•   When: Thursday, February 06, 2020 from 02:00 PM to 03:00 PM
•   Speakers: Matthew Lentz
•   Location: Engineering Building 4201
•   Export to iCal

ABSTRACT

Contemporary devices are continuously sensing and always connected, enabling powerful services for end users. These services often operate over personal data: collecting and processing input from sensors (e.g., audio, location), or rendering output to the user (e.g., financial information). This not only manifests in problems for users in terms of security and privacy but also power management due to energy constraints on portable devices.

Understandably, this data is the target of many attacks, ranging from malicious applications to compromises of platform software. Users are increasingly required to rely on these devices but have few ways to control and reason about how their sensitive data is protected, processed, and shared. In this talk, I will discuss my work that leverages hardware security extensions to construct a software enforcement layer that helps users regain control over their data. I will introduce my work on SeCloak, a minimal enforcement layer to provide users with simple, yet powerful on/off control for hardware devices (such as their microphone or camera), and then discuss my work on Accountable Paths, which enables more expressive policies and supports a wider range of assurances over the data.

In order to support continuously sensing applications, energy-constrained devices rely on aggressive power management techniques such as entering a "sleep" state whenever possible, wherein tasks are frozen and hardware resources are placed in low-power modes. However, for application-driven wakeup events (e.g., sampling sensors), transitioning between states often consumes more energy than handling the event itself. I will present my work on a new power management state "Drowsy" which tracks dependencies at the OS-level to determine exactly which tasks and hardware need to be woken up to handle an event, improving the energy efficiency of handling these events by up to 5x. I will conclude with a discussion of my future research directions.

BIO

Matthew Lentz is a Ph.D. candidate in Computer Science at the University of Maryland. He received his B.S. in Computer Engineering from the University of Maryland in 2010. He is broadly interested in research at the intersection of systems, networking, and security.