cassandra-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From oleg yusim <olegyu...@gmail.com>
Subject Re: Security labels
Date Fri, 29 Jan 2016 18:03:34 GMT
Dani,

I really appreciate you response. Actually, session timeouts and security
labels are two different topics (first is about attack when somebody
opened, say, ssh window to DB, left his machine unattended and somebody
else stole his session, second - to enable DB to support what called MAC
access model - stays for mandatory access control. It is widely used in the
government and military, but not outside of it, we all are used to DAC
access control model). However, I think you are right and I should move all
my queries under the one big roof and call this thread "Security". I will
do this today.

Now, about what you have said, I just answered the same to Jon, in Session
Timeout thread, but would quickly re-cap here. I understand that
Cassandra's architecture was aimed and tailored for completely different
type of scenario. However, unfortunately, that doesn't mean that Cassandra
is not vulnerable to the same very set of attacks relational database would
be vulnerable to. It just means Cassandra is not protected against those
attacks, because protection against them was not thought of, when database
was created. I already gave the AAA and session's timeout example in Jon's
thread, and those are just one of many.

Now what I'm trying to do, I'm trying to create a STIG - security federal
compliance document, which will assess Cassandra against SRG concepts
(security federal compliance recommendations for databases overall) and
will highlight what is not met, and can't be in current design (i.e. what
system architects should keep in mind and what they need to compensate for
with other controls on different layers of system model) and  what can be
met either with configuration or with little enhancement (and how).

That document would be of great help for Cassandra as a product because it
would allow it to be marketed as a product with existing security
assessment and guidelines, performed according to DoD standards. It would
also allow to move product in the general direction of improving its
security posture. Finally, the document would be posted on DISA site (
http://iase.disa.mil/stigs/Pages/a-z.aspx) available for every security
architect to utilize, which would greatly reduce the risk for Cassandra
product to be hacked in a field.

To clear things out - what I ask about are not my expectations. I really do
not expect developers of Cassandra to run and start implementing security
labels, just because I asked about it. :) My questions are to build my
internal knowledge of DB current design, so that I can build my security
assessment based of it, not more, not less.

I guess, summarizing what I said on top, from what I'm doing Cassandra as a
product would end up benefiting quite a bit. That is why I think it would
make sense for Cassandra community to help me with my questions even if
they sound completely of the traditional "grid".

Thanks again, I really appreciate your response and conversation overall.

Oleg

On Fri, Jan 29, 2016 at 11:20 AM, Dani Traphagen <
dani.traphagen@datastax.com> wrote:

> Also -- it looks like you're really asking questions about session
> timeouts and security labels as they associate, would be more helpful to
> keep in one thread. :)
>
>
> On Friday, January 29, 2016, Dani Traphagen <dani.traphagen@datastax.com>
> wrote:
>
>> Hi Oleg,
>>
>> I understand your frustration but unfortunately, in the terms of your
>> security assessment, you have fallen into a mismatch for Cassandra's
>> utility.
>>
>> The eventuality of having multiple sockets open without the query input
>> for long durations of time isn't something that was
>> architected...because...Cassnadra was built to take massive quantities
>> of queries both in volume and velocity.
>>
>> Your expectation of the database isn't in line with how our why it was
>> designed. Generally, security solutions are architected
>> around Cassandra, baked into the data model, many solutions
>> are home-brewed, written into the application or provided by using another
>> security client.
>>
>> DSE has different security aspects rolling out in the next release
>> as addressed earlier by Jack, like commit log and hint encryptions, as well
>> as, unified authentication...but secuirty labels aren't on anyone's radar
>> as a pressing "need." It's not something I've heard about as a
>> priority before anyway.
>>
>> Hope this helps!
>>
>> Cheers,
>> Dani
>>
>> On Friday, January 29, 2016, oleg yusim <olegyusim@gmail.com> wrote:
>>
>>> Jack,
>>>
>>> Thanks for your suggestion. I'm familiar with Cassandra documentation,
>>> and I'm aware of differences between DSE and Cassandra.
>>>
>>> Questions I ask here are those, I found no mention about in
>>> documentation. Let's take security labels for instance. Cassandra
>>> documentation is completely silent on this regard and so is Google. I
>>> assume, based on it, Cassandra doesn't support it. But I can't create
>>> federal compliance security document for Cassandra basing it of my
>>> assumptions and lack of information solely. That is where my questions stem
>>> from.
>>>
>>> Thanks,
>>>
>>> Oleg
>>>
>>> On Fri, Jan 29, 2016 at 10:17 AM, Jack Krupansky <
>>> jack.krupansky@gmail.com> wrote:
>>>
>>>> To answer any future questions along these same lines, I suggest that
>>>> you start by simply searching the doc and search the github repo for the
>>>> source code for the relevant keywords. That will give you the definitive
>>>> answers quickly. If something is missing, feel free to propose that it be
>>>> added (if you really need it). And feel free to confirm here if a quick
>>>> search doesn't give you a solid answer.
>>>>
>>>> Here's the root page for security in the Cassandra doc:
>>>>
>>>> https://docs.datastax.com/en/cassandra/3.x/cassandra/configuration/secureTOC.html
>>>>
>>>> Also note that on questions of security, DataStax Enterprise may have
>>>> different answers than pure open source Cassandra.
>>>>
>>>> -- Jack Krupansky
>>>>
>>>> On Thu, Jan 28, 2016 at 8:37 PM, oleg yusim <olegyusim@gmail.com>
>>>> wrote:
>>>>
>>>>> Patrick,
>>>>>
>>>>> Absolutely. Security label is mechanism of access control, utilized by
>>>>> MAC (mandatory access control) model, and not utilized by DAC
>>>>> (discretionary access control) model, we all are used to. In database
>>>>> content it is illustrated for instance here:
>>>>> http://www.postgresql.org/docs/current/static/sql-security-label.html
>>>>>
>>>>> Now, as per my goals, I'm making a security assessment for Cassandra
>>>>> DB with a goal to produce STIG on this product. That is one of the
>>>>> parameters in database SRG I have to assess against.
>>>>>
>>>>> Thanks,
>>>>>
>>>>> Oleg
>>>>>
>>>>>
>>>>> On Thu, Jan 28, 2016 at 6:32 PM, Patrick McFadin <pmcfadin@gmail.com>
>>>>> wrote:
>>>>>
>>>>>> Cassandra has support for authentication security, but I'm not
>>>>>> familiar with a security label. Can you describe what you want to
do?
>>>>>>
>>>>>> Patrick
>>>>>>
>>>>>> On Thu, Jan 28, 2016 at 2:26 PM, oleg yusim <olegyusim@gmail.com>
>>>>>> wrote:
>>>>>>
>>>>>>> Greetings,
>>>>>>>
>>>>>>> Does Cassandra support security label concept? If so, where can
I
>>>>>>> read on how it should be applied?
>>>>>>>
>>>>>>> Thanks,
>>>>>>>
>>>>>>> Oleg
>>>>>>>
>>>>>>
>>>>>>
>>>>>
>>>>
>>>
>>
>> --
>> Sent from mobile -- apologizes for brevity or errors.
>>
>
>
> --
> Sent from mobile -- apologizes for brevity or errors.
>

Mime
View raw message