The Last Resort: Wireless DoS Attacks

Multiple DoS attacks against various wireless (and even wired) protocols, security protocols included, are mentioned elsewhere in the chapter. In many cases these attacks can be part of a sophisticated penetration plan and assist in social engineering, man-in-the-middle attempts, stealing, or cracking secret keys. However, a desperate attacker might launch a DoS attack to "compensate" for the effort spent on failed access attempts. Besides, wireless DoS attacks per se can be launched by the competitors, for political reasons, out of curiosity, and so forth; the situation is no different from DoS attacks on public networks such as the Internet. Unfortunately, due to the nature of the RF medium and design of the core 802.11 protocols, wireless networks cannot be protected against Layer 1 and certain Layer 2 DoS attacks. This is why, in our opinion, 802.11 links should not be used for mission-critical applications in theory. In the real world, there are cases when 802.11 is the only choice, and cases of system administrators or network designers being unaware or dismissive of the problem and going forward with the WLAN installation anyway. This is why you, as a security professional, should be able to demonstrate various wireless DoS dangers to your clients. If you are a system administrator or a wireless enthusiast, you can always check out how wireless DoS attacks work on your network, perhaps to know what to expect when your WLAN is attacked and to generate IDS signatures. For your convenience, we have categorized known wireless DoS attacks:

1 Physical Layer Attacks or Jamming

There is nothing you can do about RF jamming short of triangulating the jamming device and tracking its owner. Even then the jammer owner is likely to claim that he or she did nothing illegal, because anyone is allowed to transmit anything in the ISM band. You will have to prove that the attacker's transmission is intentional and that he or she has exceeded the FCC EIRP limit (most likely this is the case) in a court of law. The jamming device can be a custom-built transmitter or a high-output wireless client card or even access point (e.g., Demarctech offers an AP with 500-mW output!) flooding the selected channel(s) with junk traffic. FakeAP, Void11, File2air, or any other 802.11 frame-generating tool can be used to run the flood. A completely custom-built jammer can employ harmonics and transmit at about 1.2 GHz or even about 600 MHz. Such a device would be easier to build than the 2.4 to 2.5 GHz jammer, and you'll need a decent, expensive frequency counter to discover the attack and its source. If one wants to build a very powerful 2.4 to 2.5 GHz jamming device, the core for such a device is elsewhere; it's called a microwave oven's magnetron. Check out Vjacheslav (Slava) Persion's Web page (http://www.voltagelabs.com/pages/projects/herf005/) for examples of microwave magnetron-based transmitters in action. The main disadvantages of Layer 1 attacks from the attacker's perspective are time, effort, and expenses to build a jammer, and the fact that such a device would have to be positioned quite close to the attacked network for an efficient attack. It is very likely that once the attack is discovered, the jammer is lost and can serve as hard evidence in court.

2 Spoofed Deassociation and Deauthentication Frames Floods

These attacks are probably the most well-known and used DoS attacks on 802.11 LANs. In the beginning of this chapter we discussed deauthentication frames floods when applied to bypassing MAC address filtering and closed ESSIDs.

Just as in the case of jamming, there is little you can do to eliminate the threat. The 802.11i developers have discussed the possibility of authenticated deauthentication (pardon the tautology) and deassociation. However, as far as we know, the idea did not get any further in practical terms. A variety of tools can be used to launch deauthentication and deassociation floods, including dinject, wlan_jack, File2air, Void11, and omerta. Void11 is probably the most devastating tool mentioned because it provides "canned" mass flood and match list flood capabilities:

arhontus# void11_hopper >/dev/null &

arhontus# void11_penetration -D wlan0 -S ihatethisnetwork -m 30

or

arhontus# void11_hopper >/dev/null &

arhontus# echo DE:AD:BE:EF:13:37 > matchlist

arhontus# void11_penetration -l matchlist -D wlan0

The capability to attack hosts from a matchlist can be very useful when implementing active defenses on your WLAN.

An extension of the deauthentication or deassociation frames flood attack is sequential multiframe attacks, such as sending deauthentication or deassociation frames followed by a forged probe responses and beacon frames flood providing incorrect information (ESSID, channel) about an access point to associate with. If 802.1x is used on the network, an EAP-Failure frame can preclude the deauthenticate or deassociate + fake probe responses frames train. Such an attack guarantees that the targeted host is dropped from the WLAN like a lead weight and will have difficulties reassociating. A forged probe responses flood might or might not have a significant detrimental effect on reassociation, depending on the passive versus active scanning priority implemented by the attacked host wireless card firmware. An example of deauthenticate + fake probe response frame attack is given in the file2air README file; this or other (void11 + FakeAP?) tools can be used to launch this type of attack.

3 Spoofed Malformed Authentication Frame Attack

This attack is implemented in practice by the fata_jack utility written for AirJack by "loud-fat-bloke" (Mark Osborne; http://www.loud-fat-bloke.co.uk). It is based on the wlan_jack code, but sends altered spoofed authentication request frames instead. As the author of the tool states, the sent frame has a destination address of the AP and a source address of the attacked client and is an authentication frame with an unknown algorithm (type 2) and a sequence number and status code both set to 0xffff.

As a result of an attack, the AP sends the impersonated client a reply frame. This frame says "Received an authentication frame with authentication sequence transaction sequence number out of expected sequence" (i.e., code 0x000e). This causes the client to become unauthenticated from the AP. In our experience, the client becomes deassociated and starts behaving erratically, exhibiting difficulties reassociating and sudden channel hops.

4 Filling Up the Access Point Association and Authentication Buffers

Many access points do not implement any protection against these buffers being overflowed and will crash after an excessive amount of connections are established or authentication requests sent. This applies to software access points as well; for example, an OpenBSD 3.1-based AP. Void11 implements both association and authentication frames floods with random flooding host interface MAC addresses. A small progtest utility that comes as an example code with libwlan for Linux HostAP also associates a great number of fake stations with an access point to see if it will crash or freeze. Alternatively, you can associate to the AP and then start fast MAC address changes at the associated interface. This variation of the association buffer overflow attack is implemented by a macfld.pl script by Joshua Wright:

arhontus# perl macfld.pl

macfld: Need to specify number of MAC's to generate with -c|--count

Usage:

  macfld [options]

    -c, --count

    -u, --usleep (microseconds)

    -f, --dataflush

    -p, --pingtest

    -i, --interface WLANINT

    -a, --apaddr

    -s, --srcaddr

    -d, --debug

    -h, --help

We strongly believe that the access point and wireless bridge manufacturers should implement these and similar tools to test their equipment before the production cycle begins.

5 Frame Deletion Attack

The idea behind this attack is to corrupt the bypassing frame's CRC-32 so that the receiving host will drop it. At the same time, the attacker sends a spoofed ACK frame to the sender telling it that the frame was successfully received. As a result, the corrupt frame is efficiently deleted without being resent. Because authenticating all CSMA/CA frames is not resource-feasible, there is nothing that can be done to stop frame deletion attacks. To corrupt the CRC, the attacker might try to send the same frame with the corrupt CRC at the same time with the legitimate sender or emit a lot of noise when the sender transmits the last 4 bytes of the frame. Providing a reliable frame CRC corruption is probably the trickiest part of the attack. Of course, if implemented successfully, such an attack is not easy to detect or defend against. However, at the time of writing, it is purely theoretical and we have yet to see someone making the theoretical practical.

6 DoS Attacks Based on Specific Wireless Network Settings

There are somewhat obscure attack possibilities based on exploiting specific Layer 2 settings of 802.11 LANs, such as the power-saving mode or virtual carrier sense (RTS/CTS)-enabled networks.

In power-saving mode attacks, a cracker can pretend to be the sleeping client and poll the frames accumulated for its target from the access point. After the frames are retrieved, the access point discards the buffer contents. Thus, the legitimate client never receives them. Alternatively, our cracker can spoof traffic indication map (TIM) frames from the access point. These frames tell the sleeping clients whether the data has arrived for them to wake up and poll it. If a cracker can deceive the clients to believe that no pending data was received by the AP, they remain asleep. In the meanwhile, the access point accumulates the unpolled packets and is forced to discard them at some point or suffer a buffer overflow. This attack is more difficult to accomplish, because the cracker has to find the way to stop the valid TIM frames from reaching the intended hosts. Finally, a cracker can spoof beacons with TIM field set or ATIM frames on ad-hoc WLANs to keep the hosts awake even if there is no data to poll. This would efficiently cancel the power-saving mode operation and increase the client host's battery drain.

The DoS attacks against the virtual carrier sense-implementing networks are prioritization attacks by nature. A cracker can constantly flood the network with request to send (RTS) frames with a large transmission duration field set, thus reserving the medium for his or her traffic and denying other hosts from accessing the communication channel. The network is going to be overwhelmed by the clear to send (CTS) responses to every RTS frame received. The hosts on the WLAN will have to obey these CTS frames and cease transmitting.

Although there are no specific tools available to launch these attacks, in practice, File2air, a hex editor, and some additional shell scripting come to mind.

7 Attacks Against 802.11i Implementations

Nothing is without a flaw, and new security standards can introduce new potential security flaws even as they fix the old ones. The risk/benefit ratio is what matters in the end, and in the case of the 802.11i security standard the balance is positive: It is better to have it than not. Nevertheless, there are a few problems with 802.11i implementations that can be exploited to launch rather sneaky DoS attacks. In this chapter we have already reviewed DoS attacks against 802.1x/EAP authentication protocols that might force an unsuspecting network administrator to switch to other, less secure means of user authentication, if persistent. Another avenue for possible DoS attacks against 802.11i-protected networks is corrupting the TKIP Michael message integrity checksum. In accordance with the standard, if more than one corrupt MIC frame is detected in a second, the receiver shuts the connection down for a minute and generates a new session key. Thus, a cracker corrupting the frame MICs a few times every 59 seconds should be able to keep the link down. However, launching this attack is not as easy as it seems. Because understanding all the "whys" and "why nots" of the MIC corruption attack requires an understanding of MIC (and TKIP in general) operations, a detailed discussion of this attack belongs in Chapter 12, where you can find it. Here we state that running this attack by sending different MIC frames with the same IV does not appear to be easy to implement or even possible. An attacker would have to resort to means similar to the CRC-32 corruption in the frame deletion attack described earlier; for example, emit a jamming signal when the part of the frame containing the MIC is transmitted. For now, like the frame deletion attack, the corrupt MIC attack remains purely theoretical.

To conclude this chapter, even the latest wireless safeguards aren't 100 percent safe. In the following discussion, you are invited to observe (or participate in) the security horrors that can follow a successful attack on a WLAN.