CSE543 - Introduction to Computer and Network Security Module: Network Security · 2008-12-12 · Security Problems in the TCP/IP Protocol Suite 6. CSE543 - Introduction to Computer

CSE543 - Introduction to Computer and Network Security Page

CSE543 - Introduction to Computer and Network Security

Module: Network Security

Professor Patrick McDanielFall 2008

1


Networking• Fundamentally about transmitting information between

two devices• Direct communication is now possible between any

two devices anywhere (just about)‣ Lots of abstraction involved‣ Lots of network components‣ Standard protocols‣ Wired and wireless‣ Works in protection environment

• What about ensuring security?

2


Network Security• Every machine is connected‣ What is trust model of the network?

• Not just limited to dogs as users‣ What other ‘dogs’ are out there?

3


Exploiting the network ...• The Internet is extremely vulnerable to attack‣ it is a huge open system ...‣ which adheres to the end-to-end principle

• smart end-points, dumb network

• Can you think of any large-scale attacks that would be 4


E2E Argument• Clark et. al discussed a property of good systems that

says features should be placed as close to resources as possible‣ In communication, this means that we want the middle

of the network to be simple, and the end-points to be smart (e.g., do everything you can at the end-points• “Dumb, minimal network”

‣ This is the guiding principle of IP (Internet)‣ Q: Does this have an effect on security?

• Note: this is a departure from the early networks which smart network, dumb terminals

5


• Bellovin’s observations about the security problems of the IP protocols‣ Not really a study of how IP is misused, e.g., IP addresses for

authentication, but really what is inherently bad about the way in which IP is setup

• A really, really nice overview of the basic ways in which security and the IP design is at odds

Security Problems in the TCP/IP Protocol Suite

6


• TCP/IP uses a three-way handshake to establish a connection

1. C -> S: QC

2. S -> C: QS, ack(QC) where sequence number QS is nonce

3. C -> S: ack(QS) … then send data

2. However assume the bad guy does not hear msg 2, if he can guess QS, then he can get S to accept whatever data it wants (useful if doing IP authentication, e.g., “rsh”)

Client Server

Adversary

Sequence number prediction

7


• The only way you really fix this problem to stop making the sequence numbers predictable:‣ Randomize them -- you can use DES or some other

mechanism to generate them randomly‣ There is an entire sub-field devoted to the creation and

management of randomness in OSes

• Also, you could look for inconsistencies in timing information‣ Assumption: the adversary has different timing than‣ OK, maybe helpful, but far from definitive

Sequence Number Prediction (fixes)

8


• RIP - routing information protocol‣ Distance vector routing protocol used for local network

‣ Routers exchange reachability and “distance” vectors for all the sub-networks within (a typically small) domain

‣ Use vectors to decide which is best, notification of changes is propagated quickly

• So, the big problem is that you receive vast amounts of data that a router uses to form the routing table‣ So, just forge that, and the game is up

‣ Manipulate paths, DOS, hijack connections, etc.

• Solutions:

Routing Manipulation

9


• ICMP is used as a control plane for IP messages‣ Ping (connectivity probe)

‣ Destination Unreachable (error notification)

‣ Time-to-live exceeded (error notification)

• These are used for good purposes, and are largely indispensable tools for network management and control‣ Error notification codes can be used to reset connections

without any

• Solution: verify/sanity check sources and content‣ ICMP “returned packets”

Internet Control Message Protocol

10


• In 1996, someone discovered that many operating systems, routers, etc. could be crash/rebooted by sending a single malformed packet‣ It turns out that you can send a IP packet larger than 65,535

(216), it would crash the system

‣ The real reason lies in the way fragmentation works• It allows somebody to send a packet bigger than IP allows

• Which blows up most fixed buffer size implementations

• … and dumps core, blue screen of death, etc.

‣ Note: this is not really ICMP specific, but easy (try it)

% ping -l 65510 your.host.ip.address

• This was a popular pastime of early hackers

The “ping of death” …

11


• Protocol used to map IP address onto the physical layer addresses (MAC) 1) ARP request: who has x.x.x.x? 2) ARP response: me!

• Policy: last one in wins• Used to forward packets on the appropriate interfaces

by network devices (e.g., bridges)

• Q: Why would you want to spoof an IP address?

Address Resolution Protocol

12


• Attack: replace good entries with your own• Leads to ‣ Session hijacking‣ Man-in-the-middle attacks‣ Denial of service, etc.

• Lots of other ways to abuse ARP.• Nobody has really come up with a good solution‣ Except smart bridges, routers that keep track of MACs

• However, some not worried‣ If adversary is in your perimeter, you are in big trouble‣ You should never should validate the source of each pack

ARP poisoning

13


• Finger user identity (my advisor hated this)‣ host gives up who is logged in, existence of identities

PSU.local Presentations > finger megan Login: megan Name: Megan SmithDirectory: /Users/megan Shell: /bin/bashLast login Mon 23 Aug 13:19 (EDT) on consoleNo Mail.No Plan.PSU.local Presentations >

• This is horrible in a distributed environment‣ Privacy, …

‣ Lots of information to start a compromise of the user.

Other flawed protocols/services

14


• Post office protocol - mail retrieval‣ Passwords passed in the clear (duh)‣ Solution: SSL, SSH, Kerberos

• Simple mail transport protocol (SMTP) - email‣ Nothing authenticated: SPAM‣ Nothing hidden: eavesdropping‣ Solution: your guess is as good as mine

• File Transfer protocol - file retrieval‣ Passwords passed in the clear (duh)‣ Solution: SSL, SSH, Kerberos

POP/SMTP/FTP

15


• DNS maps between IP address (12.1.1.3) and domain and host names (ada.cse.psu.edu)‣ How it works: the “root” servers redirect you to the top

level domains (TLD) DNS servers, which redirect you to the appropriate sub-domain, and recursively ….

‣ Note: there are 13 “root” servers that contain the TLDs for .org, .edu, and country specific registries (.fr, .ch)

Host (resolver)

cse.psu.eduroot psu.edu.edu

ada.cse.psu.edu? 130.203.16.130

DNS - The domain name system

16


• Nothing is authenticated, so really the game is over‣ You can not really trust what you hear …‣ But, many applications are doing just that.‣ Spoofing of DNS is really dangerous

• Moreover, DNS is a catalog of resources‣ Zone-transfers allow bulk acquisition of DNS data‣ … and hence provide a map for attacking the network

• Lots of opportunity to abuse the system‣ Relies heavily on caching for efficiency -- cache pollution‣ Once something is wrong, it can remain that way in caches

for a long time (e.g., it takes a long time flush)‣ Data may be corrupted before it gets to authoritative server

DNS Vulnerabilities

17


• A standard-based (IETF) solution to security in DNS‣ Prevents data spoofing and corruption‣ Public key based solution to verifying DNS data‣ Authenticates

• Communication between servers• DNS data• Public keys (a bootstrap for PKI?)

DNSsec

18


• TSIG : transaction signatures protect DNS operations‣ Zone loads, some server to server requests (master -> slave), etc.

‣ Time-stamped signed responses for dynamic requests‣ A misnomer -- it currently uses shared secrets for TSIG

(HMAC) or do real signatures using public key cryptography

• SIG0: a public key equivalent of TSIG‣ Works similarly, but with public keys‣ Not as popular as TSIG, being evaluated

• Note: these mechanisms assume clock sync. (NTP)

DNSsec Mechanisms

19


• Securing the DNS records‣ Each domain signs their “zone” with a private key‣ Public keys published via DNS‣ Indirectly signed by parent zones‣ Ideally, you only need a self-signed root, and follow keys down

the hierarchy

cse.psu.eduroot psu.edu.eduSigns Signs Signs

DNSsec Mechanisms

20


• Incremental deployability‣ Everyone has DNS, can’t assume a flag day

• Resource imbalances‣ Some devices can’t afford real authentication

• Cultural‣ Most people don’t have any strong reason to have secure

DNS ($$$ not justified in most environments)

‣ Lots of transitive trust assumptions (you have no idea how the middlemen do business)

• Take away: DNSsec will be deployed, but it is unclear whether it will be used appropriately/widely

DNSsec challenges

21


Communications Security• Harden the communication against mal-intent ...

22http://artfiles.art.com/images/-/Philip-Gendreau/Jolly-Roger-the-Pirates-Flag-Giclee-Print-C10274009.jpeg


Communications Security• A host wants to establish a secure channel to remote

hosts over an untrusted network‣ Not Login – end-users may not even be aware that

protections in place‣ Remote hosts may be internal or external

• The protection service must …‣ Authenticate the end-points (each other)‣ Negotiate what security is necessary (and how)‣ Establish a secure channel‣ Process the traffic between the end points

23


IPsec (not IPSec!)• Host level protection service‣ IP-layer security (below TCP/UDP)‣ De-facto standard for host level security‣ Developed by the IETF (over many years)‣ Now available in most operating systems

• E.g., Available in XP, OS X, Linux, BSD*, …

‣ Implements a wide range of protocols and cryptographic algorithms

• Provides ….‣ Confidentiality, integrity, authenticity, replay protection, DOS

protection

24


IPsec and the IP protocol stack• IPsec puts the two main

protocols in between IP and the other protocols‣ AH - authentication header‣ ESP - encapsulating security

payload• Tunnel vs. transport?‣ Key management/authentication‣ Policy

• Other function provided by external protocols and architectures

SMTPFTP

TCPHTTP

ESPAH

IP

UDP

25


Tunneling• “IP over IP”‣ Network-level packets are encapsulated ‣ Allows traffic to avoid firewalls

IP layer…

IP layer…

26


IPsec Protocol Suite

(IKE)Internet Key

Exchange

(AH)Authentication

Header

(ESP)Encapsulating

Security Payload

(SPS)Security Policy

SystemManual

Policy/Configuration

ManagentKey

ManagementPacket

Processing

27


Internet Key Exchange (IKE)• Built on of ISAKMP framework• Two phase protocol used to establish parameters

and keys for session‣ Phase 1: authenticate peers, establish secure channel‣ Phase 2: negotiate parameters, establish a security

association (SA)

• The details are unimaginably complex• The SA defines algorithms, keys, and policy used to

secure the session

28


IPsec

IPsec: Packet Handling (Bump …)

Network (IP)

Physical

Application

Presentation

Session

Transport

Data Link

IP Protocol Stack

SADB

29


Authentication Header (AH)• Authenticity and integrity‣ via HMAC‣ over IP headers and data

• Advantage: the authenticity of data and IP header information is protected‣ it gets a little complicated with mutable fields, which are

supposed to be altered by network as packet traverses the network

‣ some fields a immutable, and are protected

• Confidentiality of data is not preserved• Replay protection via AH sequence numbers‣ note that this replicates some features of TCP (good?)

30


IPsec AH Packet Format

IPv4 Header

IPv4 AH Packet Format

Next Header Length Reserved

Security Parameter Index

Authentication Data (variable number of 32-bit words)

AH Header Format

31

Authentication HeaderHigher Level

Protocol Data


Authentication Header (AH)• Modifications to the packet format

32

IP Header AH HeaderMAC Payload

AH Packet

EncryptedAuthenticated

IP Header Payload


IPsec Authentication• SPI: (spy) identifies the security association for this packet– Type of crypto checksum, how large it is, and how it is computed

– Really the policy for the packet

• Authentication data– Hash of packet contents include IP header as as specified by SPI

– Treat transient fields (TTL, header checksum) as zero

• Keyed MD5 Hash is default

Headers and data being sentKey KeySecret Key

MD5 Hash

33


Encapsulating Security Payload (ESP)

• Confidentiality, authenticity and integrity‣ via encryption and HMAC‣ over IP payload (data)

• Advantage: the security manipulations are done solely on user data‣ TCP packet is fully secured‣ simplifies processing

• Use “null” encryption to get authenticity/integrity only• Note that the TCP ports are hidden when encrypted‣ good: better security, less is known about traffic‣ bad: impossible for FW to filter/traffic based on port

34


IPsec ESP Packet Format

IPv4 ESP Packet Format

IP HeaderOther IP Headers

ESP Header Encrypted Data

ESP Header FormatSecurity Parameter Identifier (SPI)

Opaque Transform Data, variable length

Unencrypted Encrypted

Security Parameters Index (SPI)

Initialization Vector (optional)

Replay Prevention Field (incrementing count)

Payload Data (with padding)

Authentication checksum

DES + MD5 ESP Format

35


Encapsulating Security Payload

• Modifications to packet format

36

IP Header ESP Header Payload ESP Trailer MAC

ESP Packet

EncryptedAuthenticated

IP Header Payload


Practical Issues and Limitations• IPsec implementations‣ Often not compatible (ungh.)‣ Large footprint

• resource poor devices are in trouble• New standards to simplify (e.g, JFK, IKE2)

‣ Slow to adopt new technologies

• Issues‣ IPsec tries to be “everything for everybody at all times”

• Massive, complicated, and unwieldy

‣ Policy infrastructure has not emerged ‣ Large-scale management tools are limited (e.g., CISCO)‣ Often not used securely (common pre-shared keys)

37


Isolation• Countermeasure to physically separate the devices/

environment from mal-intent

38


Network Isolation: VPNs• Idea: I want to create a collection of hosts which

operate in a coordinated way‣ E.g., a virtual security perimeter over physical network‣ Hosts work as if they are isolated from malicious hosts

• Solution: Virtual Private Networks‣ Create virtual network topology over physical network‣ Use communications security protocol suites to secure

virtual links “tunneling”‣ Manage networks as if they are physically separate‣ Hosts can route traffic to regular networks (split-tunneling)

39


VPN Example: RW/Telecommuter

Internet

LAN

(network edge)

Physical LinkLogical Link (IPsec)

40


VPN Example: Hub and Spoke

Internet

LAN

(network edge)


41


VPN Example: Mesh

Internet

LAN

(network edge)


42


Virtual LANs (VLANs)• VPNs build with hardware‣ No encryption – none needed‣ “wire based isolation”‣ Switches increasingly support VLANs‣ Allows networks to be reorganized without rewiring

• Example usage: two departments in same hallway‣ Each office is associated with department‣ Configuring the network switch gives physical isolation‣ Note: often used to ensure QoS

43


Aside: Malware• Malware - software that exhibits malicious behavior

(typically manifest on user system)‣ virus - self-replicating code, typically transferring by shared

media, filesystems, email, etc.

‣ worm - self propagating program that travels over the network

• The behaviors are as wide ranging as imagination‣ backdoor - hidden entry point into system that allows quick

access to elevated privileges

‣ rootkit - system replacement that hides adversary behavior

‣ key logger - program that monitors, records, and potentially transmits keyboard input to adversary

44


Worms

45


Worms• A worm is a self-propagating program.• As relevant to this discussion

1. Exploits some vulnerability on a target host …2. (often) imbeds itself into a host …3. Searches for other vulnerable hosts …4. Goto (1)

• Q: Why do we care?46


The Danger• What makes worms so dangerous is that infection

grows at an exponential rate‣ A simple model:

• s (search) is the time it takes to find vulnerable host

• i (infect) is the time is take to infect a host

‣ Assume that t=0 is the worm outbreak, the number of hosts at t=j is

2(j/(s+i))

‣ For example, if (s+i = 1), what is it at time t=32?

47


The result

0

500,000,000

1,000,000,000

1,500,000,000

2,000,000,000

2,500,000,000

3,000,000,000

3,500,000,000

4,000,000,000

4,500,000,000

5,000,000,000

48


The Morris Worm• Robert Morris, a 23 doctoral student from Cornell‣ Wrote a small (99 line) program

‣ November 3rd, 1988

‣ Simply disabled the Internet

• How it did it‣ Reads /etc/password, they tries the obvious choices and

dictionary, /usr/dict words

‣ Used local /etc/hosts.equiv, .rhosts, .forward to identify hosts that are related• Tries cracked passwords at related hosts (if necessary)

• Uses whatever services are available to compromise other hosts

‣ Scanned local interfaces for network information49


Code Red• Anatomy of a worm: Maiffret (good reading)• Exploited a Microsoft IIS web-server vulnerability‣ A vanilla buffer overflow (allows adversary to run code)‣ Scans for vulnerabilities over random IP addresses‣ Sometimes would deface the served website

• July 16th, 2001 - outbreak‣ CRv1- contained bad randomness (fixed IPs searched)‣ CRv2 - fixed the randomness,

• added DDOS of www.whitehouse.gov• Turned itself off and on (on 1st and 16th of month)

‣ August 4 - Code Red II• Different code base, same exploit

50


Worms and infection• The effectiveness of a worm is determined by how good it is at

identifying vulnerable machines‣ Morris used local information at the host‣ Code Red used what?

• Multi-vector worms use lots of ways to infect‣ E.g., network, DFS partitions, email, drive by downloads …‣ Another worm, Nimda did this

• Lots of scanning strategies‣ Signpost scanning (using local information, e.g., Morris)‣ Random IP - good, but waste a lot of time scanning dark or

unreachable addresses (e.g., Code Red)‣ Local scanning - biased randomness‣ Permutation scanning - instance is given part of IP space

51


Other scanning strategies• The doomsday worm: a flash worm

‣ Create a hit list of all vulnerable hosts

• Staniford et al. argue this is feasible

• Would contain a 48MB list

‣ Do the infect and split approach

‣ Use a zero-day vulnerability

• Result: saturate the Internet is less than 30 seconds!52

0

500,000,000

1,000,000,000

1,500,000,000

2,000,000,000

2,500,000,000

3,000,000,000

3,500,000,000

4,000,000,000

4,500,000,000

5,000,000,000


Worms: Defense Strategies• (Auto) patch your systems: most, if not all, large worm outbreaks

have exploited known vulnerabilities (with patches)• Heterogeneity: use more than one vendor for your networks• Shield (Ross): provides filtering for known vulnerabilities, such that

they are protected immediately (analog to virus scanning)

• Filtering: look for unnecessary or unusual communication patterns, then drop them on the floor ‣ This is the dominant method, getting sophisticated (Arbor Networks)

OperatingSystem

Network Interface

Shield NetworkTraffic

53


Denial of Service• Intentional prevention of access to valued

resource‣CPU, memory, disk (system resources)‣DNS, print queues, NIS (services)‣Web server, database, media server (applications)

• This is an attack on availability (fidelity)• Note: launching DOS attacks is easy• Note: preventing DOS attacks is hard‣Mitigation the path most frequently traveled

54


Canonical (common) DOS - Request Flood

• Attack: request flooding‣ Overwhelm some resource with legitimate requests

‣ e.g., web-server, phone system

• Note: unintentional flood is called a flash crowd55


Example: SMURF Attacks• This is one of the deadliest and simplest of the DOS attacks

(called a naturally amplified attack)‣ Send a large number PING packet networks on the broadcast IP

addresses (e.g., 192.168.27.254)

‣ Set the source packet IP address to be your victim

‣ All hosts will reflexively respond to the ping at your victim

‣ … and it will be crushed under the load.

‣ Fraggle: UDP based SMURF

Host

Host Host Host

Host

Host

Host

Host

Host

adversary Broadcast victim

56


Distributed denial of service• DDOS: Network oriented attacks aimed at

preventing access to network, host or service‣ Saturate the target’s network with traffic‣Consume all network resources (e.g., SYN)‣Overload a service with requests

• Use “expensive” requests (e.g., “sign this data”)

‣Can be extremely costly (e.g, Amazon)

• Result: service/host/network is unavailable• Frequently distributed via other attack

• Note: IP is often hidden (spoofed)57


D/DOS (generalized by Mirkovic)

• Send a stream of packets/requests/whatever …‣ many PINGS, HTML requests, ...

• Send a few malformed packets ‣ causing failures or expensive error handling‣ low-rate packet dropping (TCP congestion control)‣ “ping of death”

• Abuse legitimate access‣ Compromise service/host‣ Use its legitimate access rights to consume the rights for

domain (e.g., local network)‣ E.g., First-year graduate student runs a recursive file

operation on root of NFS partition58


The canonical DDOS attack

Internet

LAN

(target)(zombies)

(router)(master)

(adversary)59


Adversary Network

(adversary)

(masters)

(zombies)

(target)

60


Why DDOS• What would motivate someone DDOS?‣ An axe to grind …

‣ Curiosity (script kiddies) …

‣ Blackmail

‣ Information warfare …

• Internet is an open system ...‣ Packets not authenticated, probably can’t be

• Would not solve the problem just move it (firewall)

61


Why is DDOS possible? • Interdependence - services dependent on each other‣ E.g., Web depends on TCP and DNS, which depends on

routing and congestion control, …

• Limited resources (or rather resource imbalances)‣ Many times it takes few resources on the client side to

consume lots of resources on the server side

‣ E.g., SYN packets consume lots of internal resources

• You tell me .. (as said by Mirkovic et al.)‣ Intelligence and resources not co-located

‣ No accountability

‣ Control is distributed62


DDOS and the E2E • E2E (a simplified version): We should design the

network such that all the intelligence is at the edges.‣ So that the network can be more robust and scalable ‣ Many think is the main reason why the Internet works

• Downside:‣ Also, no real ability to police the traffic/content‣ So, many security solutions break this E2E by cracking open

packets (e.g., application level firewalls)‣ DDOS is real because of this …

63


Q: An easy fix?• How do you solve distributed denial of service?

64


Simple DDOS Mitigation• Ingress/Egress Filtering‣ Helps spoofed sources, not much else

• Better Security ‣ Limit availability of zombies, not feasible‣ Prevent compromise, viruses, …

• Quality of Service Guarantees (QOS)‣ Pre- or dynamically allocate bandwidth ‣ E.g., diffserv, RSVP‣ Helps where such things are available …

• Content replication‣ E.g,. CDS‣ Useful for static content

65


DOS Prevention - Reverse-Turing Tests

• Turing test: measures whether a human can tell the difference between a human or computer (AI)

• Reverse Turning tests: measures whether a user on the internet is a person, a bot, whatever?

• CAPTCHA - completely automated public Turing test to tell computers and humans apart‣ contorted image humans can read, computers can’t‣ image processing pressing SOA, making these harder

• Note: often used not just for DOS prevention, but for protecting “free” services (email accounts)

66


DOS Prevention - Puzzles• Make the solver present evidence of “work” done ‣ If work is proven, then process request‣ Note: only useful if request processing significantly more work

than

• Puzzle design‣ Must be hard to solve‣ Easy to Verify

• Canonical Example‣ Puzzle: given all but k-bits of r and h(r), where h is a

cryptographic hash function‣ Solution: Invert h(r)‣ Q: Assume you are given all but 20 bits, how hard would it be to

solve the puzzle?67


Pushback• Initially, detect the DDOS‣ Use local algorithm, ID-esque processing‣ Flag the sources/types/links of DDOS traffic

• Pushback on upstream routers‣ Contact upstream routers using PB protocol‣ Indicate some filtering rules (based on observed)

• Repeat as necessary towards sources‣ Eventually, all (enough) sources will be filtered

• Q: What is the limitation here?

R1 R2 R3 R4R1 R2 R3 R4

68


Traceback• Routers forward packet data to source‣ Include packets and previous hop …‣ At low frequency (1/20,000) …

• Targets reconstruct path to source (IP unreliable)‣ Use per-hop data to look at‣ Statistics say that the path will be exposed

• Enact standard‣ Add filters at routers along the path

R1 R2 R3 R4

R1 R2 R3

69


DDOS Reality• None of the “protocol oriented” solutions have really

seen any adoption‣ too many untrusting, ill-informed, mutually suspicious parties

must play together well (hint: human nature)‣ solution have many remaining challenges

• Real Solution‣ Large ISP police there ingress/egress points very carefully‣ Watch for DDOS attacks and filter appropriately

• e.g., BGP (routing) tricks, blacklisting, whitelisting

‣ Products in existing that coordinate view from many points in the network to identify upswings in

‣ Interestingly, this is the same way they deal with worms ...

70

Documents

CSE543 - Introduction to Computer and Network Security Module: Network Security · 2008-12-12 · Security Problems in the TCP/IP Protocol Suite 6. CSE543 - Introduction to Computer