Static detection of energy defect patterns in Android applications

Abstract

For static analysis researchers, Android software presents a wide variety of interesting challenges. The target of our work is static detection of energy-drain defects in Android applications. The management of energy-intensive resources �e.g., GPS) creates various opportunities for software defects.Our goal is to detect statically "missing deactivation" energydrain defects in the user interface of the application. First, we define precisely two patterns of run-time energy-drain behaviors, based on modeling of Android GUI control-flow paths and energy-related listener leaks along such paths. Next, we define a static detection algorithm targeting these patterns. The analysis considers valid interprocedural control-flow paths in a callback method and its transitive callees, in order to detect operations that add or remove listeners. Sequences of callbacks are then analyzed for possible listener leaks. Our evaluation considers the detection of GUI-related energy-drain defects reported in prior work, as well as new defects not discovered by prior approaches. In summary, the detection is very effective and precise, suggesting that the proposed analysis is suitable for practical use in static checking tools for Android.

Detecting Energy Bugs in Android Apps Using Static Analysis

“…Wu et al [13] focus on detecting energy-related defects in the UI logic of an Android application. The authors identify two patterns of behavior in which location listeners are leaking.…”

On the unsoundness of static analysis for Android GUIs

“…In recent work we introduced the window transition graph [35] as another control-flow representation for Android GUI control flow. This model has been used to perform static energy-drain defect detection [33], test generation for leaks [37], and responsiveness profiling [32]. The control-flow analysis is part of the publicly available GATOR toolkit for static analysis for Android.…”

“…Many are motivated by security-related problems; a representative example is the taint analysis in FlowDroid [3], which will be discussed later. Others consider problems such as battery drain (e.g., [25,33]), leaks (e.g., [14,37]), GUI exploration (e.g., [4,36]), and responsiveness (e.g., [19,32]). In our experience, existing work (including our own work) almost never discusses the soundness of the proposed analysis.…”

“…Aspects of this control/data flow are almost always a concern for existing static analyses for Android. For example, mismanagement of GUI control flow (e.g., the callbacks for an activity's lifecycle, which are discussed shortly) is one of the major sources of battery-drain defects [20,25,33].…”

Amulet

“…Developer-focused tools have also emerged for smartphones. Tools have focused on identifying energy bugs at compile time [44], in-situ energy metering using kernel additions for energy model building [45], and then using those models to predict smartphone lifetime [22]. These tools, if carefully applied to constrained wearable platforms, could complement the developer insights gleaned from the Amulet Resource Profiler and ARP-view.…”