Exploiting geometrical structure for forensic applications of timing inference channels

Timing inference channels are a well-studied area of computer security and privacy research, but they have not been widely applied in digital forensic applications. Timing signatures (for example, of movies) are not robust against variations in the machine, the encoder, the environment, and other factors that affect timing, and unfortunately such issues have limited many researchers from using timing inference channels for revealing hidden data, detecting machine behavior, or even forensic analysis. The authors develop a geometrical interpretation in a high dimensional space of timing signatures for movies as an example of pattern-like software. The results suggest that timing signatures can be made robust against different machines, different encoders, and other environmental conditions by exploiting geometrical structure in this space. This geometrical structure helps identify the behavior of running pattern-like software that is useful for identifying digital crimes, privacy invasion matters, and network behaviors. This paper is focused on a thought experiment: how much information can an unprivileged process learn by just running on a system and observing its own timing? Although installing administrative software is the most frequent approach for understanding system behavior and detecting running software, the results show that it is feasible that such goals could be still achieved without any administrative privileges.