Smart-Guard: Defending User Input from Malware

Trusted input techniques can profoundly enhance a variety of scenarios like online banking, electronic voting, Virtual Private Networks, and even commands to a server or Industrial Control System. To protect the system from malware of the sender's computer, input needs to be reliably authenticated. Previous research in this field is based on fixed assumptions about trustworthy components and is, thus, too rigid for this use case. We present Smart-Guard, a method to protect user input into a system even if the attacker controls - to us unknown - parts of the underlying system. Our approach ensures integrity of user input even when up to two of three devices are compromised; confidentiality holds for one malicious device. In this way, Smart-Guard has flexible trust assumptions, and does not require any particular part of the system to be trusted. To prove our claims, we formally verified our protocol using the state-of-the-art protocol verifier ProVerif. Additionally, we define a new class of techniques, malware tolerance, which operate securely even when the system is infected with malware.