Presented by Pallavi Kasula Background Autonomous systems AS Border Gateway Protocol BGP Internet Censorship Decoy Routing Routing Around DecoysRAD Autonomous SystemAS Internet Comprises of interconnected Autonomous Systems ID: 718089
Download Presentation The PPT/PDF document "No Direction Home: The True cost of Rout..." is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
Slide1
No Direction Home:The True cost of Routing Around Decoys
Presented by :
Pallavi KasulaSlide2
Background
Autonomous systems (AS)
Border Gateway Protocol (BGP)
Internet Censorship
Decoy Routing
Routing Around Decoys(RAD)Slide3
Autonomous System(AS)
Internet
Comprises of interconnected Autonomous Systems
Autonomous System:
Collection of Networks with Same routing policy
Usually under single ownership, trust and administrative controlSlide4
BGP -Border Gateway Protocol
Designed to exchange routing and reachability information between autonomous systems (AS) on the Internet.
BGP is the path-vector protocol
provides routing information for autonomous systems on the Internet via its AS-Path attribute
Shortest AS_Path, Multi_Exit_DiscSlide5
Internet Censorship
the control or suppression of what can be accessed, published, or viewed on the Internet.
The extent of Internet censorship varies on a country-to-country basis
Content suppression methods include
Internet Protocol(IP) address blocking
DNS Name filtering and redirection
Circumvention using Proxy Server has been in use which needs client to connect to a specific IP address.Slide6
Decoy Routing
Decoy Routing -A mechanism capable of circumventing common network filtering strategies.
A client connects to any unblocked host service and then decoy routing is used to connect to blocked destination.
Circumvention service is placed in the network.
A single device could proxy traffic instead of host.Slide7
Routing Around Decoys
Schuchard et al. proposed Routing Against Decoys attack against decoy routing.
Main Idea- ISPs in censorship region have multiple paths to reach destination
It can instruct ISPs under it’s influence to select paths that do not pass through ISPs known to contain Decoys.Slide8
Objective of this paper
Authors have worked on true costs incurred by following RAD attack.
Various parameters have been studied such as Loss of Connectivity, Latency, path length etc.Slide9
Internet Topology
Business Relationship between ASs can be mapped to following three types according to Gao model
Customer-to-Provider (c2p)
Peer-to-Peer (p2p)
Sibling-to-sibling (s2s)Slide10
Internet Topology GraphSlide11
Internet Topology
Customer Cone : AS and its customers
Edge AS : AS with customer cone size =1
Transit AS : AS whose customer size is greater than 1 and transits other As traffic
Path : A sequence of neighbor ASes that connect source AS to destination AS.Slide12
Valid and Invalid Paths
Valid or Valley-Free(VF) Path
Every transit AS in the path a customer who is its immediate neighbor
Invalid or Non-Valley-Free (NVF) PathSlide13
BGP RoutingSlide14
RBGP RoutingSlide15
Costs of Routing
Degraded Internet Reachability
Less-Preferred Path
Longer Paths
Higher path latencies
Non-Valley-Free routes
New Transit ASes
Massive change in Transit LoadSlide16
Placing decoy Routers
RAD paper simulated two specific placements of decoys
Top - Tier
Random
But this placement in RAD is biased as decoys were primarily placed in EDGE ASsSlide17
Placing decoy Routers
Authors used following Strategic decoy Placements:
Sorted Placement - Decoys are chosen from ASs that transit more traffic for the RAD adversary.
sorted-with-ring - Set of ASs not directly controlled by RAD adversary
sorted-no-ring - Additionally exclude ASs having business relationship
Strategic random placement - ASs are chosen from a set of ASes with a particular customer size.
random-c (Random -1 is similar to one used in RAD).
random-with-ring-C and random-no-ring-CSlide18
Simulation Setup and Data Sources
Used CBGP - a popular BGP simulator with python interface to interact and query between ASs.
Geo location: “GeoLite Country” dataset to map IP addresses to countries.
AS relations : CAIDA’s inferred AS relationship dataset
AS ranking: CAIDA’s AS rank dataset
Latency: iPlane’s “Inter-PoP links” dataset to estimate BGP and RBGP path latencies.
Network origin: iPlane’s “Origin AS mapping” dataset Slide19
Simulation Results
Comparing the Internet connectivity of state-level censors.
Loss of connectivity for different RAD adversaries assuming the sorted-no-ring decoy placement strategy.Slide20
Simulation Results
Simulation results for two different scenarios :
China-World : Decoy chosen from 44000 ASs exlcuding the 199 ASs located in China. China is the adversary.
China-US :China is the RAD adversary; decoy ASes are selected only from the 13,299 ASes lo- cated in the United States.Slide21
Percentage of unreachable ASsSlide22
Non-Valley-Free pathsSlide23
Costly Valley-Free Paths
Using less preferred paths : Results have shown that the percentage of VF paths became from 6% to 21% more expensive for different placement strategies.
Longer Paths : Average increase in path length varies from 1.12 to 1.40.
Higher Latencies : Even same length paths have higher latencies due to less popular transits.Slide24
Latency Calculation
For two neighbor ASes A and B, eLat is calculated as :
where Ai represents the ith point-of-presence (PoP) of the AS A and nA is the number of A’s PoPs
For a BGP/RBGP path composed of k ASes {T1 , ..., Tk }, we define eLat to be the sum of eLat for all neighbor ASes in the path:Slide25
Simulation Results
The average increase in estimated latency due to the RAD attack.Slide26
need infrastructural changes
Edge ASes acting as transit ASes
Increased load on existing transit ASesSlide27
Traffic Volume
To simulate changes in transit loads, it is assumed that traffic volume between two ASes AS1 and AS2 is proportional to the number of IP addresses they respectively possess:
where I P s(A) is the number of IP addresses owned by the AS A
Maximum transit load increase factor for Chinese transit ASes due to the RAD attack
Text
Maximum transit load increase factor for Chinese transit ASes due to the RAD attackSlide28
Conclusions
Proposed RAD attack is extremely costly with loss of connectivity to many internet connections and lower QoS.
Strategic placement of decoy routers significantly increases cost.
Depends on connectivity of country.
Regional deployment is effective in defeating the RAD attack.
Needs more fine grained and data driven approach.Slide29
Questions?