Being the one eyed man

Incident Detection:
MacGyver Style
Ben Jackson, Mayhemic Labs
April 18th, 2012
Improving the Kill Chain
Moving on from Prevention
Finding Data in Unexpected Places
Canaries, Claymores, and Labyrinths
Simple Anomaly Detection
Thoughts expressed here are neither the opinions or beliefs of my employer.
Kill Chain
 The “Kill Chain”
 Developed by the United States Air Force
to describe the concept of target
identification, force dispatch to target,
decision and order to attack the target,
and finally the destruction of the target.
Kill Chain
Find – Locate the Target
Fix – Identify the Target
Track – Watch the Target
Target - Determine Strategy
Engage – Eliminate the Target
Assess – Determine if it worked
Things that we’re good at…
Things that we’re not good at…
Prevention is a dirty word
 Stop focusing on prevention
 Right now, your environment is
getting attacked with malware made
earlier today
 Plus, it hasn’t even been submitted to
your AV vendor
 You aren’t going to prevent that
 Prevention should be part of your
strategy, and not your strategy itself
How bad are we?
 Verizon DBIR sums it up
 92% of all data breach incidents were
discovered by a third party
 85% of all breaches took three or more
weeks to discover
 Both of these figures were up 6% from
last year
 Let’s face it, we’re doing it wrong
How can we improve?
 Lots of excuses:
“I don’t have tools”
“I don’t have manpower”
“I don’t have a budget”
“It’s too difficult”
Go cry, emo CISO
I know what you’re thinking…
 But hear me out!
Detection is the new black
 Accept you are going to
get compromised
 Liberating, isn’t it?
 Let’s work on rapidly
detecting compromises
when they happen
Let’s do it MacGyver Style!
 Watched far too much MacGyver in
my formative years
 When faced with an impossible task:
Take stock of your tools
Work the problem
Be sure to carry a swiss army knife
 Or in my case, multi-tool
Data: It’s freaking everywhere
 Some background
 In my previous life, I was a developer in
public health statistics
 “Diet Bioinformatics”
 Firm believer in thinking by the numbers
 With enough data, you can rule the
 I, for one, welcome our new Google
Data: It’s freaking everywhere
 Data is in places that you wouldn’t
 Your network keeps information just
to run smoothly
 Let’s use that data to find abnormal
Case Study
 Wayback Machine set for March 2010
 Problem: Difficult to use blacklists to
check traffic to malware sites
 Firewall logs available, but far too many
sites to check
 Equivalent to trying to identify needles in
a very large haystack. Plus, some of the
needles may not actually be there.
Budget limited tools available…
Time to get ingenious!
A Quick DNS Tutorial
A Quick DNS Tutorial
Solution: DNS Scraping
 Didn’t care about the sites users
weren’t visiting
 Only cared about ones they were visiting
 Figured by leveraging blacklists and
non recursive DNS lookups, it could
be determined what sites other hosts
were querying
How DNS Scraping Works
How DNS Scraping Works
How DNS Scraping Works
Phase 3: Profit!
 If someone was querying a
blacklisted site, further research was
 The number of sites that needed to be
searched was cut down from 2000+ to 5
or less
 This was awesome
 Mostly because I’m lazy
It’s far from perfect
 DNS records with extremely low time
to live (TTL) values give us a very
small window to locate them
 Fast flux can be as low as 5 seconds
 Number of queries can total to over
 More DNS servers? More queries
 But, better then nothing!
This statement is false!
 Everyone gets annoyed at false
 Anti Malware Software
 Intrusion Prevention Systems
 Everyone should be worried about
false negatives
 What’s happening in your network that
isn’t being detected?
 More importantly, how do we detect it?
Old and busted: Honeypots
 Honeypots are great
 They provide great information for
 Honeypots also are bad
 Lots of risk, little reward
 They’re a time suck to maintain
 Thankfully, you never needed to worry
about patching them
New Hotness: Canaries
 Somewhat like honeypots
 Well monitored
 Extremely permissive
 No one should be using them
 Only not…
 Fully patched
 Not designed to be compromised
 Not externally facing
The best of both worlds
 Give you an early warning for scans
within the environment
 Fully patched and managed, so less
likely to get compromised
 Unless someone drops some 0 day
 Permissive and monitored, so if
someone pokes at it, you know
Canaries aren’t just computers
 Pen Testers <3 Rogue APs
 OK, let’s give an attacker a rogue AP
 Unencrypted
 Default SSID
 Merrily handing out DHCP addresses
 Log everything
 Runs on a computer set up
to broadcast as a rogue AP
 Upon handing out a DHCP
address, runs a port scan on the
client computer
 Generates an alert message and
mails it along with the scan to an email address.
When the trap is sprung…
What good does it do?
 Primarily captures two kinds of people
 Bad Actors
 “Curious” Users
 Either way, folks you would want to
know are trying to poke around
 This can provide as little or as much
access at you want
Do ye seek a grail?
 Anomaly detection is the
holy grail of analysts
 Also incredibly hard to do
 What is “normal” for your network?
 More importantly, are we sure what
this is normal is actually normal?
 Are we owned already?
Do ye seek a grail? Maybe not…
 Works great on paper
 My SQL Server should only be doing SQL
Server-y things
 However, not to much in practice
 Wait, why is that SQL server FTPing out
to Sweden?
 Oh… Anti-Virus update…
 Or worse, it’s using a CDN to grab software
Let’s adjust the goalposts
 However, let’s try to apply this to
other areas
 Where does your user base surf to on
an average day?
 Both work related and non-work related
 Does it really change that much?
 People like routines
The good, the bad, and the…
The stuff we know is good
stuff we
know is
Devious Deviations
 Deviations from the baseline will give
“interesting” stuff
 The “Good”
 Legitimate traffic (Shopping, Research, etc)
 The “Bad”
 Exploit Kits, Crimeware, BlackHat SEO
 The “Ugly”
 Errr… “Not work safe” material
Why reinvent the wheel?
 Marcus Ranum created “NBS” in 2003
 “Never Before Seen”
 Simple tool that implements
something that we’re still trying to
get done correctly today
 “If we’ve never seen it before, by
definition… it’s an anomaly”
 Seen it before? No? Log it.
Network NBS
 Let’s apply that to network events
 Has that event happened before?
 No? Log it!
 Analysis is left to be done by the
 Only you will know what is normal for
your network
 Simple framework to detect “never
before seen” network events
 Currently processes firewall and
DHCP events
 Next up: Proxy and DNS
 Something we haven’t seen? Alert to
 Named after the Soviet ICBM early
warning system
Some shortcomings
 Doesn’t provide any context
 You’re on your own, bub
 Lights up like a Christmas tree for
abnormal, but authorized activity
 New server? PANIC!
 All the data extraction from logs is
left to you
 Handles JSON via ZeroMQ
In summary…
 There are lots of bits out there on
your network, harness them to your
 Use these techniques in your own
 Remember, when faced with an
impossible task, inventory your
And ask yourself…
Ur Feedback, I haz it?
Contact Info
Ben Jackson
Twitter: @innismir