Download as ppt, pdf, or txt
Download as ppt, pdf, or txt
You are on page 1of 47

Defending Against DDoS Attacks Using

Max-min Fair Server Centric Router


Throttles

David K.Y. Yau John C.S. Lu


CS Dept, Purdue University CS&E Dept,CUHK

Operating System Concepts 1.1


Motivations
 Internet is an open and democratic environment
 increasingly used for mission-critical work and
commercial applications.

 Many security threats are present or appearing


 Easy to launch, even for naïve users.
 need effective and flexible defenses to
detect/trace/counter attacks Ambitious goals
 Goals:
 protect innocent users;
 prosecute criminals

Operating System Concepts 1.2


Network Denial-of-service Attacks
 Some attacks quite subtle
 securing protocols and intrusion detection
(e.g., BGP, TCP-syn attack)
 at routing infrastructure, malicious dropping
of packets, etc (low-rate TCP)
 Others by brute force:
- flooding (e.g., UDP, valid Web Request)
 Cripples victim:
- precludes any sophisticated defense at
victim site
 Philosophical question: what is an “attacker”?
 Viewed as resource management problem

Operating System Concepts 1.3


Flooding Attack

Server

Operating System Concepts 1.4


Server-centric Router Throttle

 Installed by server when under stress,


at a set deployment routers
 can be sent by multicast

 Specifies leaky bucket rate at which


router can forward traffic to the server
 aggressive traffic for server dropped before
reaching server
 rate determined by a feedbak control
algorithm

Issues: (1) Which set of routers?


(2) What is the “proper” dropping rate?
Operating System Concepts 1.5
Router Throttle
Securely
installed by S

Throttle
Aggressive flow for S
To S

Throttle
To S’
for S’

Deployment router
C: Each victim has a leaky bucket for rate limit.
Small memory and computationoverhead!
Operating System Concepts 1.6
Key Design Problems
 Resource allocation: who is entitled to
what?
 need to keep server operating within load limits
 notion of fairness, and how to achieve it?
 Need global, rather than router-local,
fairness
 How to respond to network and user
dynamics (e.g., fluctuation of traffic)?
 Feedback control strategy is needed

Operating System Concepts 1.7


What is being fair?
 Baseline approach of dropping a fraction
“f”, say ½, of traffic for each flow won’t
work well
 a flow can cause more damage to other flows simply
by being more aggressive!

 Rather, no flow should get a higher rate


than another flow that has unmet
demands
 this way, we penalize “aggressive” flows only, but
protect the well-behaving ones

Operating System Concepts 1.8


Level-k Deployment Points
 Deployment points parameterized by an
integer k
 R(k) -- set of routers that are either k
hops away from server S, or less than k
hops away from S but are directly
connected to a host
 Fairness across global routing points R(k)

Operating System Concepts 1.10


Level-3 Deployment

Server

Operating System Concepts 1.11


Feedback Control Strategy

 Hysteresis control
 high and low water marks for server load, to
strengthen or relax router throttle

 Additive increase/multiplicative decrease


rate adjustment
 increases when server load exceeds US, and
decreases when server load falls below LS

 throttle removed when a relaxed rate does


not result in significant server load increase

Operating System Concepts 1.12


Fairness Definition

 A resource control algorithm


achieves level-k max-min fairness
among the routers R(k) if the
allowed forwarding rate of traffic
for S at each router is the router’s
max-min fair share of some rate r
satisfying LS  r  US

Operating System Concepts 1.13


Fair Throttle Algorithm

Operating System Concepts 1.14


Example Max-min Rates (L=18, H=22)
18.23
24.88 6.65

6.25
0.22 0.22
14.1
15.51
59.9 0.01
Server 6.25
17.73
6.25 1.40 17.73
20.53 0.61

0.95

0.95 0.61
Operating System Concepts 1.15
Interesting Questions

 Can we preferentially drop attacker


traffic over good user traffic?
 Can we successfully keep server
operating within design limits, so that
good user traffic that makes it gets
acceptable service?
 How stable is such a control algorithm?
How does it converge?

Operating System Concepts 1.16


Algorithm Evaluation

 Control-theoretic analysis (fluid analysis)


 algorithm stability and convergence under
different system parameters
 Packet network simulations (packet level
analysis)
 Test under UDP and TCP traffic. Also test
with Web traces
 System implementation (the real thing,
baby !!!)
 deployment costs

Operating System Concepts 1.17


Control-theoretic Model
Throttle signal from victim

Step size

Adjusted traffic from source i

When throttle signal is high, server is underloaded.


When throttle signal is low, server is overloaded.

ANALOGY!!!
Operating System Concepts 1.18
Feedback Control Model (Us=1750;Ls=1650)
Constant
Source of 20

Constant
Source of 30

Constant
Source of 25

Constant
Source of 4000

Constant
Source of 2800

Operating System Concepts 1.19


Output for good traffic (total from source 1)

Operating System Concepts 1.20


Output for attack traffic (total from source 5)

Operating System Concepts 1.21


Output for attack traffic (total from source 6)

Operating System Concepts 1.22


Total traffic to server (Us=1750;Ls=1650)

Operating System Concepts 1.23


Case 2: variable attack traffic
(Us=1750,Ls=1650)

Square Pulse

Operating System Concepts 1.24


Output of attack traffic 1

Operating System Concepts 1.25


Output of attack traffic 2

Operating System Concepts 1.26


Total traffic to server (Us=1750;Ls=1650)

Operating System Concepts 1.27


Feedback Control Model
(sources and server)

Operating System Concepts 1.28


Feedback Control Model
(server throttle signal)

Operating System Concepts 1.29


Feedback Control Model
(sources process throttle)

Operating System Concepts 1.30


Throttle Rate (L=900; U=1100)

Operating System Concepts 1.31


Server Load (L = 900; U = 1100)

Operating System Concepts 1.32


Throttle Rate (U = 1100)

Operating System Concepts 1.33


Server Load (U = 1100)

Operating System Concepts 1.34


Throttle Rate (L=1050;U=1100)

Operating System Concepts 1.35


Server Load (L=1050; U=1100)

Operating System Concepts 1.36


NS2: UDP Simulation Experiments
 Global network topology reconstructed
from real traceroute data
 AT&T Internet mapping project: 709,310 traceroute paths,
single source to 103,402 other destinations
 randomly select 5,000 paths, with 135,821 nodes of which
3879 are hosts
 Randomly select x% of hosts to be
attackers
 good users send at rate [0,r], attackers at rate [0,R]

Operating System Concepts 1.37


20% Evenly Distributed Aggressive (10:1)
Attackers

Operating System Concepts 1.38


40% Evenly Distributed Aggressive (5:1)
Attackers

Operating System Concepts 1.39


Evenly Distributed “meek” Attackers

Operating System Concepts 1.40


Deployment Extent

Operating System Concepts 1.41


NS2: TCP Simulation Experiment

 Clients access web server via HTTP 1.0 over


TCP Reno
 Simulated network subset of AT&T traceroute
topology
 85 hosts, 20% attackers
 Web clients make request probabilistically with
empirical document size and inter-request time
distributions

Operating System Concepts 1.42


Web Server Protection

Operating System Concepts 1.43


Web Server Traffic Control

Operating System Concepts 1.44


System Implementation

 On Linux router
 loadable kernel module
 CPU resource reservation
 Deployment platform
 Pentium 4/2G Hz PC
 multiple 10/100 Mb/s Ethernet
interfaces

Operating System Concepts 1.45


System Implementation: cont
 OPERA: An Open-Source Extensible Router
Architecture

http://www.cse.cuhk.edu.hk/~cslui/ANSRlab/software/o
pera/
 A Linux-based package for implementing a software
programmable router architecture with the aim to facilitate
networking experiments for the research community. Using
this architecture, one can dynamically load new extension
and services into the programmable router. Some interesting
extensions include QoS support and traceback of DDoS
attacks.)

 Dynamic module loading


 Resource reservation
 General extension framework
 Secured Communication
Operating System Concepts 1.46
Future Work

 Offered load-aware control algorithm for


computing throttle rate
 impact on convergence and stability
 Policy-based notion of fairness
 heterogeneous network regions, by size,
susceptibility to attacks, tariff payment
 Selective deployment issues
 Impact on real user applications
 Defense for other forms of DDoS like the
reflector attack, BGP cascading
failure..etc.

Operating System Concepts 1.48


Conclusions

 Extensible routers can help improve network


health
 Presented a server-centric router throttle
mechanism for DDoS flooding attacks
 can better protect good user traffic from aggressive
attacker traffic
 can keep server operational under an ongoing attack
 has efficient implementation

Operating System Concepts 1.49

You might also like