Transcript Document
Controlling High Bandwidth
Aggregates in the Network
1
Goals:
Handle congestion
Limit DoS attacks
Allow flash crowds
Identify traffic aggregates
Subset of flows responsible for congestion
Integrate provider policy
Allow provider to configure drop
mechanism
2
Related Work
IP Traceback
Tries to find source of attack
Ingress/Egress Filtering
ISP filters packets with fake source addresses
Input debugging
Uses signatures to filter attack traffic
Scheduling:
Fair Queuing
Deficit Round Robin
3
ACC Design
Apply congestion control to aggregated
traffic
Two levels of control:
Local:
Identification
Control
Global:
Pushback*
4
Issues
Collateral damage
Legitimate traffic may be inaccurately identified
and restricted
Routers may become synchronized and
simultaneously detect congestion
Insert jitter into monitoring interval
How to ensure fairness of flows
Separate identification and control
Use RED to manage queue drops
5
Application to DoS attacks:
Finding Aggregates
Match destination of each dropped IP
packet with longest matching prefix in
routing table
Periodically find most frequent prefix
See if destinations match longer prefix
E.g. maybe all dropped packets go to some
specific host.
6
Application to DoS attacks:
Rate Limiting*
Let:
wo be output bandwidth
wi be total input bandwidth
wb be bandwidth of aggregate
desired drop rate be 20%
Two conditions:
wi – wb <= 1.2*wo
wi – wb > 1.2*wo
7