RE: Re: Concepts: Security and Obscurity

["Craig Wright" <Craig.Wright () bdo com au>]

Hi Florian,
Well I have to state that you are lucky, I still get a subnet at home
that is no longer connected scanned about once a second from new
addresses. Than again, maybe it is me and I attract this ;). The
anecdote you have provided is a start, but it needs to be made
scientifically sound. Also, the results demonstrate a record of the
scans and not the survival.

There is a little more than the number of scans to be considered. I am
sure that none of us really cares (in more than general interest) how
many scan we receive. What is important is to check the survivability
without all the confounding variables which people keep adding. 

This needs to be done in a manner which is statistically valid, i.e. not
anecdote from either side of the table and it has to be replicable.

A real world Honeypot experiment would suffice. I have setup Linux and
Windows hosts in the past to check them attractiveness of these, but in
this case it would be to see the general attractiveness of a service and
than how long it took to be compromised. 

So, as an example (one amongst many possibilities) a group of hosts (or
virtual hosts) could be setup and run with SSH on half on the standard
ports and the other half on random ports. Leave these as a Honeypot and
time the survival - i.e. the time to initial compromise. Repeat the test
a number of times till a valid statistical sample is obtained.

The hosts could be mixed (i.e. Linux, Solaris etc), but they have to be
mirrored in the standard vs. obscured configs (eg 3 Linux SSH, 3 Linux
SSH on TCP 443; Solaris SSH and Solaris SSH on TCP 443) with equal
patching.

This experiment would remove the confounding variables and provide a
means of actually measuring the level of additional protection as a
factor of time provided through the addition of an "obscurity" factor
and would categorically answer the question - does obscurity provide
additional security.

The way to test the results would be to take the means of the survival
times from both the standard (SSH on TCP 22) and Obscured hosts (TCP not
equal to 22). The results from the hosts would than have to be modelled
and a simple ANOVA based test of the 2 hypothesis:
Ho:     There is no additional security through obscurity
Ha:     Obscurity gives some level of additional security
        ... Could lead us to the answer. 

In this, if there is enough evidence of a variation in survival times,
then Ha is valid and you can state that there is an improvement in
security from adding a layer of obscurity. If there is not sufficient
evidence, than Ho the Null hypothesis stands and the premise that we
have no gain in security through obscurity stands.

This is what scientific proof is about. My offer of the donation stands
and I will even help anyone who wishes to do this with time in analysis
and experimental design in an unbiased manner.

Doing this and writing it up should provide the tester a document that
they could publish, so there is a little more than proving me wrong in
it.

So again - any takers? Who wishes to attempt to prove me wrong and get
to categorically state that obscurity is a valuable tool in the security
professional's arsenal? Of course that also means you have to have an
open mind and you may be demonstrating that there is no evidenced to
support the claim the obscurity adds anything. 

Regards,
Dr Craig Wright



Craig Wright
Manager of Information Systems

Direct +61 2 9286 5497
Craig.Wright () bdo com au

BDO Kendalls (NSW)
Level 19, 2 Market Street Sydney NSW 2000
GPO Box 2551 Sydney NSW 2001
Fax +61 2 9993 9497
www.bdo.com.au

Liability limited by a scheme approved under Professional Standards Legislation in respect of matters arising within 
those States and Territories of Australia where such legislation exists.

The information in this email and any attachments is confidential.  If you are not the named addressee you must not 
read, print, copy, distribute, or use in any way this transmission or any information it contains.  If you have 
received this message in error, please notify the sender by return email, destroy all copies and delete it from your 
system. 

Any views expressed in this message are those of the individual sender and not necessarily endorsed by BDO Kendalls.  
You may not rely on this message as advice unless subsequently confirmed by fax or letter signed by a Partner or 
Director of BDO Kendalls.  It is your responsibility to scan this communication and any files attached for computer 
viruses and other defects.  BDO Kendalls does not accept liability for any loss or damage however caused which may 
result from this communication or any files attached.  A full version of the BDO Kendalls disclaimer, and our Privacy 
statement, can be found on the BDO Kendalls website at http://www.bdo.com.au or by emailing administrator () bdo com au.

BDO Kendalls is a national association of separate partnerships and entities.

-----Original Message-----

From: Florian Rommel [mailto:frommel () gmail com] 
Sent: Monday, 16 April 2007 4:24 PM
To: Craig Wright
Cc: levinson_k () securityadmin info; security-basics () securityfocus com
Subject: Re: Re: Concepts: Security and Obscurity

Hello to everyone. I actually think this discussion is very fruitful.
It provides a good way of proofing/unproofing the concept.

I would like to add one thing though.

Craig, your argument is that you need proof of declined hack attempts.

I would like to take the simple example of SSH. SSH bruteforcing is
still going through the roof. I have a DSL connection and when my SSH
server was on port 22 I received about 50-100 false logins per day.
Not much you say and I agree but for a home connection it is. However
I then moved ssh to port 443 (SSL) as I am not running a secure
webserver. I need the standard port access to be able to ssh from work
to home and I have had 0 bruteforce attempts per day now. Would that
not qualify as some sort of Security through obscurity?

//Flosse
http://blog.2blocksaway.com

PS: I haven't gotten a 443 attempt either though port 80 does get
"accessed" quite a lot.



On 4/15/07, Craig Wright <Craig.Wright () bdo com au> wrote:
> No Karl, you have not provided mathematical proof or something that
serves to prove your point.
> I stated survivability - the number of scans by service not the key to
this test. The number of scans and attacks are differnt factors. A scan
is not an attack. Now as you state, proving a negative for all cases is
near impossible, but you have to prove the positive, and this is not
being done. You have not as yet proved proof.
> As I have stated, please provide some proof. Demonstrate how obscurity
works. Either provide an experiment or a peer reviewed paper.
Speculation is not proof. You keep stating that there are other cases to
my proofs and I have stated that disproving a negative is often a futile
effort. Please provide real proof and not just state that your views are
proof.
> The number of scans example is not a survivability case and is not
proof for your assertion.
> Craig
>
>
>
> Craig Wright
> Manager of Information Systems
>
> Direct +61 2 9286 5497
> Craig.Wright () bdo com au
>
> BDO Kendalls (NSW)
> Level 19, 2 Market Street Sydney NSW 2000
> GPO Box 2551 Sydney NSW 2001
> Fax +61 2 9993 9497
> www.bdo.com.au
>
> Liability limited by a scheme approved under Professional Standards
Legislation in respect of matters arising within those States and
Territories of Australia where such legislation exists.
> The information in this email and any attachments is confidential.  If
you are not the named addressee you must not read, print, copy,
distribute, or use in any way this transmission or any information it
contains.  If you have received this message in error, please notify the
sender by return email, destroy all copies and delete it from your
system.
> Any views expressed in this message are those of the individual sender
and not necessarily endorsed by BDO Kendalls.  You may not rely on this
message as advice unless subsequently confirmed by fax or letter signed
by a Partner or Director of BDO Kendalls.  It is your responsibility to
scan this communication and any files attached for computer viruses and
other defects.  BDO Kendalls does not accept liability for any loss or
damage however caused which may result from this communication or any
files attached.  A full version of the BDO Kendalls disclaimer, and our
Privacy statement, can be found on the BDO Kendalls website at
http://www.bdo.com.au or by emailing administrator () bdo com au.
> BDO Kendalls is a national association of separate partnerships and
entities.
> ________________________________
>
>
> From: listbounce () securityfocus com on behalf of
levinson_k () securityadmin info
> Sent: Sat 14/04/2007 2:53 PM
> To: security-basics () securityfocus com
> Subject: Re: Re: Concepts: Security and Obscurity
>
>
>
>
> > > > In a test that is determined scientifically and without bias,
> > > > the results show that obscurity does not reduce risk and is thus
not a
> > > > benefit.
> > >
> > > I'd love to see such a study.  It does not exist.
> >
> > Actually, I believe the honeynet project compiles statistics on how
well
> > obfuscation of ports works, and last I read they have decided it
makes
> > no difference at all. Services running on nonstandard ports are
> > attacked just as much as services on standard ports over time.
>
> It is easy to demonstrate this is false.
>
> http://www.incidents.org/top10.html
>
> The top ports receiving unsolicited scans are all well known,
published server ports:
> TCP 8080
> TCP 2967 (symantec)
> TCP 445
> TCP 139
> TCP 1434
> TCP 5900
>
> Put a server on any other port, and your number of attacks is going to
be demonstrably lower than the numbers above.  Hence, reduced risk by
obscurity.
> Besides, given that so much hacking nowadays is financially motivated
and aims at compromising the most systems starting with low hanging
fruit, I don't see how could anyone could prove that non-standard ports
are attacked just as often as standard ports.
> Anyways, obfuscation of ports is just one example of obscurity, and
any study of that countermeasure would not be applicable to all forms of
obscurity.  That's why I objected to the absurd claim that it has been
mathematically proven that all forms of obscurity are ineffectual, and
objected to the attempts here to point out some examples of bad
obscurity in order to prove that obscurity is universally bad.
Certainly some forms of obscurity are ineffectual.  I only need to point
out one beneficial form of obscurity to invalidate such universal
statements.  People talking about math should realize my side is more
likely to be proven true.
> There, I gave mathematical data suggesting that obscurity
significantly reduces the number and type of threats it was intended to
reduce.  Let's see some statistics proving otherwise.
> > > > Obscurity does not work.
> > >
> > > It is impossible for you to make that assertion for all
> > environments and situations.
> >
> > Yes it is possible to make that assertion, based on logic and hard
math.
> > Security has nothing at all to do with raw numbers of break in
> > attempts,
>
> Incorrect.  Security is based on risk management and (quantitative)
risk assessment, which are mathematical formulas that evaluate the
likelihood of certain risks occurring in a given year, e.g. raw numbers
of break in attempts.  Furthermore, risk assessment, while mathematical,
is pretty meaningless unless you apply it to specific situations,
because the value, threats and existing countermeasures of a particular
system are variables that have to be known and inserted into the
mathematical formula.  That's why I say you cannot assert that obscurity
is never a (cost) effective measure at reducing risk.
> Obscurity absolutely can and often does reduce certain kinds of risks,
such as risk of script kiddies and viruses, frequently at very low cost.
I can't see how anyone can debate that point.  Though some here clearly
do not see any value
> > and everything to do with how resilient a system is to any
> > and all attacks.
>
> That's not how security and countermeasure evaluation, e.g.
quantitative risk assessment, work.  Countermeasures are designed to
mitigate JUST SPECIFIC THREATS, not all of them.  It is meaningless to
evaluate countermeasures by including threats that they were never
designed to mitigate.  Firewalls don't protect against social
engineering, but that doesn't mean you don't need one.
> > The "obscurity factor" is utterly irrelevant because
> > it has no impact what so ever on actual security. Using offered
> > examples, if your passwords are good ones it makes absolutely no
> > difference how many times an attacker tries to guess them because
they
> > simply can't make enough attempts in any sane time frame to do any
> > damage. Inversely, a single attempt is all it might take to "crack"
a
> > weakly protected system regardless of what port it's made on. So the
> > only security one could possibly gain by limiting the numbers of
> > attempts is of type "false sense".
>
> Not true.  It is an obvious truism that most all computers, especially
those on the Internet, are going to be vulnerable to unpatched zero day
vulnerabilities from time to time.  Once a vulnerability is exploited by
a network worm or easily downloadable script tool, your likelihood of
being compromised (a key component in quantitative risk assessment)
increases.  If you change the port on which your server listens, you
evade those attacks, and your likelihood of being compromised decreases
significantly.

> Please note here that by your purely theoretical definition, the
system is just as secure in both cases, because its configuration and
resistance to attack have not changed at all.  And yet, in the real
world, the system has a reduced risk and/or reduced number of compromise
events (which is the key result in quantitative risk assessment formulas
used to judge security).
> > conclusion that it can't be any other way. Obscurity carries with it
> > precisely as much potential for disaster as it does its ability to
"hide
> > something". That direct relationship exists by the very definition
of
> > obscurity.
>
> Most of the supposed dangers, risks and costs of obscurity are
actually risks of incompetent administration and failures of other
recommended security countermeasures such as the system procedures and
configuration being documented.  If your sysadmin assumes a system is in
the default configuration and takes a damaging action based on that
assumption, that's arguably not the fault of obscurity, and that damage
would arguably be just as likely to happen without obscurity, when you
have an incompetent sysadmin plus inadequate documentation.
> > And before we meander off into an endless debate about "would have"
and
> > "should have", I'll point out that all that is irrelevant. Obscurity
> > adds far more complexity than it affords protection, and no amount
of
> > after the fact  tail chasing can change the fact that this is a bad
> > thing at its core.
>
> Another broad, unsupportable generalization.  Tell me how something
like changing an FTP banner adds prohibitively costly complexity.
Obscurity includes a lot of different things.
> > This is the brittleness experts warn you about. It's a real life
issue,
> > not some theoretical mumbo-jumbo. By performing tasks in
"nonstandard"
> > ways you're as likely to confound the good guys as the bad. Not only
> > does obscurity not work, if it has any real effect at all it's
> > more likely to be a negative one than not. :(
>
> Again, quantitative risk assessment comes to the rescue.  Risk
assessment is an example of theory that is useful in the real world.
When using risk assessment to evaluate whether or not a countermeasure
is beneficial, you quantify and compare the amount that risks go up and
down.  You are not using or demonstrating mathematics when you state
that the increased risk/cost of obscurity's complexity outweighs the
other security risks that obscurity decreases.  Are you jumping to
conclusions, or do you have data to show that proves that in most all
environments, systems and obscurity-related countermeasures,
> > There
> > may be brief respites and fluctuations, but they're invariably
> > discovered and quite often attacked even harder than services on
> > standard ports, for obvious reasons.
>
> I don't see how that's very likely.  Putting hundreds of thousands of
servers on the same nonstandard port would not be a good implementation
of obscurity.  Attacking a poor implementation of anything is not really
relevant to whether or not a good implementation of it has merit.
> Besides, unless you're talking hundreds of thousands of systems using
the same non-standard port, you're still pretty much talking about
determined human attackers.  I thought I made it clear that obscurity is
not intended as a countermeasure to determined human attackers, social
engineering, earthquakes, etc.
> kind regards,
> Karl Levinson
> http://securityadmin.info