An essential forensic capability is to infer the sequence of actions performed by a suspect in the commission of a crime. Unfortunately, for cyber investigations, user activity timeline reconstruction remains an open research challenge, currently requiring manual identification of datable artifacts/logs and heuristic-based temporal inference. In this paper, we propose a memory forensics capability to address this challenge. We present Timeliner, a forensics technique capable of automatically inferring the timeline of user actions on an Android device across all apps, from a single memory image acquired from the device. Timeliner is inspired by the observation that Android app Activity launches leave behind key self-identifying data structures. More importantly, this collection of data structures can be temporally ordered, owing to the predictable manner in which they were allocated and distributed in memory. Based on these observations, Timeliner is designed to (1) identify and recover these residual data structures, (2) infer the user-induced transitions between their corresponding Activities, and (3) reconstruct the devicewide, cross-app Activity timeline. Timeliner is designed to leverage the memory image of Android's centralized ActivityManager service. Hence, it is able to sequence Activity launches across all apps -even those which have terminated. Our evaluation shows that Timeliner can reveal substantial evidence (up to an hour) across a variety of apps on different Android platforms.