ATTACKING WPA3: NEW VULNERABILITIES AND EXPLOIT FRAMEWORK presented at HITB Singapore 2022

by Mathy Vanhoef,

Summary : In this presentation, we perform an audit of WPA3’s new features. We focus on Management Frame Protection, which prevents the popular deauthentication attack, and we study the new Simultaneous Authentication of Equals handshake. This uncovered several 0-day vulnerabilities, ranging from attacks that allow an adversary to trivially disconnect users from the network, to remotely crashing an access point, and revealed a vulnerability that allows an adversary to intercept all traffic of a victim under the right circumstances. Proof-of-concepts are implemented using a new Wi-Fi testing framework, allowing researchers to easily test their devices for several of the identified vulnerabilities.

The layout of the presentation is as follows:

The new Simultaneous Authentication of Equals handshake, also called Dragonfly, which prevents dictionary attacks and provides forward secrecy.
The mandatory usage of Management Frame Protection which most notably prevents ‘deauthentication attacks’ where a client is forcibly disconnected from the network.
The introduction of the SAE Public Key protocol to better secure “coffee shop” hotspots where authenticated is based on a shared password.
I’ll also briefly mention known weaknesses in WPA3, namely the Dragonblood attacks, and how they relate to the work being presented. Summarized, the Dragonblood attacks covered side-channel leaks in the SAE (Dragonfly) handshake that enabled an adversary to still perform dictionary attacks.
After this introduction, the presentation will focus on the security of Management Frame Protection (MFP). This feature most notably prevents certain denial-of-service attacks. For instance, it prevents the famous deauthentication attack. However, we discovered various novel DoS attacks that remain possible even when WPA3 and MFP are being used. For instance, we will explain the following newly discovered attacks:

An adversary can trivially disconnect Linux, macOS, iOS, iPadOS, and Windows clients, by spoofing beacon frames that contain a channel switch announcement. Beacon frames are not protected, even when using MFP, meaning an adversary can trivially perform this attack by spoofing beacon frames.
Against Linux and Windows, we also demonstrate that spoofing beacon frames that announce an unsupported channel bandwidth causes the client to disconnect from the network, even when MFP is being used.
The presentation will also briefly mention design flaws in MFP that still enable an adversary to perform attacks when the victim is in the process of connecting to the network.
Last but not least, we will demonstrate how initiating a large number of SAE handshakes causes the D-Link DIR-X1860 router to crash. We conjecture that this vulnerability affects many more vendors and devices.
We will mention that these attacks were implemented in our new Wi-Fi testing framework. This enables others to easily test their own devices against these attacks. This framework also allows other researchers to easily implement other Wi-Fi attacks.
We will also cover attacks agains the novel SAE-PK protocol. In particular, two network-based attacks will be discussed:

An adversary can use techniques such as ARP poisoning to still intercept the traffic of all clients in the network. This can be prevented by disabling client-to-client communication.
Second, because the group key used to encrypt broadcast and multicast frames is shared by all clients in the network, this key can be used by any client to spoof broadcast and multicast traffic. Moreover, we will demonstrate that an adversary can even use the group key to spoof unicast traffic against several victims.
Finally, the presentation will briefly mention the possibility of constructing rainbow tables to still break the SAE-PK password. In other words, rainbow tables can be used to convert the SAE-PK password into a known public and private key, which then enables the adversary to create a rogue clone of the network (but this is a theoretic attack and hence only briefly mentioned in passing – it’s still good for the audience to know that it exists).