We've made denial-of-service attacks into a separate section for several reasons. First, denial-of-service attacks are extremely difficult to protect against, and especially so with wireless. Any attacker with a bigger amplifier, antenna, or using more power, can deny service to an individual or group at the RF level. As a result, RF attacks are difficult but not impossible to prevent. The military, for instance, uses spread spectrum, frequency hopping, and probably ultrawide-band systems to mitigate the possibility of an attacker jamming important frequencies. We don't have that luxury because our equipment is readily available to our attackers. Therefore, RF-based denial-of-service attacks against IEEE 802.11-based networks (and actually any consumer wireless standard) are nearly impossible to prevent. An attacker with the know-how and access to the right equipment can mount a denial-of-service attack against your wireless network.
Another class of denial-of-service attack against the network and cryptographic protocols, specifically layer 2 or the MAC layer, is preventable. Unfortunately, neither the old nor the new Wi-Fi standards opted to protect against this form of attack?TGi debated the cost of protecting against layer 2 denial-of-service attacks, but opted for downward compatibility with the old standard rather than protection against denial-of-service attacks.
You may have noticed that the management frames, for example, Associate-request, don't have any integrity protection. That is, these frames can easily be forged by an attacker. An attacker can deny service to a station/client or to an entire access point, and in some cases across a LAN.
The attack is trivial with the right software (see Chapter 16). If the attacker wishes to prevent a station from using the Wi-Fi LAN, he has several choices. First, when the attacker can see the AP to which the station is associated, he simply forges a Disassociation or Deauthentication frame and sends it to either the AP or STA. The AP/STA, thinking that the station wishes to leave (or the AP no longer can service the STA), grants the request and closes the association. Unfortunately, both of these management frames (Disassociation and Deauthentication) permit the attacker to use the broadcast MAC address as the target. This results in all stations associated with the targeted AP being knocked off the AP.
Second, the attacker can deny service to a station when he can see an AP on the same wired LAN as the AP to which the target station is associated. In this case, the attacker sends a forged Association-request message with the target station's MAC address to an AP on the same wired LAN. The AP that receives the association request approves (because we don't authenticate until after association) it and sends out a layer 2 update frame to the wired LAN. The router or switch now begins forwarding traffic to the AP that just sent the layer 2 update, and the actual station no longer receives any traffic. Obviously, both of these attacks need to be constantly run to prevent service to a particular station, and this is one of the reasons why TGi opted not to protect against it.
Another method of denying service to a group is similar and involves loading up an AP with bogus stations such that the resources on the AP are exhausted and the AP either reboots or no longer permits new stations to associate to it.
TGi decided not to protect against these attacks because the majority of the participants felt that a determined attacker can always resort to an RF-based attack. The majority of the members of TGi also were concerned with potential problems with backward compatibility if integrity protection were added to management frames. Unfortunately, these attacks have been implemented in an open source tool (see Chapter 16).
Michael is a lightweight message authenticity algorithm (see Chapter 11). Because Michael provides only 20 bits of protection against message modification attacks, countermeasures were designed to prevent active attacks. These countermeasures are effective at preventing the creation of a forged message. However, they also introduce the potential for a denial-of-service attack against the entire AP.
A capable attacker can accomplish the WPA denial-of-service attack. The attack isn't trivial to accomplish, but it doesn't require rocket science either. Essentially, the attacker must accomplish three tasks. The first is to stop a valid packet from reaching the AP, and then, second, to modify the packet such that the ICV remains valid. The third and final task is to send the modified packet before a packet with a higher TSC is received by the AP.
Accomplishing all three tasks is easy once a man-in-the-middle attack is established between the AP and an STA. Because the attacker controls the connection between to the STA and AP, he can easily perform the first and third tasks. Performing the second task requires applying what we discussed earlier in this chapter and is shown in Appendix A to modify the message.
Once the attacker sends two such modified packets to the AP within a minute, the AP shuts down for exactly one minute, preventing traffic from all stations associated with the AP from communicating. The AP must also rekey stations immediately upon beginning service after the one-minute delay.
While this is a particularly brutal DoS attack, the same results are obtained by using a much easier attack?the management frame DoS described earlier.