60
1 Chapter 30 – Internet Security and Firewall Design 30.13 Firewalls and Internet Access “Intrane t”

Chapter 30 – Internet Security and Firewall Design 30.13 Firewalls and Internet Access

  • Upload
    nora

  • View
    45

  • Download
    0

Embed Size (px)

DESCRIPTION

Chapter 30 – Internet Security and Firewall Design 30.13 Firewalls and Internet Access. “Intranet”. 30.13 Firewalls and Internet Access - continued Successful access control and content protection requires a careful combination of: ► restrictions on network topology - PowerPoint PPT Presentation

Citation preview

Page 1: Chapter 30 –  Internet Security and Firewall Design 30.13  Firewalls and Internet Access

1

Chapter 30 – Internet Security and Firewall Design

30.13 Firewalls and Internet Access

“Intranet”

Page 2: Chapter 30 –  Internet Security and Firewall Design 30.13  Firewalls and Internet Access

2

30.13 Firewalls and Internet Access - continued

Successful access control and content protection requires a careful combination of:

► restrictions on network topology

► intermediate information staging

► packet filters

30.14 Multiple Connections and Weakest Links

Refers to first item above.

In general, an organization’s intranet has multiple connections to the Internet.

Must form a security perimeter by installing a firewall at each connection.

All firewalls must be configured to have same access restrictionselse entry through “weakest link.”

Page 3: Chapter 30 –  Internet Security and Firewall Design 30.13  Firewalls and Internet Access

3

Chapter 30 – Internet Security and Firewall Design

30.13 Firewalls and Internet Access

“Intranet”

Page 4: Chapter 30 –  Internet Security and Firewall Design 30.13  Firewalls and Internet Access

4

30.15 Firewall Implementation and Packet Filters

Refers to 3rd item.

We have previously seen the addition of additional capability to a router – NAT.

Now we add another capability – packet filter.

Recall: ► restrictions on network topology► intermediate information staging► packet filters

Usually, a packet filter allows a manager to identify classes of datagrams by specifying arbitrary combinations of:

► source IP address► destination IP address► protocol► source port► destination port► arrival interface

Page 5: Chapter 30 –  Internet Security and Firewall Design 30.13  Firewalls and Internet Access

5

A packet filter is stateless; it treats each datagram in isolation, not “remembering” datagrams that arrived earlier and keeping

no record of this event, apart from possibly writing to a log.

30.15 Firewall Implementation and Packet Filters - continued

We hope that the packet filter will operate at wire speed, not delaying incoming IP datagram traffic.

Page 6: Chapter 30 –  Internet Security and Firewall Design 30.13  Firewalls and Internet Access

6

Figure 7.2

Recall row-by-row table search in routing:

Page 7: Chapter 30 –  Internet Security and Firewall Design 30.13  Firewalls and Internet Access

7

30.15 Firewall Implementation and Packet Filters - continued

When an IP datagram arrives, the packet filter will work through this table, row by row. If the datagram matches the specification on any row, the datagram will be filtered/blocked/discarded.

128.5.0.0

The ports are not in the IP datagram header, so modified router must “drill down” into data.

Page 8: Chapter 30 –  Internet Security and Firewall Design 30.13  Firewalls and Internet Access

8

Like NAPT, packet filtering gets router involved in layer 4!

(looking inside “data” in IP datagram, not just header)

Transport

Page 9: Chapter 30 –  Internet Security and Firewall Design 30.13  Firewalls and Internet Access

9

30.16 Security and Packet Filter Specification

This packet filter has specified a small list of services to be blocked.

This does not work well, because:

► the number of well-known (i.e. server) ports is large and growing► some Internet traffic does not travel to or from the well-known

ports (e.g. organization can run WWW server on port 8080, instead of 80)

► listing ports of well-known services leaves the firewall vulnerable to tunneling (needs inside accomplice).

This suggests reversing the idea of the filter:

Instead of specifying types of datagram that should be filtered, specify types that should be forwarded.

Everything else is filtered.

Page 10: Chapter 30 –  Internet Security and Firewall Design 30.13  Firewalls and Internet Access

10

30.17 Consequences of Restricted Access for Clients

Problem with this scheme:

It prevents a client inside the firewall from receiving a reply from a server outside the firewall.

Why? Because the client chooses a source port at random, in the range 1024 to 65,536. In the server’s reply the client’s source port becomes the destination port. The packet filter would have to be configured to forward all of these possibilities.

Page 11: Chapter 30 –  Internet Security and Firewall Design 30.13  Firewalls and Internet Access

11

30.18 Stateful Firewalls

Recall that basic packet filters are stateless. They treat each IP datagram separately and keep no record of datagrams received.

Stateful firewalls watch outgoing requests and adapt the filter rules to accommodate the replies.

Example:

Internal client sends TCP connection request to external WWW server.

Stateful firewall records this as the two endpoints of the requested connection:

( IPsource, Portsource, IPdest, 80 )

When the server returns a connection accept the firewall will recognize this as a response to the request, and forward it to the client.

This is additional to the packet filter, so actions can still be prohibited, as determined by the administrator.

Page 12: Chapter 30 –  Internet Security and Firewall Design 30.13  Firewalls and Internet Access

12

30.18 Stateful Firewalls – continued

In the previous example, what if no reply is received to the connection request after a reasonable time?

The record of the connection must be purged – “soft state”

How does the stateful firewall know when a TCP connection is terminated, so that the record can be deleted?

Firewall must watch for the two FIN segments (“connection monitoring”)

Page 13: Chapter 30 –  Internet Security and Firewall Design 30.13  Firewalls and Internet Access

13

Figure 12.15

Basically, the firewall must be following this state-transition diagram for each of the active connections!

Page 14: Chapter 30 –  Internet Security and Firewall Design 30.13  Firewalls and Internet Access

14

30.19 Content Protection and Proxies

Recall that successful access control requires a careful combination of:

► restrictions on network topology► intermediate information staging► packet filters

Proxies refer to the second item.We have been concentrating on access, but we may also

have to protect content.

This is almost impossible at the packet-filter level, since content can be divided among many datagrams, which can arrive in any order and may be fragmented.

This is going far beyond the original idea of a wire-speed firewall!

The firewall must mimic the ultimate destination host by assembling the entire message for inspection – application proxy.

Page 15: Chapter 30 –  Internet Security and Firewall Design 30.13  Firewalls and Internet Access

15

30.19 Content Protection and Proxies - continued

“Transparent” proxy – apart from delay, client/user is unaware that there is a proxy.

“Non-transparent” – client is configured to access proxy when it tries to access the external server.

PROXY

Page 16: Chapter 30 –  Internet Security and Firewall Design 30.13  Firewalls and Internet Access

16

30.20 Monitoring and Logging

If you’re the network administrator, do it!

Or else you don’t know what’s happening.

Page 17: Chapter 30 –  Internet Security and Firewall Design 30.13  Firewalls and Internet Access

17

7.11 Establishing Routing Tables

For now, assume routing tables are loaded manually;

In chapters 13 and 15 we’ll see protocols that allow routers to learn routes from each other.

End of Chapter 7.

Background to Chapter 13 - 15

Page 18: Chapter 30 –  Internet Security and Firewall Design 30.13  Firewalls and Internet Access

18

BHM

ATL

*

Page 19: Chapter 30 –  Internet Security and Firewall Design 30.13  Firewalls and Internet Access

19

8.11 Route Change Requests from Routers – continued

This is not a general mechanism for route changes. It is restricted to routers sending to directly-connected hosts.

Figure 8.7 – R5 cannot redirect R1 to use the shorter path from S to D

But R1 could tell S to use R6 for traffic to D, provided that R6 is in R1’s routing table as “next hop” for destination D

Page 20: Chapter 30 –  Internet Security and Firewall Design 30.13  Firewalls and Internet Access

20

13.6 Automatic Route Propagation

“Routing protocols serve two important functions. First, they compute a set of shortest paths. Second, they respond to network failures or topology changes by continually updating the routing information.”

A network administrator cannot respond manually to failures fast enough.

Figure 13.3

13.7 Distance Vector (Bellman-Ford) Routing

This is the first type of automatic routing protocol that we shall study.

At start-up routing tables include only the directly-connected networks.

Page 21: Chapter 30 –  Internet Security and Firewall Design 30.13  Firewalls and Internet Access

21

“Distance” for direct connection has been changed from 0 to 1to agree with chapter 15.

Figure 13.3

Routers advertise their capabilities to their directly-connected neighbors, using IP local broadcast capability.

Page 22: Chapter 30 –  Internet Security and Firewall Design 30.13  Firewalls and Internet Access

22

13.7 Distance Vector (Bellman-Ford) Routing - continued

Periodically, routers broadcast copies of their routing tables to all directly-connected routers.

Consider router J sending to router K.

We think of J as advertising “I can get you to network X at a cost of Y”

“cost” means the number of routers along the path to X (router J plus subsequent routers).

Router K will update its routing table on the basis of the information received from J.

Page 23: Chapter 30 –  Internet Security and Firewall Design 30.13  Firewalls and Internet Access

23

Router K’s initial routing table

To see how it works, assume that at some later time router K has learned routes and its routing table looks like this:

Routers J, L, M, and Q are directly-reachable from K

Page 24: Chapter 30 –  Internet Security and Firewall Design 30.13  Firewalls and Internet Access

24

Router K now receives an update message from directly-connected router J

Recall that J says “I can get you to network X at a cost of Y”

Router K’s routing table Update message from J

Update items marked with arrow cause K to change its routing table.

Page 25: Chapter 30 –  Internet Security and Firewall Design 30.13  Firewalls and Internet Access

25

Router K’s routing table Update message from J

Resulting Changes to K’s routing table:

► to Net 4 – distance 4 – via J (a better route has been discovered)

K will now advertise “I can get you to Net 4 at a cost of 4 via J”“I can get you to Net 21 at a cost of 5 via J”“I can get you to Net 42 at a cost of 4 via J”

► to Net 21 – distance 5 – via J (a new route has been discovered)

► to Net 42 – distance 4 – via J (something has gone wrong with the old route beyond J )

Page 26: Chapter 30 –  Internet Security and Firewall Design 30.13  Firewalls and Internet Access

26

What routers are where? (detective work!)

Page 27: Chapter 30 –  Internet Security and Firewall Design 30.13  Firewalls and Internet Access

27

13.7 Distance Vector (Bellman-Ford) Routing – continued

Advantages:

► Distance-vector algorithms are easy to implement.

Disadvantages:

► All routers must participate

► In a relatively static environment they compute the shortest paths and propagate correct routes to all destinations.

► When routes change rapidly the computations may not stabilize (changes propagate slowly – diffusion)

► In a large internet the update messages get large (size is proportional to the number of networks in the

internet, so distance-vector algorithms “do not scale well”)

Page 28: Chapter 30 –  Internet Security and Firewall Design 30.13  Firewalls and Internet Access

28

13.9 Link-State SPF) Routing

An alternative to distance-vector routing is link-state routing.

These are known as Shortest Path First

(a misnomer, since all routing algorithms compute the shortest path)

Every router has a graph (CS 250/350) showing all other routers and the networks to which they connect. Nodes in the graph are the routers; links in the graph are direct connections between routers.

Periodically each router tests the reachability of all directly-connected routers (i.e. tests whether each of its links is “up” or “down”)

The router multicasts this information to all other routers.

If a receiving router detects a change in link status, the router recomputes shortest paths to all possible destinations, using Dijkstra’s algorithm.

Page 29: Chapter 30 –  Internet Security and Firewall Design 30.13  Firewalls and Internet Access

29

Link-State Routing.

Advantages:

► size of the update messages sent by a router is proportional to the number of links it has

(i.e. update messages are much smaller than those in vector-distance, so link-state “scales better”)

Disadvantages:

► computational load on routers.

► each router computes routes independently from original data (not relying on intermediate routers)

Page 30: Chapter 30 –  Internet Security and Firewall Design 30.13  Firewalls and Internet Access

30

14.5 Autonomous System Concept

We cannot run an automatic routing protocol for the entire Global Internet.

How should the Internet be partitioned into sets of routers so that each set can run a routing update protocol?

Networks and routers are owned by organizations and individuals.Within each, an administrative authority can guarantee that internal routes remain consistent and viable.

For purposes of routing, a group of networks controlled by a single administrative authority is called an autonomous system (AS) identified by an autonomous system number.

Comer suggests thinking about an autonomous system as corresponding to a large ISP (but UAB is an AS, number 3452)

Page 31: Chapter 30 –  Internet Security and Firewall Design 30.13  Firewalls and Internet Access

31

One router can be chosen to inform the outside world of networks within the organization (assume desire for universal connectivity - temporarily ignore security!)

This router also learns about outside networks and distributes this information internally.

Page 32: Chapter 30 –  Internet Security and Firewall Design 30.13  Firewalls and Internet Access

32

Figure 14.2

Within an autonomous system, the administration chooses a routing method.

Between autonomous systems, the Border Gateway Protocol (BGP-4) is used.

14.6 Exterior Gateway Protocols and Reachability

R1 gathers information about networks in AS1 and BGPs the info to R2

R2 gathers information about networks in AS2 and BGPs the info to R1.

Page 33: Chapter 30 –  Internet Security and Firewall Design 30.13  Firewalls and Internet Access

33

Chapter 15: Routing Within an Autonomous System (RIP, OFPF)

15.3 Routing Information Protocol

RIP is a straightforward implementation of distance-vector routing.

Routers run RIP in “active mode,” broadcast update messages to directly-connected neighbors every 30 seconds.

Hosts listen and learn, but do not broadcast.

Page 34: Chapter 30 –  Internet Security and Firewall Design 30.13  Firewalls and Internet Access

34

15.3 Routing Information Protocol – continued

RIP rules:

► routers send updates every 30 seconds

► receiving routers do not replace an existing routewith one of equal cost (hop count)

► the maximum hop count is 16 (“infinity”)

► receivers use 180-second timeout on entries (“soft state”)

We will use fig 15.2 to illustrate how RIP works.

Page 35: Chapter 30 –  Internet Security and Firewall Design 30.13  Firewalls and Internet Access

35

Initially: R5 not running

Other routers have only direct connections.

N3 1 dir

N4 1 dir

N2 1 dirN3 1 dir

N2 1 dir

N3 1 dir

N1 1 dir

N2 1 dir

N1 3 R2N2 2 R2

N1 2 R1N1 2 R1

N1 3 R5 N2 2 R5

Page 36: Chapter 30 –  Internet Security and Firewall Design 30.13  Firewalls and Internet Access

36

15.4 Slow Convergence Problem

Fig 15.4 (a)

Page 37: Chapter 30 –  Internet Security and Firewall Design 30.13  Firewalls and Internet Access

37

R1

R2

Send to R2

Send to R1

and R3

N1 1

N1 2 R1

N1 2

N1 16

N1 3 R2 N1 3

N1 4 R1

N1 4

N1 5 R2 N1 5

N1 6 R1N1 6

At this point we have a routing loop!

Fails!

Page 38: Chapter 30 –  Internet Security and Firewall Design 30.13  Firewalls and Internet Access

38

15.4 Slow Convergence Problem

Fig 15.4

Page 39: Chapter 30 –  Internet Security and Firewall Design 30.13  Firewalls and Internet Access

39

15.5 Solving the Slow Convergence Problem

Problem arises from sending back a route to the router that sent it.

“Split horizon updates” prevent this.

Router K must not send routes to Net 24 and Net 42 back to router J

Easy to implement: recall figure 13.4:Router K’s routing table

This is done in RIPv2

Page 40: Chapter 30 –  Internet Security and Firewall Design 30.13  Firewalls and Internet Access

40

15.5 Solving the Slow Convergence Problem – continued

Other techniques:

after receipt of information that a network is unreachable:

► “hold down”

ignore further information about that network

for hold-down period (60 seconds)

► “poison reverse” with “triggered updates” continue to advertise path to that network, with cost 16 send immediate special update –

don’t wait for the regular 30-second schedule.

Page 41: Chapter 30 –  Internet Security and Firewall Design 30.13  Firewalls and Internet Access

41

15.9 RIP2 Extensions and Message Format

Figure 15.6

COMMAND: 1 = request, 2 = response

Route to Network 1 Goes next to this D-C router

And this is the total distance to the destination over this route.

Page 42: Chapter 30 –  Internet Security and Firewall Design 30.13  Firewalls and Internet Access

42

15.9 RIP2 Extensions and Message Format – continued

In RIPv1 routers broadcast their messages, so that every computer in the local network had to process the message. This is wasteful.

RIPv2 makes use of multicasting to the class–D “RIP2 routers” address 224.0.0.9.

This sends messages specifically (only) to routers on the local network.

Page 43: Chapter 30 –  Internet Security and Firewall Design 30.13  Firewalls and Internet Access

43

15.9 RIP2 Extensions and Message Format – continued

RIP messages travel encapsulated in UDP datagrams

Both source and destination ports are 520 (not client/server).

15.10 The Disadvantage of RIP Hop Counts

Using hop counts as a metric does not always yield routes with the least delay or the highest capacity.

Page 44: Chapter 30 –  Internet Security and Firewall Design 30.13  Firewalls and Internet Access

44

15.11 Delay Metric

HELLO protocol measures delay of competing routes and selects route with least delay.

15.12 Delay Metrics and Oscillation

HELLO is susceptible to oscillation between two routes with similar delay.

Page 45: Chapter 30 –  Internet Security and Firewall Design 30.13  Firewalls and Internet Access

45

15.15 The Open SPF Protocol (OSPF)

An Implementation of link-state routing.

Features:

► open standard (not proprietary)

► type-of-service routing

► load balancing – “if a manager specifies multiple routes to a given destination at the same cost, OSPF distributes traffic

over all routes equally.”

► can partition internets into areas

► exchanges between routers can be authenticated

► supports host-specific, subnet-specific, classful and class-less

routes

Page 46: Chapter 30 –  Internet Security and Firewall Design 30.13  Firewalls and Internet Access

46

15.16 Routing with Partial Information

“Routers at the center of the Internet have a complete set of routes to all possible destinations; such routers do not use default routes.”

(288,000 entries in routing tables in 2009 +14% /year)

Most other routers do not have complete information they use default routes.

Page 47: Chapter 30 –  Internet Security and Firewall Design 30.13  Firewalls and Internet Access

47

15.16 Routing with Partial Information - continued

Using default routes for most routers has two consequences:

► local routing errors can go undetected – one router’s default may send datagrams to the wrong next-hop router (perhaps

outside the autonomous system), but that router may quietly forward the datagram to the correct next hop

(perhaps back inside the autonomous system);

► routing update messages exchanged by routers can be much smaller than if the messages contained all possible

destinations (our original motivation for using default routes).

Page 48: Chapter 30 –  Internet Security and Firewall Design 30.13  Firewalls and Internet Access

48

No N3 Default R1

N3 2 R3

Sub-optimal routing

Page 49: Chapter 30 –  Internet Security and Firewall Design 30.13  Firewalls and Internet Access

49

Page 50: Chapter 30 –  Internet Security and Firewall Design 30.13  Firewalls and Internet Access

50

Lab Session 5 – Packet Filtering

1. Physical Connections

INSIDE: as usual (192.168.1.0)

OUTSIDE: UAB class B address 138. 26. 0. 0

CIS subnet 138. 26. 66. 0

mask 255. 255. 255. 0

we will subnet further 255. 255. 255. 240

Packet filter

Page 51: Chapter 30 –  Internet Security and Firewall Design 30.13  Firewalls and Internet Access

51

OUTSIDE: UAB class B address 138. 26. 0. 0 CIS subnet 138. 26. 66. 0

mask 255. 255. 255. 0we will subnet further 255. 255. 255. 240

255. 255. 255. 240

11111111.11111111.11111111.11110000

Subnets are 138. 26. 66. 0

66. 16 ..

66. 240 use this

000000010010 …

1111

Available host IP addresses are: 138.26.66.241 thru 138.26.66.254

138 . 26 . 66 .

0001

Page 52: Chapter 30 –  Internet Security and Firewall Design 30.13  Firewalls and Internet Access

52

138.26.66.241 138.26.66.242

2. Configure the Interfaces

eth0 on CENTER: 138.26.66.254

Page 53: Chapter 30 –  Internet Security and Firewall Design 30.13  Firewalls and Internet Access

53

3. Install Routes and Check Operation

same as Lab 4, except take opportunity to use default routes

Page 54: Chapter 30 –  Internet Security and Firewall Design 30.13  Firewalls and Internet Access

54

Recall Packet Filter:

First idea: if match, discard datagram

Second idea: if match, accept datagram

Linux can implement packet filter with module called iptables –

can do either of these via “policy” -- DROP or ACCEPT

Page 55: Chapter 30 –  Internet Security and Firewall Design 30.13  Firewalls and Internet Access

55

There are 3 tables: INPUT FORWARD and OUTPUT

Default policy for all three is ACCEPT

This is the “null firewall” (Section 4)

eg. Routing changes

Page 56: Chapter 30 –  Internet Security and Firewall Design 30.13  Firewalls and Internet Access

56

In lab session 5 sections 5 thru 9 we implement various packet filter configurations by making changes only in the FORWARD table.

Page 57: Chapter 30 –  Internet Security and Firewall Design 30.13  Firewalls and Internet Access

57Echo request from LEFT arriving at RIGHT2 will appear to have come from CENTER’s IP address (Network Address Translation)

Page 58: Chapter 30 –  Internet Security and Firewall Design 30.13  Firewalls and Internet Access

58

11. Masquerading with FTP

This was designed to demonstrate NAPT

TCP connection request from LEFT arriving at RIGHT2 will appear to have come from CENTER’s IP address

In general the source port number will not need to be changed, but will be if two clients choose same random source port number.

You will see that FTP in “active” mode does not work, but “passive” mode works.

Page 59: Chapter 30 –  Internet Security and Firewall Design 30.13  Firewalls and Internet Access

59

Page 60: Chapter 30 –  Internet Security and Firewall Design 30.13  Firewalls and Internet Access

60

March 26, 2011: Microsoft pays Nortel $7.5 million for 666,624 IPv4 addresses, raising questions if the IPv4 black market has arrived.