2. Outline
Overview Issues and Threats in Network
Security
Review basic network technology
TCP/IP in particular
Attacks specific to particular technologies
4. Increased Security Complexity
Different operating systems
Computers, Servers, Network Devices
Multiple Administrative Domains
Need to open access
Multiple Paths and shared resources
Anonymity
5. OSI Reference Model
• The layers
– 7: Application, e.g., HTTP, SMTP, FTP
– 6: Presentation
– 5: Session
– 4: Transport, e.g. TCP, UDP
– 3: Network, e.g. IP, IPX
– 2: Data link, e.g., Ethernet frames, ATM cells
– 1: Physical, e.g., Ethernet media, ATM media
• Standard software engineering reasons for thinking
about a layered design
6. Message mapping to the layers
SVN update message
Packet2
D
P
S
P
D
P
S
P
Packet1
D
P
S
P
D
P
S
P
D
A
S
A
Packet1
D
P
S
P
D
A
S
A
Pack
2
Communications bit stream
D
P
S
P
D
A
S
A
Packet1
D
M
S
M
D
P
S
P
D
A
S
A
Pack
2
D
M
S
M
L7 App
L4 TCP
L3 IP
L2 Eth
8. Switches
• Original ethernet broadcast all packets
• Layer two means of passing packets
– Learn or config which MAC's live behind which ports
– Only pass traffic to the appropriate port
• Span ports
– Mirror all traffic
9. Physical Denial of Service
Radio
Jamming
Cables
Cutting or mutilating
10. Network Layer - IP
Moves packets between computers
Possibly on different physical segments
Best effort
Technologies
Routing
Lower level address discovery (ARP)
Error Messages (ICMP)
11. IPv4
• See Wikipedia for field details
– http://en.wikipedia.org/wiki/IPv4
Version IHL Type of service Total length
Identification DF MF Frag Offset
Time to live Protocol Header checksum
Source address
Destination Address
0 or more words of options
12. Ipv4 Addressing
• Each entity has at least one address
• Addresses divided into subnetwork
– Address and mask combination
– 192.168.1.0/24 or 10.0.0.0/8
– 192.168.1.0 255.255.255.0 or 10.0.0.0 255.0.0.0
– 192.168.1.0-192.168.1.255 or 10.0.0.0-
10.255.255.255
• Addresses in your network are “directly”
connected
– Broadcasts should reach them
– No need to route packets to them
13. Address spoofing
• Sender can put any source address in packets
he sends:
– Can be used to send unwelcome return traffic to
the spoofed address
– Can be used to bypass filters to get unwelcome
traffic to the destination
• Reverse Path verification can be used by
routers to broadly catch some spoofers
14. Address Resolution Protocol (ARP)
• Used to discover mapping of neighboring
ethernet MAC to IP addresses.
– Need to find MAC for 192.168.1.3 which is in your
interface's subnetwork
– Broadcast an ARP request on the link
– Hopefully receive an ARP reply giving the correct
MAC
– The device stores this information in an ARP cache
or ARP table
15. ARP cache poisoning
• Bootstrap problem with respect to security. Anyone can send
an ARP reply
– The Ingredients to ARP Poison,
http://www.governmentsecurity.org/articles/TheIngredientstoARPPoison.
php
• Classic Man-in-the-middle attack
– Send ARP reply messages to device so they think your machine is
someone else
– Better than simple sniffing because not just best effort.
• Solutions
– Encrypt all traffic
– Monitoring programs like arpwatch to detect mapping changes
• Which might be valid due to DHCP
16. Basic IPv4 Routing
• Static routing. Used by hosts, firewalls and routers.
– Routing table consists of entries of
• Network, Next hop address, metric, interface
– May have routing table per incoming interface
– To route a packet, take the destination address and find the best
match network in the table. In case of a tie look at the metric
• Use the corresponding next hop address and interface to send the packet
on.
• The next hop address is on the same link as this device, so you use the
next hop’s data-link address, e.g. ethernet MAC address
– Decrement “time to live” field in IP header at each hop. Drop packet
when it reaches 0
• Attempt to avoid routing loops
• As internet got bigger, TTL fields got set bigger. 255 maximum
17. Routing example
• Receive a packet destined to 192.168.3.56 on inside
interface
• Local routing table for inside interface
– 192.168.2.0/30, 127.0.0.1, 1, outside
– 192.168.5.0/29, 127.0.0.1, 1, dmz
– 192.168.3.0/24, 192.168.5.6, 1, dmz
– 192.168.3.0/24, 192.168.1.2, 3, outside
– 0.0.0.0/0, 192.168.1.2, 1, outside
• Entries 3 and 4 tie. But metric for 3 is better
• Entries 1 and 2 are for directly connected networks
18. Source Based Routing
• In the IP Options field, can specify a source
route
– Was conceived of as a way to ensure some traffic
could be delivered even if the routing table was
completely screwed up.
• Can be used by the bad guy to avoid security
enforcing devices
– Most folks configure routers to drop packets with
source routes set
19. IP Options in General
• Originally envisioned as a means to add more
features to IP later
• Most routers drop packets with IP options set
– Stance of not passing traffic you don’t understand
– Therefore, IP Option mechanisms never really took off
• In addition to source routing, there are security
Options
– Used for DNSIX, a MLS network encryption scheme
20. Dynamic Routing Protocols
• For scaling, discover topology and routing rather than
statically constructing routing tables
– Open Shortest Path First (OSPF): Used for routing within an
administrative domain
– RIP: not used much anymore
– Border Gateway Protocol (BGP): Used for routing between
administrative domains. Can encode non-technical transit
constraints, e.g. Domain X will only carry traffic of paying
customers
• Receives full paths from neighbors, so it avoids counts to infinity.
21. Dynamic Routing
• Injecting unexpected routes a security concern.
– BGP supports peer authentication
– BGP blackholing is in fact used as a mechanism to
isolate “bad” hosts
– Filter out route traffic from unexpected (external)
points
– OSPF has MD5 authentication, and can statically
configure neighbor routers, rather than discover
them.
• Accidents are just as big of a concern as
malicious injections
22. Internet Control Message Protocol
(ICMP)
• Used for diagnostics
– Destination unreachable
– Time exceeded, TTL hit 0
– Parameter problem, bad header field
– Source quench, throttling mechanism rarely used
– Redirect, feedback on potential bad route
– Echo Request and Echo reply, ping
– Timestamp request and Timestamp reply, performance ping
– Packet too big
• Can use information to help map out a network
– Some people block ICMP from outside domain
23. Smurf Attack
• An amplification DoS attack
– A relatively small amount of information sent is expanded to
a large amount of data
• Send ICMP echo request to IP broadcast addresses.
Spoof the victim's address as the source
• The echo request receivers dutifully send echo replies
to the victim overwhelming it
• Fraggle is a UDP variant of the same attack
25. Transport Level – TCP and UDP
• Service to service communication.
– Multiple conversations possible between same pair of
computers
• Transport flows are defined by source and destination ports
• Applications are associated with ports (generally just destination
ports)
– IANA organizes port assignments http://www.iana.org/
• Source ports often dynamically selected
– Ports under 1024 are considered well-known ports
– Would not expect source ports to come from the well-known
range
26. Reconnaissance
Port scanning
Send probes to all ports on the target
See which ones respond
Application fingerprinting
Analyze the data returned
Determine type of application, version, basic
configuration
Traffic answering from port 8080 is HTTP, Apache
or Subversion
27. Datagram Transport
• User Datagram Protocol (UDP)
– A best-effort delivery, no guarantee, no ACK
– Lower overhead than TCP
– Good for best-effort traffic like periodic updates
– No long lived connection overhead on the endpoints
• Some folks implement their own reliable protocol over UDP to
get “better performance” or “less overhead” than TCP
– Such efforts don’t generally pan out
• TFTP and DNS protocols use UDP
• Data channels of some multimedia protocols, e.g., H.323 also
use UDP
29. DHCP
• Built on older BOOTP protocol (which was built on even older
RARP protocol)
– Used by diskless Suns
• Enables dynamic allocation of IP address and related
information
• Runs over UDP
• No security considered in the design, obvious problems
– Bogus DHCP servers handing out addresses of attackers
choice
– Bogus clients grabbing addresses
• IETF attempted to add DHCP authentication but rather late in
the game to do this.
• Other solutions
– Physically secure networks
– Use IPSec
30. Reliable Streams
• Transmission Control Protocol (TCP)
– Guarantees reliable, ordered stream of traffic
– Such guarantees impose overhead
– A fair amount of state is required on both ends
• Most Internet protocols use TCP, e.g., HTTP,
FTP, SSH, H.323 control channels
31. TCP Header
Source Port Destination Port
Sequence Number
Acknowledgement number
HDR
Len
U
R
G
A
C
K
P
S
H
R
S
T
S
Y
N
F
I
N
Window
Size
Checksum Urgent Pointer
Options (0 or more words)
33. Syn flood
• A resource DoS attack focused on the TCP three-way
handshake
• Say A wants to set up a TCP connection to B
– A sends SYN with its sequence number X
– B replies with its own SYN and sequence number Y and an ACK of
A’s sequence number X
– A sends data with its sequence number X and ACK’s B’s sequence
number Y
– Send many of the first message to B. Never respond to the
second message.
– This leaves B with a bunch of half open (or embryonic) connections
that are filling up memory
– Firewalls adapted by setting limits on the number of such half open
connections.
34. SYN Flood
Machine A Machine B
SYN:
seqno=100
SYN:
seqno=511
ACK = 100
SYN: seqno=89
SYN:
seqno=176
SYN:
seqno=344
35. SYN Flood Constrainer
Machine A FW
SYN:
seqno=100
SYN:
seqno=511
ACK = 100
ACK=511
SYN:
seqno=176
SYN:
seqno=344
Machine B
SYN: seqno=56
SYN:
seqno=677
ACK = 56
ACK=677
36. Another Syn Flood solution:
SYN cookie
Encode information in the sequence number, so
receiver does not need to save anything for half
open connection
t = counter , m = MSS, s = crypto function
computed over IP addresses and server port and t
(24 bits)
Seqno = (t mod 32) || m encoded in 3 bits || s (24
bits)
On receiving ACK, get original seqno by
subtracting 1
Check 1 to verify timeout
Recompute s to verify addresses and ports
37. SYN Flood
Machine A Machine B
SYN:
seqno=100
SYN:
seqno=511
ACK = 100
SYN: seqno=89
SYN:
seqno=176
SYN:
seqno=344
38. Session Hijacking
Take over a session after the 3 way handshake
is performed
After initial authentication too
Local
Can see all traffic.
Simply inject traffic at a near future sequence
number
Blind
Cannot see traffic
Must guess the sequence number
40. Application Protocols
• Single connection protocols
– Use a single connection, e.g. HTTP, SMTP
• Dynamic Multi-connection Protocols, e.g. FTP and
H.323
– Have a well known control channel
– Negotiate ports and/or addresses on the control channel for
subsidiary data channels
– Dynamically open the negotiated data channels
• Protocol suites, e.g. Netbios and DNS
41. Spoofing Applications
• Often times ridiculously easy
• Fake Client
– Telnet to an SMTP server and enter mail from
whoever you want
– Authenticating email servers
• Require a password
• Require a mail download before server takes send
requests
• Fake server
– Phishing: misdirect user to bogus server
42. Default Settings
Many applications installed with default users
and passwords
Wireless routers, SCADA systems
Default passwords for many of these systems
are easily found on the Internet
http://www.cirt.net/cgi-bin/passwd.pl
43. Domain Name System (DNS)
• Hierarchical service to resolve domain names to IP addresses.
– The name space is divided into non-overlapping zones
– E.g., consider shinrich.cs.uiuc.edu.
– DNS servers in the chain. One for .edu, one for .uiuc.edu,
and one for .cs.uiuc.edu
• Can have primary and secondary DNS servers per zone. Use
TCP based zone transfer to keep up to date
• Like DHCP, no security designed in
– But at least the DNS server is not automatically discovered
– Although this information can be dynamically set via DHCP
44. DNS Problems
• DNS Open relays
– Makes it look like good DNS server is authoritative
server to bogus name
– Enables amplification DoS attack
http://www.us-cert.gov/reading_room/DNS-recursion0330
06.pdf
DNS Cache Poisoning
– Change the name to address mapping to something
more desirable to the attacker
http://www.secureworks.com/research/articles/cachepois
oning
– Dan Kaminsky raised issue again last summer
http://www.linuxjournal.com/content/understanding-kamin
46. DNS Communication
Use UDP
Requests and responses have matching 16 bit
transaction Ids
Servers can be configured as
Authoritative Nameserver
Officially responsible for answering requests for a domain
Recursive
Pass on requests to other authoritative servers
Both (this can be the problem)
47. DNS Open Relay
Y: D N S S erver
A uthoritative for big.com
R ecursion enabled for all Internet
Z : A ttacker
X : Victim
S rc= X dst=Y
W hat is address of bob.com ?
S rc=Y dst=X
bob.com = 1.2.3.4
48. Good DNS Deployment
Y : D N S S erver
R ecursive
Only accepts local requests
Internet
Z: A ttacker
X : Victim
S rc=X dst=Y
W hat is address of bob.com ?
W : D N S S erver
A uthoritative for big.com
S rc= X dst=W
W hat is address of big.com?
S rc=X dst= W
W hat is address of bob.com?
49. DNS Cache Poisoning
Older implementations would just accept
additional information in a reply
e.g. A false authoritative name server
Fixed by bailiwick checking. Additional records only
include entries from the requested domain
Now to spoof a reply must anticipate the correct
transaction ID
Only 16 bits
Random selection of ID isn't always the greatest
50. Bailiwick Checks
$ dig @ns1.example.com www.example.com
;; ANSWER SECTION:
www.example.com. 120 IN A 192.168.1.10
;; AUTHORITY SECTION:
example.com. 86400 IN NS
ns1.example.com.
example.com. 86400 IN NS
ns2.example.com.
;; ADDITIONAL SECTION:
ns1.example.com. 604800 IN A 192.168.2.20
ns2.example.com. 604800 IN A 192.168.3.30
www.linuxjournal.com. 43200 IN A 66.240.243.113
52. Kaminsky's Observations
Most implementations don't randomize source
ports (making the TID collision more likely)
Try to poison through the additional information
(side stepping the bailiwick check)
$ dig doesnotexist.example.com
;; ANSWER SECTION:
doesnotexist.example.com. 120 IN A 10.10.10.10
;; AUTHORITY SECTION:
example.com. 86400 IN NS
www.example.com.
;; ADDITIONAL SECTION:
www.example.com. 604800 IN A 10.10.10.20
53. DNSSEC
• Seeks to solve the trust issues of DNS
• Uses a key hierarchy for verification
• Has been under development for over a
decade and still not really deployed
This year articles say root servers for .edu, .org,
and .com will be deployed in 2010, 2011
timeframe.
• Provides authentication, not confidentiality
• DNS Threat Analysis in RFC 3833.
54. Key Points
Network is complex and critical
Many flaws have been simple implementation
problems
Poor configuration also can cause widespread
problems
Other guys problems can affect me
Next, what can you do about it?
Editor's Notes
#24:
This is a diagram of a “Smurf” or “Fraggle” attack.
The single stream from the perpetrator to the broadcast LANrepresents the flow of information from the perpetrator to the broadcast LAN, usually several packets per second of ICMP echo (“Smurf”) or UDP echo (“Fraggle”) traffic spoofed to look like it is coming from the victim’s system.
If the router at the edge of the LAN forwards the broadcast ping to the LAN, each device on the LAN will respond with an echo-reply (ICMP) or will bounce the traffic (UDP), creating a multiplication of the original traffic flow. The traffic is then directed to the victim.
There are usually several bounce sites involved, used to increase the factor by which traffic is multiplied.
This attack is characterized by many ICMP echo reply packets at the victim’s site or many UDP packets involving the diagnostic “echo” port.