Device Attacker Models: Fact and Fiction

According to standard fiction, a user is able to securely keep long term keys on his device. However, in fact his device may become infected with malware, and an adversary may obtain a copy of his key. We propose an attacker model in which devices are "periodically trustworthy" - they may become infected by malware, and then later become trustworthy again after software patches and malware scans have been applied, in an ongoing cycle. This paper proposes a solution to make the usage of private keys by attackers detectable by using public transparently-maintained logs to monitor the usage of long-term secret keys.