MOBILESoft '20: Proceedings of the IEEE/ACM 7th International Conference on Mobile Software Engineering and Systems

Full Citation in the ACM Digital Library

SESSION: Software quality

Security testing of second order permission re-delegation vulnerabilities in Android apps

In Android, inter-app communication is a cornerstone feature where apps exchange special messages called Intents in order to integrate with each other and deliver a rich end-user experience. In particular, in case an app is granted special permission, it can dispatch privileged Intents to request sensitive tasks to system components.

However, a malicious app might hijack a defective privileged app and exploit it as a proxy, to forward attacking Intents to system components. We call this threat "Second Order Permission Re-delegation" vulnerability.

In this paper, we present (i) a detailed description of this novel vulnerability and (ii) our approach based on static analysis and automated test cases generation to detect (and document) instances of this vulnerability. We empirically evaluated our approach on a large set of top Google Play apps. Results suggest that this novel vulnerability is neglected by state of the art, but that it is common even among popular apps. In fact, our approach found 27 real vulnerabilities with fast analysis time, while a state-of-the-art static analysis tool could find none of them.

DFarm: massive-scaling dynamic Android app analysis on real hardware

Dynamic analysis is an important tool for assessing software quality during testing. It not only helps analysts identify performance bottlenecks and functional errors, but also provides a means for finding security vulnerabilities. For example, analysts can determine the servers to which a mobile app connects, which sensitive data it transfers, and which cryptographic protocols it uses for the transfer. While many approaches for monitoring a running Android app exist, most work silently assumes that a suitable execution environment is available. When analyzing hundreds of apps at the same time, however, a single phone on the analyst's desk is not enough. Emulators are not always an alternative as we show, because apps can behave differently on real hardware.

In this paper, we discuss the challenges for providing a large-scale testing environment with real Android devices on physical hardware. We further present DFarm, a software and hardware system to configure and control hundreds of Android phones in a private testing cloud. We discuss electrical wiring, USB and WiFi connectivity, automatic configuration, and load balancing. We evaluate DFarm on a range between 1 and more than 70 devices. We show that it provides near-linear scaling for dynamic app analysis when adding new devices, while retaining the original device's computation and network performance.

Making Android apps monkey-friendly

Monkey testing is a random testing technique in which a stream of pseudo-random events are automatically fired on the GUI of the application under test, usually with the purpose of robustness testing or responsiveness analysis. A line of research is dedicated to addressing the limitations of monkey testing for Android apps. However, all the existing works try to improve the underlying algorithms or techniques used by the monkey testing tools. In this vision paper, we propose the idea of improving the effectiveness of monkey testing by automatically refactoring the application under test. We provide two sample scenarios in which this idea can be used to address limitations of monkey testing for Android applications.

Improving app quality despite flawed mobile analytics

Analytics can help improve the quality of software; the improvements are affected by the fidelity of the analytics. The impact of poor fidelity may vary depending on the type of data being collected, for example, for crashes low fidelity may be sufficient.

The mobile ecosystem includes a platform where apps run and an app store that intermediates between developers and users. Google's Android ecosystem provides all the developers with analytics about various qualities of their app through a service called Android Vitals that automatically collects data on how their app is performing.

My research found ways to improve app quality through using mobile analytics, including Android Vitals. It also found fidelity flaws in several analytics tools provided by Google. They confirmed and validated some flaws and chose not to discuss others.

SESSION: Mining and reviews

AndroidPropTracker: mining lifetime properties of Android projects

Android operating system introduces new releases frequently. This fact led to the existence of several Android Application Programming Interfaces (APIs) which is one of the causes of the Android fragmentation phenomenon. As a consequence of fragmentation, many apps became not ready for new Android releases. Aiming at investigation of the readiness of Android apps, we developed a software repository mining tool to understand how ready apps are (and were) for Android releases. The tool tracks the changes of Android projects properties over time, contributing for a deeper analysis through collecting data since the beginning of the projects. It allows researchers to examine when exactly Android properties were changed, how many times they were changed, as well as all their values along time. This mechanism can support researchers to understand the evolution of Android projects and to answer research questions. In addition, developers can use the tool to track their apps evolution and perform comparisons and analysis with other open source apps. The tool can help developers to have a broader view of their apps evolution as well as to analyze competitor apps evolution.

ReviewViz: assisting developers perform empirical study on energy consumption related reviews for mobile applications

Improving energy efficiency of mobile applications is a topic that has gained a lot of attention recently. It has been addressed in a number of ways such as identifying energy bugs and developing a catalog of energy patterns. Previous work shows that users discuss the battery related issues (energy inefficiency or energy consumption) of the apps in their reviews. However, there is no work that addresses the automatic extraction of the battery related issues from users' feedback.

In this paper, we report on a visualization tool that is developed to empirically study machine learning algorithms and text features to automatically identify the energy consumption specific reviews with the highest accuracy. Other than the common machine learning algorithms, we utilize deep learning models with different word embeddings to compare the results. Furthermore, to help the developers extract the main topics that are discussed in the reviews, two state of the art topic modeling algorithms are applied. The visualizations of the topics represent the keywords that are extracted for each topic along with a comparison with the results of string matching.

The developed web-browser based interactive visualization tool is a novel framework developed with the intention of giving the app developers insights about running time and accuracy of machine learning and deep learning models as well as extracted topics. The tool makes it easier for the developers to traverse through the extensive result set generated by the text classification and topic modeling algorithms. The dynamic-data structure used for the tool stores the baseline-results of the discussed approaches and are updated when applied on new datasets. The tool is open sourced to replicate the research results.1

Embracing mobile app evolution via continuous ecosystem mining and characterization

While an indicator of its vibrancy, the rapid evolution of a mobile ecosystem also causes challenges to mobile software engineers in developing and maintaining quality products, and to users concerning the usability and security of resulting apps. In this context, it is crucially important to arm mobile software engineers with effective and practical tool support that is informed and enabled by a comprehensive understanding of the evolutionary dynamics of this ecosystem. Targeting Android, we envision to build an infrastructure that is capable of systematically and continuously mining a mobile software ecosystem. Using this infrastructure, we then perform large-scale longitudinal characterization studies of the ecosystem to understand its evolutionary dynamics with a focus on the behavioral evolution patterns of, and ecological interaction among, three ecosystem elements: the mobile platforms, user apps built on the platforms, and users associated with the apps (including end users and developers). Further, the characterization results enable proactive app quality and sustainable app security. We also report our current progress in this effort with initial results, and discuss risks and next steps.

Collaborative earthquake detection and response using smart devices

An earthquake detection and response system is crucial due to the fatal nature of earthquakes. Recently, extensive research efforts have been made to detect earthquakes in real time and give an alert by incorporating with server-based systems or by working as a stand-alone device. However, such approaches still have many challenges due to the cost of network and server infrastructures or the inaccuracy of detection algorithms. In this work, we introduce a collaborative approach using smart devices located in a near area to detect earthquakes and issue an alert with detailed action plans. A smartphone's accelerometer works as a seismic sensor to monitor ground motions when it is in a steady state. Once an earthquakelike motion is detected, it is sent to nearby devices through D2D communications to confirm it as an earthquake. Our experimental results show that the proposed approach can be effective in a home environment or an earthquake early warning system is not present.

SESSION: Empirical software engineering

Leave my apps alone!: a study on how Android developers access installed apps on user's device

To enable app interoperability, the Android platform exposes installed application methods (IAMs), i.e., APIs that allow developers to query for the list of apps installed on a user's device. It is known that information collected through IAMs can be used to precisely deduce end-users interests and personal traits, thus raising privacy concerns. In this paper, we present a large-scale empirical study investigating the presence of IAMs in Android apps and their usage by Android developers.

Our results highlight that: (i) IAMs are widely used in commercial applications while their popularity is limited in open-source ones; (ii) IAM calls are mostly performed in included libraries code; (iii) more than one-third of libraries that employ IAMs are advertisement libraries; (iv) a small number of popular advertisement libraries account for over 33% of all usages of IAMs by bundled libraries; (v) developers are not always aware that their apps include IAMs calls.

Based on the collected data, we confirm the need to (i) revise the way IAMs are currently managed by the Android platform, introducing either an ad-hoc permission or an opt-out mechanism and (ii) improve both developers and end-users awareness with respect to the privacy-related concerns raised by IAMs.

Experimental comparison of features and classifiers for Android malware detection

Android platform has dominated the smart phone market for years now and, consequently, gained a lot of attention from attackers. Malicious apps (malware) pose a serious threat to the security and privacy of Android smart phone users. Available approaches to detect mobile malware based on machine learning rely on features extracted with static analysis or dynamic analysis techniques. Different types of machine learning classifiers (such as support vector machine and random forest) deep learning classifiers (based on deep neural networks) are then trained on extracted features, to produce models that can be used to detect mobile malware. The usually-analyzed features include permissions requested/used, frequency of API calls, use of API calls, and sequence of API calls. The API calls are analyzed at various granularity levels such as method, class, package, and family.

In the view of the proposals of different types of classifiers and the use of different types of features and different underlying analyses used for feature extraction, there is a need for a comprehensive evaluation on the effectiveness of the current state-of-the-art studies in malware detection on a common benchmark. In this work, we provide a baseline comparison of several conventional machine learning classifiers and deep learning classifiers, without fine tuning. We also provide the evaluation of different types of features that characterize the use of API calls at class level and the sequence of API calls at method level. Features have been extracted from a common benchmark of 4572 benign samples and 2399 malware samples, using both static analysis and dynamic analysis.

Among other interesting findings, we observed that classifiers trained on the use of API calls generally perform better than those trained on the sequence of API calls. Classifiers trained on static analysis-based features perform better than those trained on dynamic analysis-based features. Deep learning classifiers, despite their sophistication, are not necessarily better than conventional classifiers, especially when they are not optimized. However, deep learning classifiers do perform better than conventional classifiers when trained on dynamic analysis-based features.

Empirical study on code smells in iOS applications

Code smells are recurring patterns in code that have been identified as bad practices. They have been analysed extensively in Java desktop applications. For mobile applications most of the research has been done for Android with very little research done for iOS. Although Android has the largest market share, iOS is a very popular platform. Our goal is to understand the distribution of code smells in iOS applications. For this analysis we used a collaborative list of open source iOS applications from GitHub. We combined code smells defined by Fowler and object oriented code smells studied on Android. We developed a tool that can detect these code smells in Swift applications. We discovered that iOS applications are most often affected by Lazy Class, Long Method and Message Chain code smells. Most often occurring code smells are Internal Duplication, Lazy Class and Long Method.

SESSION: Software development and evolution

Are apps ready for new Android releases?

Context: Android operating system always brings new releases and updates to improve security, increase performance and bring a better user experience. When Google announces a new release, a whole chain of changes is triggered in cascade, causing many compatibility issues. Objective: This study focus at performing a quantitative and qualitative analysis on the state of apps readiness for new Android releases over time. Method: We performed an empirical study to map apps readiness to different Android versions. We developed a Repository Mining Tool to analyse 8420 open-source repositories, detecting 2118 Android projects and when they were adapted to different Android versions along their lifetimes. Results: Our results show that Android apps have became "less ready" over time. We found that 76.45% of the analysed apps were ready for Android Lollipop 5.0 (API level 21) release, in October 2014. Though only 5.46% were ready for Android 10 (API level 29), in September 2019. In addition, our results show that when apps are adapted to an Android version, 59.41% perform the adaptation until the new Android release month, 95% are adapted twelve months after the release, and 99.16% are adapted two years later. Conclusion: Our findings reveal implications that affect not only the Android or mobile development research field and developers, they also reveal implications that points to Google's policies and Android final users as well.

APIMigrator: an API-usage migration tool for Android apps

To provide their functionality, mobile apps interact extensively with the application programming interface (API) of the underlying operating system. Given that this API evolves frequently, app developers are periodically required to migrate API usages in their apps to ensure that the apps behave as expected when running on the new API. To help developers with this tedious, error-prone, and time-consuming task, we defined a technique for automated API migration and implemented it in a tool called APIMigrator that supports Android apps. APIMigrator (1) automatically migrates API usages within an app by leveraging how developers of other apps migrated corresponding API usages and (2) validates the migrations through differential testing. We evaluated APIMigrator on a benchmark of 15 real-world apps and obtained promising results. Overall, our tool was able to migrate 85% of the API usages considered and validate 68% of these migrations. We provide a demo video of the tool at

Doodle2App: native app code by freehand UI sketching

User interface development typically starts with freehand sketching, with pen on paper, which creates a big gap in the software development process. Recent advances in deep neural networks that have been trained on large sketch stroke sequence collections have enabled online sketch detection that supports many sketch element classes at high classification accuracy. This paper leverages the recent Google Quick, Draw! dataset of 50M sketch stroke sequences to pre-train a recurrent neural network and retrains it with sketch stroke sequences we collected via Amazon Mechanical Turk. The resulting Doodle2App website offers a paper substitute, i.e., a drawing interface with interactive UI preview and can convert sketches to a compilable single-page Android application. On 712 sketch samples Doodle2App achieved higher accuracy than the state-of-the-art tool Teleport. A video demo is at

Real-time multi-user spatial collaboration using ARCore

This paper proposes a collaboration application that allows multiuser to add extra contents to live video streaming, based on augmented reality annotation in real-time. Compared to the previous work, we think the integration of remote collaboration and a co-located collaborative way is one of the novelty points of the proposed application. The AR-based collaborative system can render annotations directly on an environment which helps local users easily recognize the original intention that the remote helper wants to deliver. We introduce how the application work.

SESSION: Energy consumption

Should energy consumption influence the choice of Android third-party HTTP libraries?

In mobile devices, the battery is a limited resource and mobile apps are designed with this constraint in mind. To speed up development, app developers often use third-party libraries. Researchers have found that third-party libraries for ads and billing excessively use mobile resources. Other often used third-party libraries such as Android third-party HTTP libraries have received less attention from research regarding energy consumption. To fill this gap, we investigated whether popular Android third-party HTTP libraries vary in energy consumption. In addition, we checked whether there is a correlation between performance and energy consumption. To achieve this goal, we performed a controlled experiment. We created 45 different versions of a custom app and explored the energy consumption and performance of eight popular Android third-party HTTP libraries in five typical use cases. We found that there is a significant variance of energy consumption between the selected Android third-party HTTP libraries. We assume that the energy drivers are related to the internal structure of the Android third-party HTTP libraries, in particular to the handling of asynchronous tasks and the creation of multiple threads in the background. We did not find significant correlation between performance and energy consumption in most of the versions. Our results will help app developers make better choices when selecting Android third-party HTTP libraries.

Greenspecting Android virtual keyboards

During this still increasing mobile devices proliferation age, much of human-computer interaction involves text input, and the task of typing text is provided via virtual keyboards. In a mobile setting, energy consumption is a key concern for both hardware manufacturers and software developers. Virtual keyboards are software applications, and thus, inefficient applications have a negative impact on the overall energy consumption of the underlying device. Energy consumption analysis and optimization of mobile software is a recent and active area of research. Surprisingly, there is no study analyzing the energy efficiency of the most used software keyboards and evaluating the performance advantage of its features.

In this paper, we studied the energy performance of five of the most used virtual keyboards in the Android ecosystem. We measure and analyze the energy consumption in different keyboard scenarios, namely with or without using word prediction. This work presents the results of two studies: one where we instructed the keyboards to simulate the writing of a predefined input text, and another where we performed an empirical study with real users writing the same text.

Our studies show that there exist relevant performance differences among the most used keyboards of the considered ecosystem, and it is possible to save nearly 18% of energy by replacing the most used keyboard in Android by the most efficient one. We also showed that is possible to save both energy and time by disabling keyboard intrinsic features and that the use of word suggestions not always compensate for energy and time.

Evaluating the impact of caching on the energy consumption and performance of progressive web apps

Context. Since today mobile devices have limited battery life, the energy consumption of the software running on them can play a strong role with respect to the success of mobile-based businesses. Progressive Web Applications (PWAs) are built using common web technologies like HTML, CSS, and JavaScript and are commonly used for providing a better user experience to mobile users. Caching is the main technique used by PWA developers for optimizing network usage and for providing a meaningful experience even when the user's device is offline.

Goal. This paper aims at assessing the impact of caching on both the energy consumption and performance of PWAs.

Method. We conducted an empirical experiment targeting 9 real PWAs developed by third-party developers. The experiment is designed as a 1 factor - 2 treatments study, with the usage of caching as the single factor and the status of the cache as treatments (empty vs populated cache). The response variables of the experiment are (i) the energy consumption of the mobile device and (ii) the page load time of the PWAs. The experiment is executed on a real Android device running the Mozilla Firefox browser.

Results. Our results show that PWAs do not consume significantly different amounts of energy when loaded either with an empty or populated cache. However, the page load time of PWAs is significantly lower when the cache is already populated, with a medium effect size.

Conclusions. This study confirms that PWAs are promising in terms of energy consumption and provides evidence that caching can be safely exploited by PWA developers concerned with energy consumption. The study provides also empirical evidence that caching is an effective technique for improving the user experience in terms of page loading time of PWAs.

SESSION: Security and privacy

Representing string computations as graphs for classifying malware

Android applications rely heavily on strings for sensitive operations like reflection, access to system resources, URL connections, database access, among others. Thus, insight into application behavior can be gained through not only an analysis of what strings an application creates but also the structure of the computation used to create theses strings, and in what manner are these strings used. In this paper we introduce a static analysis of Android applications to discover strings, how they are created, and their usage. The output of our static analysis contains all of this information in the form of a graph which we call a string computation. We leverage the results to classify individual application behavior with respect to malicious or benign intent. Unlike previous work that has focused only on extraction of string values, our approach leverages the structure of the computation used to generate string values as features to perform classification of Android applications. That is, we use none of the static analysis computed string values, rather using only the graph structures of created strings to do classification of an arbitrary Android application as malware or benign. Our results show that leveraging string computation structures as features can yield precision and recall rates as high as 97% on modern malware. We also provide baseline results against other malware detection tools and techniques to classify the same corpus of applications.

On the elicitation of privacy and ethics preferences of mobile users

Nowadays, mobile users are constantly being connected and increasingly asked to express their personal preferences in the digital world. User preferences deal with simple device settings options, like notification alarms, as well as relevant ethical choices relating to the user behavior, privacy ones included (e.g., concerning the unauthorized disclosure and mining of personal data, as well as the access to restricted resources). All these preferences define the user, they are the building blocks of her digital identity and will be increasingly important given the growing rise of autonomous technologies and their ethical implications. The settings that enable these preferences are often hard to locate and hard to understand, even in popular apps and operating systems. Moreover, they can expose privacy, be employed for profiling or exploited for malicious activities. In this landscape, we devise the introduction of a Personal Preferences Automation Module (PPAM) capable of automatically inferring, applying and enforcing user choices in multiple scenarios ranging from speeding up simple time consuming tasks, to managing sensitive ethical choices. The wide range of sensors and devices that can be found in the mobile domain makes it a privileged context in which to employ the proposed module. In this paper, we present two application scenarios and describe the proposed approach at work on them.

Vision: alleviating Android developer burden on obfuscation

Mobile applications (apps) have gained an increasing importance in the field of software engineering as they are becoming one of the most widely used type of software. In the Android ecosystem, obfuscation tools are available to optimize, reduce the size and protect the intellectual properties of apps. However, despite the clear advantages provided by obfuscation most apps do not use it, often because of the difficulties induced by the usage of obfuscation which requires writing rules to keep a usable app. In this paper, we identify the concrete challenges encountered by app developers who wish to use obfuscation in their apps. In addition, we propose an approach using crowdsourcing to automatically generate rules, when static analysis is not sufficient. With the knowledge gained from hundreds of projects, we hope to lighten the burden on developers when writing rules.