2. Contents
⢠Working idea
⢠Paper introduction
⢠Techniques & Evaluation
⢠Conclusion of paper
⢠What I have to do
3. Paper introduction
⢠PowerSpy, Dan Boneh
⢠USENIX Security, 2015 / Acceptance ratio : 15.7%(67/426)
⢠Learn userâs location from reading phoneâs power consumption
4. Contributions
⢠Showed power meter available on medern phones can reveal
potentially private information
⢠Developed the machine learning techniques for infer location
information from power meter data.
⢠Discussed potential continuation, countermeasures for this work.
5. Assumptions
⢠Malicious application is installed and running in background
⢠Only has permission to access power data, network communicationâ¨
- Cannot measure the power consumed by the cellular radio alone.
⢠Prior knowledge of the area/routes through which the victim moves.
⢠Cannot locate a phone that is standing still.
16. ⢠And we can get battery info without any permissions
Motivation
17. Challenges
⢠Pre-measurement may have different speed or stops.
⢠Have to identify the targetâs power proďŹle in many pre-collected
proďŹles along different routes
⢠Exact location of the target may be ambiguous
⢠Target may travel the road which the attacker partially measured.
34. Evaluation
⢠Nexus 4, Nexus 5, HTC
⢠Enough communication is occuring
⢠Driver might be using nav sw or streaming music
⢠Unexpected events (e.g., phone call) can be normalized using MA