Transcript Botnets
Collection of connected programs
communicating with similar programs to
perform tasks
Legal
IRC bots to moderate/administer channels
Origin of term botnet
Illegal
Bots usually added through infections
Communicate through standard network
protocols
Named after malware that created the
botnet
Multiple botnets can be created by same malware
▪ Controlled by different entities
“Bot master” can control entire group of
computers remotely through Command and
Control(C&C) system
Botnets used for various purposes
Distributed Denial of Service Attacks(DDOS)
SMTP mail relays for spam
Click Fraud
▪ Simulating false clicks on advertisements to earn money
Theft of information
▪
▪
▪
▪
Application serial numbers
Login information
Financial information
Personal information
Bitcoin mining
Three main connection models
Centralized
P2P-based
Unstructured
Central point(server) that forwards messages
to bots
Advantages
Simple to implement
Customizable
Disadvantages
Easier to detect and destroy
Most botnets use this model
Mainly used to avoid problems with centralized
model
Does not use server as central location
Instead the bots are connected to each other
Advantages
Very hard to destroy
Commands can be injected at any point
Hard for researchers to find all bots
Disadvantages
Harder to implement and design
Bots will not actively contact other bots or
botmaster
Only listens for incoming connections
Botmsater randomly scans internet for bots
When bot is found botmaster sends encrypted
commands
Botnets use well defined communication protocols
Helps blend in with traffic
Protocol examples
IRC
▪ Most common
▪ Used for one-to-many or one-on-one
HTTP
▪ Difficult to be detected
▪ Allowed through most security devices by default
P2P
▪ More advanced communication
▪ Not always allowed on network
Two main detection methods
Signature-based
▪ Relies on knowing connection methods
▪ Cannot detect new threats
Anomaly-based
▪ Relies on anomalies from base-line traffic
▪ High false-positive rates
▪ Not useful in cases where base-line traffic cannot be
established
Malware writers constantly looking for new
ways to avoid detection
Recent botnets employ new methods to
avoid detection
Fast flux
Domain flux
Use a set of IP addresses that all correspond
to one domain name
Use short TTL(Time To Live) and large IP
pools
Can be grouped in two categories.
Single flux
Double flux
Domain resolves to different IP in different
time ranges
User accesses same domain twice
First time DNS query returns 11.11.11.11
TTL expires on DNS query
User performs another DNS query for domain
DNS server returns 22.22.22.22
More sophisticated counter-detection
Repeated changes of both flux agents and
registration in DNS servers
Authoritative DNS server part of fluxing
Provides extra redundancy
Critical step in detecting fast flux network is
to distinguish fast fluxing attack
network(FFAN) and fast fluxing service
network(FFSN)
All agents in FFSN should be up 24/7
Agents within FFAN have unpredictable alive time
▪ Botmaster does not have physical control over bots
Two metrics developed to distinguish these
Average Online Rate(AOR)
Minimum Available Rate(MAR)
Uses AOR and MAR to track FFANs and FFSNs
Broken up into four components
Dig tool
▪ Gather information and add new IP addresses to database
Agents monitor
▪ Sends HTTP requests records response
IP lifespan records database
▪ Stores service status
Detector
▪ Judges between FFAN and FFSN by using AOR and MAR
To avoid single point of failure domain flux
was created
Uses a set of domain names that are
constantly, and automatically, generated
Occasionally correspond to IP address
Bots and server both run domain name
generation algorithm.
Bots try to contact C&C server by using
generated domain names
If no answer is received at one, it moves on
Torpig was botnet that used domain flux
Eventually taken over by researchers
First calculated domain names by current
week and current year
“weekyear.com” or “weekyear.net”
If those fail it moves on to calculated the daily
domain
If all other methods fail, a Torpig bot will try
to connect to a hard-coded domain within its
configuration files
Reverse-engineering domain generation
algorithm not always possible
Only a few domains will resolve to IP addresses
One detection method is to watch DNS query
failures
Small percentage will be user error/poor configuration
Larger part of errors will be from malicious activity
With enough data one should be able to find
patterns in DNS query errors
Fast Flux networks mitigated by blacklisting
domain name associated with flux
Contact registrar
ISP block requests in DNS
ISP monitor DNS queries to domain
Domain flux is harder to mitigate
In order to register domain names before attackers
one must know the algorithm used
Automated techniques to block DNS queries not
always accurate
Registrars used by attackers usually do not listen to
abuse reports
BredoLab
Created May, 2009
30,000,000 bots
Mariposa
Created 2008
12,000,000 bots
Zeus
Banking credentials for all major banks
3,600,000 bots in US alone
Customizable