a number of reasons.  One reason is that
the TPM is relatively slow.  Another reason is to limit the size of the
measurement list.  For this reason, duplicate hashes aren't added to
the measurement list or extended into the TPM.

When a dentry is removed from cache, its also removed from IMA's iint
cache.  A subsequent file read would result in adding the measurement
and extending the TPM again.  ima_lookup_digest_entry() is called to
prevent adding the duplicate entry. 

Lakshmi is trying to address the situation where an event changes a
value, but then is restored to the original value.  The original and
subsequent events are measured, but restoring to the original value
isn't re-measured.  This isn't any different than when a file is
modified and then reverted.

Instead of changing the name like this, which doesn't work for files,
allowing duplicate measurements should be generic, based on policy.

Mimi