Internet Resiliency to Attacks and Failures under BGP Policy Routing - - PowerPoint PPT Presentation

▶
internet resiliency to attacks and failures under bgp
SMART_READER_LITE
LIVE PREVIEW

Internet Resiliency to Attacks and Failures under BGP Policy Routing - - PowerPoint PPT Presentation

Jun 26, 2023 300 likes •926 views

Internet Resiliency to Attacks and Failures under BGP Policy Routing D. Dolev, S. Jamin, O. Morkyn, Y. Shavitt Presenter: Shiva Kasiviswanathan Pennsylvania State University University Park Internet Resiliency to Attacks and Failures under BGP

slide-1
SLIDE 1

Internet Resiliency to Attacks and Failures under BGP Policy Routing

  • D. Dolev, S. Jamin, O. Morkyn, Y. Shavitt

Presenter: Shiva Kasiviswanathan Pennsylvania State University University Park

Internet Resiliency to Attacks and Failures under BGP Policy Routing – p. 1/12

slide-2
SLIDE 2

Layout of the Presentation

Introduction to Resiliency

Internet Resiliency to Attacks and Failures under BGP Policy Routing – p. 2/12

slide-3
SLIDE 3

Layout of the Presentation

Introduction to Resiliency Drawbacks of Related Works

Internet Resiliency to Attacks and Failures under BGP Policy Routing – p. 2/12

slide-4
SLIDE 4

Layout of the Presentation

Introduction to Resiliency Drawbacks of Related Works BGP protocol

Internet Resiliency to Attacks and Failures under BGP Policy Routing – p. 2/12

slide-5
SLIDE 5

Layout of the Presentation

Introduction to Resiliency Drawbacks of Related Works BGP protocol AS Connectivity and Internet Topology

Internet Resiliency to Attacks and Failures under BGP Policy Routing – p. 2/12

slide-6
SLIDE 6

Layout of the Presentation

Introduction to Resiliency Drawbacks of Related Works BGP protocol AS Connectivity and Internet Topology Graph Theoretic Modeling

Internet Resiliency to Attacks and Failures under BGP Policy Routing – p. 2/12

slide-7
SLIDE 7

Resiliency in Internet

Internet collection of Autonomous systems.

Internet Resiliency to Attacks and Failures under BGP Policy Routing – p. 3/12

slide-8
SLIDE 8

Resiliency in Internet

Internet collection of Autonomous systems. Path 1: Stability of Routing protocols.

Internet Resiliency to Attacks and Failures under BGP Policy Routing – p. 3/12

slide-9
SLIDE 9

Resiliency in Internet

Internet collection of Autonomous systems. Path 1: Stability of Routing protocols. Path 2: Random Failures of nodes or Attack on key elements.

Internet Resiliency to Attacks and Failures under BGP Policy Routing – p. 3/12

slide-10
SLIDE 10

Resiliency in Internet

Internet collection of Autonomous systems. Path 1: Stability of Routing protocols. Path 2: Random Failures of nodes or Attack on key elements. Internet topology as Scale Free, i.e., follows Power law.

Internet Resiliency to Attacks and Failures under BGP Policy Routing – p. 3/12

slide-11
SLIDE 11

Resiliency in Internet

Internet collection of Autonomous systems. Path 1: Stability of Routing protocols. Path 2: Random Failures of nodes or Attack on key elements. Internet topology as Scale Free, i.e., follows Power law. Phase Transition Phenomena: Random Deletion doest’t disconnect.

Internet Resiliency to Attacks and Failures under BGP Policy Routing – p. 3/12

slide-12
SLIDE 12

Resiliency in Internet

Internet collection of Autonomous systems. Path 1: Stability of Routing protocols. Path 2: Random Failures of nodes or Attack on key elements. Internet topology as Scale Free, i.e., follows Power law. Phase Transition Phenomena: Random Deletion doest’t disconnect. Internet Susceptible to attacks on high degree nodes.

Internet Resiliency to Attacks and Failures under BGP Policy Routing – p. 3/12

slide-13
SLIDE 13

Resiliency in Internet

Internet collection of Autonomous systems. Path 1: Stability of Routing protocols. Path 2: Random Failures of nodes or Attack on key elements. Internet topology as Scale Free, i.e., follows Power law. Phase Transition Phenomena: Random Deletion doest’t disconnect. Internet Susceptible to attacks on high degree nodes.

Internet Resiliency to Attacks and Failures under BGP Policy Routing – p. 3/12

slide-14
SLIDE 14

Drawbacks of Related Works

Treat Internet as undirected graph.

Internet Resiliency to Attacks and Failures under BGP Policy Routing – p. 4/12

slide-15
SLIDE 15

Drawbacks of Related Works

Treat Internet as undirected graph. Routing in Internet is governed by policies with aid of BGP.

Internet Resiliency to Attacks and Failures under BGP Policy Routing – p. 4/12

slide-16
SLIDE 16

Drawbacks of Related Works

Treat Internet as undirected graph. Routing in Internet is governed by policies with aid of BGP. Physical Path alone not enough for information exchange.

Internet Resiliency to Attacks and Failures under BGP Policy Routing – p. 4/12

slide-17
SLIDE 17

Drawbacks of Related Works

Treat Internet as undirected graph. Routing in Internet is governed by policies with aid of BGP. Physical Path alone not enough for information exchange. A valid path conforming to the policies of AS should exist.

Internet Resiliency to Attacks and Failures under BGP Policy Routing – p. 4/12

slide-18
SLIDE 18

Drawbacks of Related Works

Treat Internet as undirected graph. Routing in Internet is governed by policies with aid of BGP. Physical Path alone not enough for information exchange. A valid path conforming to the policies of AS should exist. Data used reflect partial view of Internet.

Internet Resiliency to Attacks and Failures under BGP Policy Routing – p. 4/12

slide-19
SLIDE 19

Drawbacks of Related Works

Treat Internet as undirected graph. Routing in Internet is governed by policies with aid of BGP. Physical Path alone not enough for information exchange. A valid path conforming to the policies of AS should exist. Data used reflect partial view of Internet. Data collected from few nodes and contain neighbors on shortest BGP path.

Internet Resiliency to Attacks and Failures under BGP Policy Routing – p. 4/12

slide-20
SLIDE 20

Drawbacks of Related Works

Treat Internet as undirected graph. Routing in Internet is governed by policies with aid of BGP. Physical Path alone not enough for information exchange. A valid path conforming to the policies of AS should exist. Data used reflect partial view of Internet. Data collected from few nodes and contain neighbors on shortest BGP path. Also BGP doesn’t advertise backup path.

Internet Resiliency to Attacks and Failures under BGP Policy Routing – p. 4/12

slide-21
SLIDE 21

AS connectivity and Internet Topology

BGP protocol enables each administrative domain to decide which path to use.

Internet Resiliency to Attacks and Failures under BGP Policy Routing – p. 5/12

slide-22
SLIDE 22

AS connectivity and Internet Topology

BGP protocol enables each administrative domain to decide which path to use. Paths in Internet not shortest paths, but shortest conforming to AS policies.

Internet Resiliency to Attacks and Failures under BGP Policy Routing – p. 5/12

slide-23
SLIDE 23

AS connectivity and Internet Topology

BGP protocol enables each administrative domain to decide which path to use. Paths in Internet not shortest paths, but shortest conforming to AS policies. Commercial agreements between AS create relationships.

Internet Resiliency to Attacks and Failures under BGP Policy Routing – p. 5/12

slide-24
SLIDE 24

AS connectivity and Internet Topology

BGP protocol enables each administrative domain to decide which path to use. Paths in Internet not shortest paths, but shortest conforming to AS policies. Commercial agreements between AS create relationships. Customer-Provider, Peer to Peer, Siblings.

Internet Resiliency to Attacks and Failures under BGP Policy Routing – p. 5/12

slide-25
SLIDE 25

AS connectivity and Internet Topology

BGP protocol enables each administrative domain to decide which path to use. Paths in Internet not shortest paths, but shortest conforming to AS policies. Commercial agreements between AS create relationships. Customer-Provider, Peer to Peer, Siblings. Peer to Peer and Siblings differ in amount of transit information.

Internet Resiliency to Attacks and Failures under BGP Policy Routing – p. 5/12

slide-26
SLIDE 26

AS connectivity and Internet Topology

BGP protocol enables each administrative domain to decide which path to use. Paths in Internet not shortest paths, but shortest conforming to AS policies. Commercial agreements between AS create relationships. Customer-Provider, Peer to Peer, Siblings. Peer to Peer and Siblings differ in amount of transit information. Paper by Lixin Gao defines algorithm for inferring relation- ships from BGP paths.

Internet Resiliency to Attacks and Failures under BGP Policy Routing – p. 5/12

slide-27
SLIDE 27

Connectivity and Internet Topology Contd..

Define uphill path as customer-provider links.

Internet Resiliency to Attacks and Failures under BGP Policy Routing – p. 6/12

slide-28
SLIDE 28

Connectivity and Internet Topology Contd..

Define uphill path as customer-provider links. Define downhill path as provider-customer links.

Internet Resiliency to Attacks and Failures under BGP Policy Routing – p. 6/12

slide-29
SLIDE 29

Connectivity and Internet Topology Contd..

Define uphill path as customer-provider links. Define downhill path as provider-customer links. One legal AS path is uphill followed by downhill.

Internet Resiliency to Attacks and Failures under BGP Policy Routing – p. 6/12

slide-30
SLIDE 30

Connectivity and Internet Topology Contd..

Define uphill path as customer-provider links. Define downhill path as provider-customer links. One legal AS path is uphill followed by downhill. Other is uphill followed by peer link followed by downhill.

Internet Resiliency to Attacks and Failures under BGP Policy Routing – p. 6/12

slide-31
SLIDE 31

Connectivity and Internet Topology Contd..

Define uphill path as customer-provider links. Define downhill path as provider-customer links. One legal AS path is uphill followed by downhill. Other is uphill followed by peer link followed by downhill. AS relationship studies also reveal that Internet is hierar- chical.

Internet Resiliency to Attacks and Failures under BGP Policy Routing – p. 6/12

slide-32
SLIDE 32

AS Graph Model

AS is modeled as directed graph with nodes as AS systems and edges if they have peering relationship and BGP neighbors.

Internet Resiliency to Attacks and Failures under BGP Policy Routing – p. 7/12

slide-33
SLIDE 33

AS Graph Model

AS is modeled as directed graph with nodes as AS systems and edges if they have peering relationship and BGP neighbors. Uphill edges, downhill edges, peer to peer directed edges in both ways, siblings undirected edge.

Internet Resiliency to Attacks and Failures under BGP Policy Routing – p. 7/12

slide-34
SLIDE 34

AS Graph Model

AS is modeled as directed graph with nodes as AS systems and edges if they have peering relationship and BGP neighbors. Uphill edges, downhill edges, peer to peer directed edges in both ways, siblings undirected edge. The reachability algorithm maintains reachability map.

Internet Resiliency to Attacks and Failures under BGP Policy Routing – p. 7/12

slide-35
SLIDE 35

AS Graph Model

AS is modeled as directed graph with nodes as AS systems and edges if they have peering relationship and BGP neighbors. Uphill edges, downhill edges, peer to peer directed edges in both ways, siblings undirected edge. The reachability algorithm maintains reachability map. For each node find the nodes that can be reached in the AS constraint graph.

Internet Resiliency to Attacks and Failures under BGP Policy Routing – p. 7/12

slide-36
SLIDE 36

AS Graph Model

AS is modeled as directed graph with nodes as AS systems and edges if they have peering relationship and BGP neighbors. Uphill edges, downhill edges, peer to peer directed edges in both ways, siblings undirected edge. The reachability algorithm maintains reachability map. For each node find the nodes that can be reached in the AS constraint graph. Simple Adaption of BFS search algorithm.

Internet Resiliency to Attacks and Failures under BGP Policy Routing – p. 7/12

slide-37
SLIDE 37

AS Reachability Algorithm

Each node is labeled by link through with it is reached.

Internet Resiliency to Attacks and Failures under BGP Policy Routing – p. 8/12

slide-38
SLIDE 38

AS Reachability Algorithm

Each node is labeled by link through with it is reached. A node can be reached only if reaching it gives new possibilities.

Internet Resiliency to Attacks and Failures under BGP Policy Routing – p. 8/12

slide-39
SLIDE 39

AS Reachability Algorithm

Each node is labeled by link through with it is reached. A node can be reached only if reaching it gives new possibilities. States of nodes are: None, Up, Down, Slide.

Internet Resiliency to Attacks and Failures under BGP Policy Routing – p. 8/12

slide-40
SLIDE 40

AS Reachability Algorithm

Each node is labeled by link through with it is reached. A node can be reached only if reaching it gives new possibilities. States of nodes are: None, Up, Down, Slide. None: Node not traversed yet.

Internet Resiliency to Attacks and Failures under BGP Policy Routing – p. 8/12

slide-41
SLIDE 41

AS Reachability Algorithm

Each node is labeled by link through with it is reached. A node can be reached only if reaching it gives new possibilities. States of nodes are: None, Up, Down, Slide. None: Node not traversed yet. Up: Node was in none, slide and down and there is a uphill path from it.

Internet Resiliency to Attacks and Failures under BGP Policy Routing – p. 8/12

slide-42
SLIDE 42

AS Reachability Algorithm

Each node is labeled by link through with it is reached. A node can be reached only if reaching it gives new possibilities. States of nodes are: None, Up, Down, Slide. None: Node not traversed yet. Up: Node was in none, slide and down and there is a uphill path from it.

Internet Resiliency to Attacks and Failures under BGP Policy Routing – p. 8/12

slide-43
SLIDE 43

AS Reachability Algorithm Contd...

Down: node was in none state, there is a downhill path.

Internet Resiliency to Attacks and Failures under BGP Policy Routing – p. 9/12

slide-44
SLIDE 44

AS Reachability Algorithm Contd...

Down: node was in none state, there is a downhill path. Slide: node was none or down and there exists a peer path.

Internet Resiliency to Attacks and Failures under BGP Policy Routing – p. 9/12

slide-45
SLIDE 45

AS Reachability Algorithm Contd...

Down: node was in none state, there is a downhill path. Slide: node was none or down and there exists a peer path. Algorithm gives the priority in order: uphill path, peer to peer path, downhill path.

Internet Resiliency to Attacks and Failures under BGP Policy Routing – p. 9/12

slide-46
SLIDE 46

AS Reachability Algorithm Contd...

Down: node was in none state, there is a downhill path. Slide: node was none or down and there exists a peer path. Algorithm gives the priority in order: uphill path, peer to peer path, downhill path. Each node is visited thrice, so complexity O(|E|).

Internet Resiliency to Attacks and Failures under BGP Policy Routing – p. 9/12

slide-47
SLIDE 47

Metrics for Defining Resiliency

Some metrics are Average Diameter, Average Shortest path length ˜

d, giant component size S, number of

connected pairs K, Diameter-Inverse-k-DIK.

Internet Resiliency to Attacks and Failures under BGP Policy Routing – p. 10/12

slide-48
SLIDE 48

Metrics for Defining Resiliency

Some metrics are Average Diameter, Average Shortest path length ˜

d, giant component size S, number of

connected pairs K, Diameter-Inverse-k-DIK. DIK measures both the expected distance between two nodes and the probability of a path existing between two arbitrary nodes.

DIK = ˜ d k

Internet Resiliency to Attacks and Failures under BGP Policy Routing – p. 10/12

slide-49
SLIDE 49

Metrics for Defining Resiliency

Some metrics are Average Diameter, Average Shortest path length ˜

d, giant component size S, number of

connected pairs K, Diameter-Inverse-k-DIK. DIK measures both the expected distance between two nodes and the probability of a path existing between two arbitrary nodes.

DIK = ˜ d k

These measures can’t be applied in this case as reachability not same as connectivity.

Internet Resiliency to Attacks and Failures under BGP Policy Routing – p. 10/12

slide-50
SLIDE 50

Metrics for Defining Resiliency

Some metrics are Average Diameter, Average Shortest path length ˜

d, giant component size S, number of

connected pairs K, Diameter-Inverse-k-DIK. DIK measures both the expected distance between two nodes and the probability of a path existing between two arbitrary nodes.

DIK = ˜ d k

These measures can’t be applied in this case as reachability not same as connectivity. First measure R captures the reachability of Internet. Let

r(u, v) ∈ 0, 1 denote the reachability between an arbitrary

distinct pair of nodes u, v ∈ AS.

Internet Resiliency to Attacks and Failures under BGP Policy Routing – p. 10/12

slide-51
SLIDE 51

Metrics for Defining Resiliency

Let πr denote the number of distinct node pairs in graph with r = 1. Let Or denote the theoretical limit for πr.

R = πr Or

Internet Resiliency to Attacks and Failures under BGP Policy Routing – p. 11/12

slide-52
SLIDE 52

Metrics for Defining Resiliency

Let πr denote the number of distinct node pairs in graph with r = 1. Let Or denote the theoretical limit for πr.

R = πr Or

Second metric quantifies the size of strongly connected component graph. Create a reachability graph with an edge between u and v iff r(u, v) = 1.

Internet Resiliency to Attacks and Failures under BGP Policy Routing – p. 11/12

slide-53
SLIDE 53

Metrics for Defining Resiliency

Let πr denote the number of distinct node pairs in graph with r = 1. Let Or denote the theoretical limit for πr.

R = πr Or

Second metric quantifies the size of strongly connected component graph. Create a reachability graph with an edge between u and v iff r(u, v) = 1. Largest strongly connected in original graph corresponds to clique in this graph.

Internet Resiliency to Attacks and Failures under BGP Policy Routing – p. 11/12

slide-54
SLIDE 54

Metrics for Defining Resiliency

Let πr denote the number of distinct node pairs in graph with r = 1. Let Or denote the theoretical limit for πr.

R = πr Or

Second metric quantifies the size of strongly connected component graph. Create a reachability graph with an edge between u and v iff r(u, v) = 1. Largest strongly connected in original graph corresponds to clique in this graph. Paper used a greedy heuristic for finding maximal clique in the graph.

Internet Resiliency to Attacks and Failures under BGP Policy Routing – p. 11/12

slide-55
SLIDE 55

Metrics for Defining Resiliency

Let πr denote the number of distinct node pairs in graph with r = 1. Let Or denote the theoretical limit for πr.

R = πr Or

Second metric quantifies the size of strongly connected component graph. Create a reachability graph with an edge between u and v iff r(u, v) = 1. Largest strongly connected in original graph corresponds to clique in this graph. Paper used a greedy heuristic for finding maximal clique in the graph. Find a core node and find maximal clique from it.

Internet Resiliency to Attacks and Failures under BGP Policy Routing – p. 11/12

slide-56
SLIDE 56

Resiliency of Internet: Results

Four Datasets: LZ, OR1, OR2, UM.

Internet Resiliency to Attacks and Failures under BGP Policy Routing – p. 12/12

slide-57
SLIDE 57

Resiliency of Internet: Results

Four Datasets: LZ, OR1, OR2, UM. LZ dataset from RIPE Routing Information Service is the result of combining two exchange points one in London,

  • ther in Zurich.

Internet Resiliency to Attacks and Failures under BGP Policy Routing – p. 12/12

slide-58
SLIDE 58

Resiliency of Internet: Results

Four Datasets: LZ, OR1, OR2, UM. LZ dataset from RIPE Routing Information Service is the result of combining two exchange points one in London,

  • ther in Zurich.

To test Resiliency target attack at Internets most highly connected AS.

Internet Resiliency to Attacks and Failures under BGP Policy Routing – p. 12/12

slide-59
SLIDE 59

Resiliency of Internet: Results

Four Datasets: LZ, OR1, OR2, UM. LZ dataset from RIPE Routing Information Service is the result of combining two exchange points one in London,

  • ther in Zurich.

To test Resiliency target attack at Internets most highly connected AS. Since reachability is less than connectivity, on deletion of few high degree nodes size of largest component falls faster.

Internet Resiliency to Attacks and Failures under BGP Policy Routing – p. 12/12

slide-60
SLIDE 60

Resiliency of Internet: Results

Four Datasets: LZ, OR1, OR2, UM. LZ dataset from RIPE Routing Information Service is the result of combining two exchange points one in London,

  • ther in Zurich.

To test Resiliency target attack at Internets most highly connected AS. Since reachability is less than connectivity, on deletion of few high degree nodes size of largest component falls faster. Conclusion is that Internet is susceptible to attack on high degree nodes.

Internet Resiliency to Attacks and Failures under BGP Policy Routing – p. 12/12

slide-61
SLIDE 61

Resiliency of Internet: Results

Four Datasets: LZ, OR1, OR2, UM. LZ dataset from RIPE Routing Information Service is the result of combining two exchange points one in London,

  • ther in Zurich.

To test Resiliency target attack at Internets most highly connected AS. Since reachability is less than connectivity, on deletion of few high degree nodes size of largest component falls faster. Conclusion is that Internet is susceptible to attack on high degree nodes. Resiliency to random failures in directed graph and undi- rected graph was found not to be significant.

Internet Resiliency to Attacks and Failures under BGP Policy Routing – p. 12/12