Quoting Matthew F. Caldwell ([EMAIL PROTECTED]):

>         I think you will find that, out of those products 90% plus where
> created by non-security professionals.  And those created by security
> companies, security analyst had very little input into the creation of
> how things should be done.  And about 2% do what they say they can do. 

I think the most important question here is: What is a _Risk_. Sure,
Incidents are easily defined, Vulnerabilities, too, sometimes expert
systems even analyze Policies and Standards based on a probability
database, but Risk is much more than that.

I aggree with you. Would I want to travel the Amazon Rainforest, I'd
certainly not hire the someone who wrote the book on "Growth of Trees
in the Amazon Region" but someone who's been there for a while and
seen the dangers of this particular area with his own eyes.
Professionalism, multiple years of experience and "been there, done
that" are definitely the most important factors when Risk is to be
analyzed, qualified and quantified.

But this is where my questions start. Maybe we're discussing two
different things here, but while I can see how a tool would assess
vulnerabilities (Nessus, etc.) and incidents (snort et al), I fail to
understand how Risk is being assessed. Risk is a predictive term, it's
bound to psychological factors, technical incertainities (such as
routing loops, IP-address similarities between the assessed site and a
more likely target, joe-job likelyhood, etc.), policies and their
inforcement, etc.

As I wrote in my email to you, I am certain you guys have a great
product. I just don't see how "Risk" can be assessed automatically or
even with the help of automated tools. Sure, technical aid is needed,
I personally like to use netcat, lynx (for bugtraq-queries, etc.),
telnet and a set of python scripts, but so far I have not found this
magical "one size fits all" software that would assess Risk for me.

As a good example let me tell the story of this week: There's this
company that just recently conducted a "Risk" assessment via a third
party. Said third party used several scanners to scan for
vulnerabilities and installed a sniffer at the central Catalyst
SPANning the whole internal network and DMZ. One of the findings was
"Your telnet port is open, this is a HIGH RISK". The finding referred
to the Cisco 75xx router in front of the network.

Now, is this

a) A Risk? Not necessarily. What the tool failed to assess was the
question if anybody ever used this telnet port. And if there were rate
limits in place to prevent someone from connecting a couple of
thousand times to this port exploiting a remote DoS vulnerability in
this special IOS version. Unless someone was stupid enough to log in
to the route via Telnet (there's a console attached to it, too,
sporting SSH login capabilities) and NOT to use the SecureID setup, I
don't see a big Risk here.

b) A vulnerability? Not necessarily. If SecureID can be positively
proven predictable (I saw a few reference implementations which failed
for these 4-digit-keypunch versions), this looks different, sure, same
if someone finds a buffer-overflow or something else in Cisco's IOS
telnet implementation, but in general ... well.

c) An incident? Was the port open for everyone before? Are there
access-lists in place to prevent this kind of connection, usually?

BUT, and here it comes, there IS a high risk associated with this
router. A risk, no tool in the world could assess - the router was
running a flawed version of IOS which allowed for the arbitrary
injection of falsified BGP statements. An attacker with the right
tolls could easily falsify some routes, pose as zonelabs.com, wait
until some of the internal users download a supsected "upgrade" to
their Personal Firewall Software and - boingo - roam the internal
network.

In the world of vulnerability scanning, the process of obtaining
version information on open ports and matching them against a list of
patterns, tools have their place. Risk assessments, however, go so
much deeper into a predictive and psychological area that tools can
not even begin to compliment the work that needs to be done there.

Risk Analysis, at least for me, is the process of correlating human-
and technical data with predictive statements, a good portion of
psychological evaluations, business requirements, "best practice"
analysis, and a lot of knowledge about the inner workings of the
infrastructure in question.

I guess, what i boils down to is: You're right. In the hands of a
skilled Risk Analyst a good tool can do a lot of good things. But
that's only if the same analyst could have obtained the same results
with the help of netcat, tcpdump and perl. A tool cannot replace
expertise and it can only so much compliment existant expertise. If
you found out how to increase the aid a tool can offer, cool. If it's
not too intrusive, I am sure some of us would like to learn about it
and how we could use it, in technical terms not the marketing way :)

jonas

Reply via email to