A plethora of organizations, companies, and foremost universities and educational institutions are using WPA2-Enterprise protocol to allow their end-users to connect to provided Wi-Fi networks. When both the provider’s and the end-user’s devices are configured properly, it is considered one of the safest Wi-Fi connection protocols with the added benefits of having a unique password for every Wi-Fi user. However, a known evil twin attack can be performed to steal users’ Wi-Fi login credentials, if the devices are not configured correctly. Considering the widespread use of Wi-Fi-enabled smartphones and rising concerns regarding users’ privacy, we focus on the privacy aspects of WPA2-Enterprise vulnerabilities mainly on the widespread Eduroam network. We show that device deanonymization is a concerning liability of many Eduroam networks. More than 87% of 1650 devices collected during a two-month test on our university are vulnerable to MAC address deanonymization attack. Furthermore, by analyzing the Eduroam Configuration Assistant Tool of 1066 different institutions around the world, 67% of exported Eduroam profiles having the Wi-Fi device reveal the user’s identity in the clear, thus linking the users with the device’s MAC address. Indeed, the analysis of the configuration profiles has been confirmed by performing the deanonymization attack on a large-scale international music festival in our country, where 70% of the devices have been vulnerable. Additionally, we showcase the psychological aspects of secure Eduroam users, where some are willing to modify secure configuration profiles to gain aspects to certain blocked features. As a result, the attacker is granted with user credentials and IMSI number and provided with access to all Eduroam-related services.

Classical password/PIN-based authentication methods have proven to be vulnerable to a broad range of observation attacks (such as key-logging, video-recording or shoulder surfing attacks). In order to mitigate these attacks, a number of solutions have been proposed, most of them being cognitive authentication schemes (challenge-response protocols that require users to perform some kind of cognitive operations). In this paper, we show successful passive side-channel timing attacks on two cognitive authentication schemes, a well-known Hopper-Blum (HB) protocol and a U.S. patent Mod10 method, previously believed to be secure against observation attacks. As we show, the main security weakness of these methods comes from detectable variations in the user's cognitive load that results from cognitive operations during the authentication procedure. We carried out theoretical analysis of both Mod10 and HB methods, as well as an experimental user study of Mod10 method with 58 participants to validate the results of our timing attacks. We also propose security enhancements of these schemes aimed to mitigate the timing side-channel attacks. The proposed enhancements show the existence of a strong tradeoff between security and usability, indicating that the security of cognitive authentication schemes comes at a non-negligible usability cost (e.g., increased overall login time). For this reason, the designers of new cognitive authentication schemes should not ignore possible threats induced by side-channel timing attacks.