Transcript Denial of service (DOS) - Computer Science and Engineering

Denial of Service (DoS)
By
Vijay C Uyyuru, Prateek Arora, & Terry Griffin
Overview
•
•
•
•
•
•
Introduction
Background
Benchmarks and Metrics
Requirements
Summary of Methods
Conclusion
Vijay C Uyyuru
Prateek Arora
Terry Griffin
What is denial of service attack?
• When a denial of service (DoS) attack
occurs, a computer or a network user is
unable to access resources like e-mail and
the Internet. An attack can be directed at
an operating system or at the network.
Denial of Service
DoS
Bad guy
Compromised
host
Victim
Third parties
What is distributed denial of
service?
• A distributed denial of service (DDoS)
attack is accomplished by using the
Internet to break into computers and using
them to attack a network. Hundreds or
thousands of computer systems across
the Internet can be turned into “zombies”
and used to attack another system or
website.
Distributed Denial of Service
• DDoS
Bad guy
Master
agent
Victim (s)
Slave agents
(zombies, bots)
Owned
host
Third parties
Brief history and trends
• DoS attacks started at around early ’90s.
• At the first stage they were quite "primitive",
involving only one attacker exploiting maximum
bandwidth from the victim, denying others the
ability to be served. This was done mainly by
using simple methods of ping floods, SYN floods
and UDP floods.
• These attacks had to be "manually"
synchronized by a lot of attackers in order to
cause an effective damage.
Brief history and trends
• The shift to automating this synchronization,
coordination and generating a parallel massive
attack became public in 1997, with the release of
the first publicly available DDoS attacks tool,
Trinoo.
• In the following years, few more tools were
published – TFN (tribe flood network), TFN2K,
and Stacheldraht ("Barbed wire" in German).
Massive attack on public sites
Massive attack on public sites
• The subject came to public awareness only after
a massive attack on public sites on February
2000. During a period of three days the sites of
Yahoo.com, amazon.com, buy.com, cnn.com &
eBay.com were under attack.
• Analysts estimated that Yahoo! Lost $500,000 in
e-commerce and advertising revenue when it
was knocked offline for three hours.
Interesting Facts
• It turned out that about fifty computers at
Stanford University, and also computers at the
University of California at Santa Barbara, were
amongst the zombie computers sending pings in
these DoS attacks.
• A study during a period of three weeks in
February 2001 showed that there were about
4000 DoS attacks each week. Most DoS attacks
are neither publicized in the news media nor
prosecuted in courts.
Other major attack
• May 2001 - hackers overloaded Weather.com
routers and those of its Web hosting company
with bogus traffic. To counter the attack,
weather.com moved to another dedicated router
and installed filtering software to protect
switches and servers, as well as intrusion
detection software to record all ongoing activity.
It took the company 7 hours to bring the site
back up.
How does an attack work?
• One way to attack a company’s network or
website is to flood its systems with
information.
• Web and e-mail servers can only handle a
finite amount of traffic and an attacker
overloads the targeted system with
packets of data.
Impact
• Denial-of service attacks can essentially disable
the computer or the network. Depending on the
nature of the enterprise, this can disable your
organization.
• Some denial-of-service attacks can be executed
with limited resources against a large,
sophisticated site. This type of attack is
sometimes called an “asymmetric attack”.
• For example, an attacker with an old PC and a
slow modem may be able to disable much faster
and more sophisticated machines or network.
Dollar amount of losses by type!
Attack classification
•
DoS attacks exploit the asymmetric nature of
certain types of network traffic. One attack
method seeks to cause the target to use more
resources processing traffic than the attacker
does sending the traffic. Another method is to
control multiple attackers. Therefore DoS
attacks can be classified into three categories
1. Bandwidth/Throughput Attacks
2. Protocol Attacks
3. Software Vulnerability Attacks
Bandwidth/Throughput Attacks
•
•
•
•
Ping Flood Attack (ICMP echo)
SYN Flood Attack (DoS attack)
DDoS Attack (Distributed SYN Flood)
UDP Flood Attacks
Ping Flood Attack
• An attempt by an attacker on a high
bandwidth connection to saturate a
network with ICMP echo request packets
in order to slow or stop legitimate traffic
going through the network.
SYN Flood Attack
DDoS Attack
• The idea behind this attack is focusing
Internet connection bandwidth of many
machines upon one or a few machines.
This way it is possible to use a large array
of smaller (or “weaker”) widely distributed
computers to create the big flood effect.
UDP Flood Attacks
• UDP protocol is a connectionless unreliable
protocol which doesn't require session
negotiation between client and server
application. UDP provides easy to use interface
for producing large quantity of packets.
• A common attack which exploits UDP simply
floods the network with UDP packets destined to
a victim's host. Due to the relative simplicity of
this protocol an attacker can produce large
bandwidth capacity with relatively small effort.
Protocol Attacks
• Smurf Attack
• DNS name server Attack
Smurf Attack
• In this attack, spoofed IP packets containing
ICMP Echo-Request with a source address
equal to that of the attacked system and a
broadcast destination address are sent to the
intermediate network.
• Sending a ICMP Echo Request to a broadcast
address triggers all hosts included in the
network to respond with an ICMP response
packet, thus creating a large mass of packets
which are routed to the victim's spoofed
address.
Smurf Attack (contd.)
DNS name server Attack
• The most common method seen involves an intruder
sending a large number of UDP-based DNS
requests to a Nameserver using a spoofed source
IP address. Any Nameserver response is sent back
to the spoofed IP address as the destination.
• In this scenario, the spoofed IP address represents
the victim of the denial of service attack. The
Nameserver is an intermediate party in the attack.
The true source of the attack is difficult for an
intermediate or a victim site to determine due to the
use of spoofed source addresses.
DNS name server Attack (contd.)
Software Vulnerability Attacks
• Land Attack
• Ping of Death Attack
• Fragmentation Attack and Teardrop Attack
Land Attack
• In this attack, an attacker sends spoofed TCP SYN packets, with the
same source and destination addresses as the victim's host
address.
• In some TCP/IP stack implementations those kinds of packets may
cause the victim's host to crash.
• Any remote user that can send spoofed packets to a host can crash
or "hang" that host.
• Possible solution for this attack is to block IP-spoofed packets.
Attacks like those of the Land tool rely on the use of forged packets,
that is, packets where the attacker deliberately falsifies the origin
address. With the current IP protocol technology, it is impossible to
eliminate IP-spoofed packets. However, you can reduce the
likelihood of your site's networks being used to initiate forged
packets by filtering outgoing packets that have a source address
different from that of your internal network.
Land Attack (contd.)
• In cases where the victim's host is a router, this
attack may result in a routing loop consuming
large quantities of bandwidth (unless filtered in
advance).
• One of the variations of this attack targets a
certain TCP service provided by the victim. In
this case the attacker uses the same source and
destination ports which used by the victim's
service. This may consume the victim's host
CPU resources.
Land Attack (contd.)
• Here DUT is the Device Under Test
Ping of Death Attack
• Ping of Death is an attempt by an attacker
to crash, reboot or freeze a system by
sending an illegal ICMP (over IP) packet to
the host under attack.
• The TCP/IP specification allows for a
maximum packet size of up to 65536
octets. In some TCP stack implementation
encountering packets of greater size may
cause the victim's host to crash.
Ping of Death Attack (contd.)
• Most implementations of the ICMP
protocol use packet header size of 8
octets but allow the user to specify larger
packet header sizes.
• In the attack, the ICMP packet is sent in
the form of a fragmented message which,
when reassembled is larger than the
maximum legal IP packet size.
Ping of Death Attack (contd.)
Teardrop Attack
•
A normal packet is sent. A second packet is sent which has a fragmentation
offset claiming to be inside the first fragment. This second fragment is too
small to even extend outside the first fragment. This may cause an
unexpected error condition to occur on the victim host which can cause a
buffer overflow and possible system crash on many operating systems.
•
Teardrop attacks target a vulnerability in the way fragmented IP packets are
reassembled. Fragmentation is necessary when IP Datagrams are larger
than the maximum transmission unit (MTU) of a network segment across
which the Datagrams must traverse. In order to successfully reassemble
packets at the receiving end, the IP header for each fragment includes an
offset to identify the fragment's position in the original un-fragmented
packet. In a Teardrop attack, packet fragments are deliberately fabricated
with overlapping offset fields causing the host to hang or crash when it tries
to reassemble them.
Teardrop Attack (contd.)
•
In the following figure, a source test port simulates a Teardrop attack by
sending one, and then many IP packet fragments with overlapping
Fragment Offset fields. This attack traffic is first sent to the Device Under
Test (DUT) interface connected to the source test port and then to the
DUT's loopback address. The DUT's ability to drop this attack traffic is
verified. Finally, normal background traffic is sent at the same time as attack
traffic, so the DUT's performance during a Teardrop attack can be
measured.
Frequency & Scope
• How prevalent are denial-of-service
attacks in the Internet today?
• Researchers at the Cooperative Association for Internet
Data Analysis (CAIDA) address this question in their
paper, “Inferring Internet Denial-of-Service Activity”.
Using a technique called backscatter analysis, the
researchers monitored unsolicited traffic to unpopulated
address space. Their theory is that DoS traffic that uses
random spoofed source addresses will generate some
response traffic to the entire Internet address space,
including unpopulated space.
Frequency & Scope (contd.)
• Their results in February’ 2001 were that using
backscatter analysis, they observed 12,805 attacks on
over 5,000 distinct Internet hosts belonging to more than
2,000 distinct organizations during a three-week period.
• In addition, CAIDA reports that 90% of attacks last for
one hour or less; 90% are TCP based attacks, and
around 40% reach rates of 500 Packets Per Second
(PPS) or greater.
• Analyzed attacks peaked at around 500,000 PPS. Other
anecdotal sources report larger attacks consuming 35
megabits per second (Mbps) for periods of around 72
hours, with high-volume attacks reaching 800 Mbps.
Damage & Costs
• Hidden Costs: There may be hidden costs associated with denialof-service attacks. For example, the direct target of a DoS attack
may not be the only victim. An attack against one site may affect
network resources that serve multiple sites.
• Bandwidth wastage: Resources we share with other parties
(upstream bandwidth) may be consumed by an attack on someone
else—another customer of our Internet service provider is attacked,
so our upstream connections and routers are not as available to
handle our legitimate traffic. Thus, even when we are not the target
of an attack, we might experience increased network latency and
packet loss, or possibly a complete outage.
Damage & Costs (contd.)
• Logging costs: We may have additional costs because
of the need to size notification resources (such as logs,
mail spools, and paging services) to absorb attackrelated events. Logging systems need to cope with
significant deviations in the amount of data logged during
attacks.
• Extra network channels: Ideally, logging systems
should use an out-of-band channel so that logging traffic
does not add to the volume of DoS traffic that may be
passed to the internal network. Centralized logging
systems, considered a best security practice, may be
stressed by receiving log data from multiple locations.
Mail queues may fill up during a prolonged outage.
Damage & Costs (contd.)
• Insurance & Bandwidth cost: Network traffic
generated by the attack can result in incremental
bandwidth costs—when we pay per byte, we
also pay for the increased traffic caused by the
attack.
• In addition, our upstream Internet provider might
or might not be amenable to waiving penalty
charges caused by flood traffic.
• Other issues that create hidden costs are
insurance or legal fees or possible third-party
liability resulting from our involvement in an
attack.
How to handle DoS
•
Protecting – Among the aspects of protecting our systems and our
business, are looking at network design, discussing our agreement with
your ISP, putting detection mechanisms and a response plan in place, and
perhaps taking out an insurance policy. Proper preparation is essential for
effective detection and reaction. Unfortunately, some sites begin their cycle
with detection and reaction, triggering preparation steps after a “lessons
learned” experience.
•
Detecting – Our ability to detect attacks directly affects our ability to react
appropriately and to limit damages. Among the approaches we can take are
instituting procedures for analyzing logs and using automated intrusion
detection systems.
•
Reacting – Reaction steps, hopefully put in place as part of preparing for an
attack, include following our response plan, implementing specific steps
based on the type of attack, calling our ISP, enabling backup links, moving
content, and more. Technical steps include traffic limiting, blocking, and
filtering.
Real world targets and metrics
Following are few real world examples of various targets of
DOS attacks:
• A worm called MyDoom started propagating which had a
real target in mind - www.sco.com. It was engineered to
launch a Denial Of Service (DOS) attack against SCO
starting on February 1. Damage and total cost estimates
from MyDoom are still in progress, but CEI now
estimates the total may exceed $ 4 billion, making it one
of the most costly cyber attacks on record.
• In January 2001 a series of DoS attacks overwhelmed
the multicast infrastructure with an unusually large
number of Source Active messages. An Internet worm,
called the Ramen Worm, triggered these attacks with the
simple attack mechanism.
Summary
Dos Attack Types
Flood Attacks
Logic / Software
Attacks
Summary
Flood Attacks
TCP SYN
Flood Attack
Smurf IP
Attack
UDP Flood
Attack
ICMP Flood
Attack
Summary
TCP SYN
Flood Attack
Taking advantage of the flaw of TCP three-way handshaking behavior, an attacker makes
connection requests aimed at the victim server with packets with unreachable source
addresses. The server is not able to complete the connection requests and, as a result, the
victim wastes all of its network resources. A relatively small flood of bogus packets will tie
up memory, CPU, and applications, resulting in shutting down a server.
Graphic: http://www.narizone.it/sezioni/firewall/Immagini/SYN_sequence.gif
Summary
Smurf IP
Attack
An attacker sends forged ICMP echo
packets to broadcast addresses of
vulnerable networks. All the systems
on these networks reply to the victim
with ICMP echo replies. This rapidly
exhausts the bandwidth available to
the target, effectively denying its
services to legitimate users.
Summary
UDP Flood
Attack
UDP is a connectionless protocol and it does not require any connection setup procedure to
transfer data. A UDP Flood Attack is possible when an attacker sends a UDP packet to a
random port on the victim system. When the victim system receives a UDP packet, it will
determine what application is waiting on the destination port. When it realizes that there is no
application that is waiting on the port, it will generate an ICMP packet of destination
unreachable to the forged source address. If enough UDP packets are delivered to ports on
victim, the system will go down.
Summary
ICMP Flood
Attack
icmp
icmp
icmpicmp
icmpicmp
icmp icmp
icmp
An ICMP flood occurs when ICMP pings overload a system with so many echo
requests that the system expends all its resources responding until it can no
longer process valid network traffic.
Summary
Dos Attack Types
Flood Attacks
Logic / Software
Attacks
Summary
Logic / Software
Attacks
Ping of Death
Teardrop
Land
Echo/Chargen
Summary
Ping of Death
Expected Packet Size
Actual Packet Size
An attacker sends an ICMP ECHO request packet that is much larger than the
maximum IP packet size to victim. Since the received ICMP echo request packet
is bigger than the normal IP packet size, the victim cannot reassemble the
packets. The OS may be crashed or rebooted as a result.
Summary
Teardrop
An attacker sends two fragments that cannot be reassembled properly by manipulating the
offset value of packet and cause reboot or halt of victim system. Many other variants such as
targa, SYNdrop, Boink, Nestea Bonk, TearDrop2 and NewTear are available.
Summary
Land
198.215.34.56 198.215.34.56
An attacker sends a forged packet with the same source and destination
IP address. The victim system will be confused and crashed or rebooted
Summary
Echo/Chargen
• The character generator (CharGen) service is designed is primarily used for
testing purposes.
• Remote users/intruders can abuse this service by exhausting system resources.
• Spoofed network sessions that appear to come from that local system's echo
service can be pointed at the CharGen service to form a "loop."
• This session will cause huge amounts of data to be passed in an endless loop that
causes heavy load to the system.
• When this spoofed session is pointed at a remote system's echo service, this
denial of service attack will cause heavy network traffic/overhead that considerably
slows your network down.
Conclusion / Question
What makes DoS attacks possible?
What makes DoS attacks possible?
end-to-end paradigm
• middle is passive (packet forwarding)
• sender and receiver to all the work
What makes DoS attacks possible?
Internet security is highly interdependent.
• Keeping your machine secure may not be enough
• Your security relies too much on other machines on the
net
What makes DoS attacks possible?
Internet resources are limited.
• Each Internet entity (host, network, service) has
limited resources that can be consumed by too
many users.
What makes DoS attacks possible?
Intelligence and resources are not collocated
• Intelligence mostly in the hosts
• middle mostly worried about high throughput, not
decision making (like filtering)
What makes DoS attacks possible?
Accountability is not enforced.
• IP spoofing gives attackers a powerful mechanism to
escape accountability
What makes DoS attacks possible?
Control is distributed.
• Internet management is distributed
• each network is run according to local policies
• no way to enforce global deployment of a particular
security mechanism or security policy
• often impossible to investigate cross-network traffic
behavior
Dos Defense
How do we defend against a
Dos attack??
Dos Defense
Separate Client and Server Addresses
• The IP address space can be divided into a set of client
addresses and a set of server addresses.
• allow clients to initiate connections to servers, but not
vice versa
• nor servers to initiate connections to servers.
Dos Defense
Nonglobal Client Addresses
– path based addressing
Dos Defense
RPF Checking of Server Addresses
– Using path-based client addresses severely restricts
source-address spoofing by a client, but it does not
restrict spoofing by servers.
– Reverse Path Forwarding largely prevents a server
from spoofing the address of a server in a different
domain.
Dos Defense
Middlewalls
• simple special-purpose high-speed firewalls being
deployed in the core of the Internet at inter-domain
boundaries to serve as a filter of sorts
• Gives Upstream access control to a server under
stress
Conclusion
• Susceptibility to attacks could be alleviated with
better Internet Architectures (goal of class).
• Don’t leave all the decision making to the
machines on either end of a connection
• Provide ‘intelligent’ support along the path (e.g.
No Blind forwarding of packets)
• Create “Hardened” networks
References
•
•
•
•
•
•
•
•
•
•
•
•
•
•
http://www.google.co.in
http://www.itoc.usma.edu/workshop/2005/Papers/Follow%20ups/Making%20Gar
bage%20Collection%20Dependable.pdf
http://www.cert.org/archive/pdf/Managing_DoS.pdf
http://en.wikipedia.org/wiki/Smurf_attack
http://searchsecurity.techtarget.com/sDefinition/0,,sid14_gci822096,00.html
http://www.computereconomics.com/article.cfm?id=932
www.caida.org/tools/measurement/Mantra/mantra-publications/INFOCOM03b.ps.gz
http://www.iss.net/security_center/advice/Exploits/TCP/SYN_flood/default.htm
http://www.cert.org/tech_tips/denial_of_service.html
http://www.cs.ucl.ac.uk/staff/m.handley/papers/dos-arch.pdf
A taxonomy of DDoS attack and DDoS defense mechanisms
http://cactus.eas.asu.edu/Partha/Papers-PDF/2003/Shu-ICCCN.pdf
advanced.comms.agilent.com
http://www.anml.iu.edu/ddos/types.html#flood