A first look at Android applications in Google Play related to COVID-19

Due to the convenience of access-on-demand to information and business solutions, mobile apps have become an important asset in the digital world. In the context of the COVID-19 pandemic, app developers have joined the response effort in various ways by releasing apps that target different user bases (e.g., all citizens or journalists), offer different services (e.g., location tracking or diagnostic-aid), provide generic or specialized information, etc. While many apps have raised some concerns by spreading misinformation or even malware, the literature does not yet provide a clear landscape of the different apps that were developed. In this study, we focus on the Android ecosystem and investigate Covid-related Android apps. In a best-effort scenario, we attempt to systematically identify all relevant apps and study their characteristics with the objective to provide a first taxonomy of Covid-related apps, broadening the relevance beyond the implementation of contact tracing. Overall, our study yields a number of empirical insights that contribute to enlarge the knowledge on Covid-related apps: (1) Developer communities contributed rapidly to the COVID-19, with dedicated apps released as early as January 2020; (2) Covid-related apps deliver digital tools to users (e.g., health diaries), serve to broadcast information to users (e.g., spread statistics), and collect data from users (e.g., for tracing); (3) Covid-related apps are less complex than standard apps; (4) they generally do not seem to leak sensitive data; (5) in the majority of cases, Covid-related apps are released by entities with past experience on the market, mostly official government entities or public health organizations.


Introduction
The outbreak of the Covid-19 pandemic in late December 2019 has quickly and severely impacted countries worldwide, becoming one of the most important health crisis of the 21 st century (Spinelli and Pellino 2020;Remuzzi and Remuzzi 2020).The infectious agent, a coronavirus, identified as responsible for this disease is notoriously difficult to pin down: while it leaves many infected people without symptoms, it can lead to a common cold for some and even severe respiratory disorders to others (Clerkin et al. 2020).So far, the Covid-19 has brought about a human tragedy, with hundreds of thousands of lives lost, as well as an economic downturn due to the lockdown of over three billion people (half of humanity) (Mahase 2020;Dudel et al. 2020).
The scale of Covid-19 effects has urged stakeholders at different levels (government, local authorities, private companies, academia, NGOs, and citizens) to plan and implement measures for addressing the virus spread.In particular, while pharmaceutical research is embarked on a long journey to develop a vaccine, non-pharmaceutical innovations are sought to contribute to responding to the outbreak.Chief among the technologies that are employed, Information and Communication Technology has been widely leveraged in various capacities across all regions and targeting broadly all levels of society.For example, news and sensitization messages have been viral thanks to the use of internetbased services such as social networks.Given the widespread use of handheld devices such as smartphones, users are keen to install applications (often referred to as apps in the mobile realm) that have specific purposes for entertainment, business, productivity, news, and social networking.Following the outbreak of the pandemic, authorities, non-governmental organizations, and independent developers have engaged in an app development race to provide readily-available digital tools to the modern citizen.
We focus in this paper on the case of the Android ecosystem.With the largest market share on mobiles (86% in 2020 (IDC 2020)), Android constitutes a prime choice for developers and users alike.Initial reporting on Covid-19 related apps are focused on the problems that such apps raise: (1) He et al. (2020) have already explored the case of coronavirus-themed Android malware; (2) Google, the maintainer of the Play Store (i.e., the official Android app market) has decided to crack down on Covid-19 apps to combat misinformation, sometimes with an excessive zeal (i.e., legitimate apps can be temporarily banned just for sharing Covid-19 information (Carman 2020;Google 2020)).
Our study is of a broader and more generic dimension.It is about characterizing the applications that are related to the Covid-19 outbreak: What are they for?Who developed them?To what extent can they be considered dangerous?These are some of the questions that we undertake to investigate.To that end, we have considered 184 563 apps released in the time window of July 2019-May 2020 and collected in the AndroZoo dataset (Allix et al. 2016).From this initial set of Android apps, by following well-defined heuristics, we were able to identify 44 Covid-related apps.Given the limita-tions of the crawling of AndroZoo, we also considered other online resources such as the GitHub search engine, specialized Covid-19 technology blogs, etc., and we finally raise the number of collected Covid-related apps to 92.We extract different features from these apps and provide summary statistics on their characterization.
This paper presents our analyses exploring the use of permissions and libraries, the presence of leaks, the malicious status, the code size and complexity, the authorship, and the described purposes.Mainly, we establish that most of the apps are made for informing people, monitoring their health, and tracing users with the goal of preventing the spread of the virus.In addition, we note that Covid-related apps are not flagged by malware detectors.Yet, we found that some of them have been removed from Google Play.Then, after assessing the complexity of Covid-related apps and comparing them with standard apps, we found that on average, Covid-related apps are less complex, which has been shown to often be indicative of quality for apps (Jošt et al. 2013).Finally, we applied state-of-the-art security and privacy scanners in order to check for potential cryptographic API misuses and data leaks in the code.
In summary, we present the following contributions: -We present the first systematic study of Covid-related apps that explores their characteristics and compare them with other non-Covid-related apps.-We build a taxonomy of Covid-related apps based on their described goals.
-We apply literature analysis tools on Covid-related apps and discuss their results.

Research questions
The main objective of this work is to analyze and understand Covid-related apps.To do so, we empirically observe apps characteristics by extracting features that provide insights toward understanding those apps.Hence, to accomplish this objective, we plan to answer the following research questions: -RQ 1: What are Covid-related apps used for?-RQ 2: When did Covid-related apps start to appear in Google Play? -RQ 3: Do Covid-related apps have specific characteristics?-RQ 4: Are Covid-related Android apps more complex than standard apps?-RQ 5: To what extent were Covid-related apps removed from the official Google Play and why? -RQ 6: Who are Covid-related apps' developers?-RQ 7: Are Covid-related apps flagged by privacy & security scanners?
The remainder of this paper is organized as follows.First, we present an initial Android apps dataset and some background for the reader in Section 2.
Then, in Section 3, we give details about the experimental setup of our study.In Section 4, we provide the results obtained from our experiments.We discuss the threats to validity of our study in Section 5. Finally, we discuss related work in Section 6 and conclude in Section 7.
2 Dataset and Background on App Analysis Initial Dataset: In order to perform our experiments and to exhaustively answer the research questions presented in Introduction, we need to rely on a comprehensive dataset.Consequently, we used the state-of-the-art largest dataset available, namely AndroZoo (Allix et al. 2016).At the time of writing, AndroZoo contains more than 11 million Android apps (June 2020) that have been collected from different sources, such as Google Play and other third-party providers (F-Droid, Anzhi, AppChina, etc.).As it is continuously growing, researchers still heavily rely on AndroZoo for collecting apps and experimenting on it (Ranganath and Mitra 2020;Shar et al. 2020;He et al. 2020;Xu et al. 2020).
Since the Covid-19 outbreak is quite recent, we did not consider the 11 million Android apps present in AndroZoo.For our study, we instead considered and collected from AndroZoo all the apps ranging from July, 1 st 2019 to May, 25 th 2020, leading to a total number of 184 432 collected apps.Note that AndroZoo does not contain information about the release date of an app, i.e., there is no information about when an app has been uploaded on the market.For this reason, we approximate the date of an app by considering the date of the dex files in apks.Figure 1 shows the distribution of collected APKs according to the month of the date of the dex files in APKs.Dataset Augmentation: It is a known problem that the date metadata in APKs (approximated as the date of the dex files) is not always reliable (Li et al. 2018).For this reason, we decided to augment our dataset (1) by not considering the time window from July 2019 to May 2020 and (2) by selecting any apps from AndroZoo whose package name contained the words covid-19, covid19 or coronavirus.With this heuristic, we were able to retrieve 131 apps from AndroZoo.
⇒ Finally, our initial dataset is constituted of 184 563 Android apps.Short Background on Android Apps: In general, Java is the language used by developers to create Android applications, though other languages can be used such as Kotlin, or even languages such as C and C++ thanks to the Native Development Kit (NDK).For common Java applications, the source code is compiled into Java bytecode and run into the Java virtual machine.However, within the Android operating system lies the Dalvik Virtual Machine which is an optimized version of the Java Virtual Machine.Thus, regarding mobile apps, the source code is first compiled into Java bytecode, then translated into Dalvik bytecode and written in DEX files.

Number of apps collected from AndroZoo by month
Fig. 1: Monthly distribution of Android apps considered in this study tion Package (APK).This format is used for distributing apps on devices, not only via official and unofficial markets but also via any other channels where an independent file can be distributed.Indeed, Android devices allow users to install applications from any sources once a configuration option is toggled off.
An APK is typically a zip file containing the following files: (1) Metadata files, (2) the certificate(s) used to sign the application, (3) a lib folder containing platform-dependent compiled code, (4) compiled and non-compiled resource files, (5) one or multiple DEX files, generally named classesX.dexwith X an integer, (6) an AndroidManifest.xmlfile describing the application (package name, components, version, access rights, etc.).

Experiment Setup
In this section, we describe the setup of our experiments.More specifically, in Section 3.1 we describe how we create a dataset of Covid-related apps by searching apps in both our initial dataset extracted from AndroZoo, and on the web.Then we describe in Section 3.2 how the information needed for answering the research questions was extracted.

Dataset curation
The large majority of apps contained in our initial dataset of Android apps are not related to the Covid-19 outbreak.Consequently, we need to curate this dataset.Our first idea was to rely on a clear definition of what is a Covid-related app.However, we realized that finding a correct and precise definition is not obvious.For instance, the english Wikipedia page related to Covid-19 (Wikipedia 2020) defines Covid-related apps as "mobile software applications that use digital tracking to aid contact tracing in response to the COVID-19 pandemic, i.e. the process of identifying persons ("contacts") who may have been in contact with an infected individual."(accessedearly June 2020).We quickly considered this definition as too restrictive since we found several Covid-related apps that are not about "digital tracing".As a result, rather than relying on a definition of what is a Covid-related app we (1) implemented several heuristics based on hypotheses, and (2) performed several quality checks to filter out irrelevant apps.
Hypothesis 1: Our first hypothesis is that a Covid-related app contains strings (e.g., in a class name or method name) related to Covid-19.Based on this hypothesis, we defined several regular expressions that we apply on various fields on an Android app.More specifically, we proceed as follows: Let A be a set of Android apps and apk an app of this set.Let be:  2, we can see the number of apps that were retrieved per keyword (note that several keywords can be present in a given app).
Quality Check #1: We can see in Figure 2 that the keywords corona and quarantine match a significantly larger number of apps.After investigation, we found that the keyword quarantine is used in strings by developers to check if the smartphones are rooted, i.e., if the user has super-user (root) privilege.Indeed, the presence of some packages can characterize the fact that a smartphone is rooted.Among such packages, one is named com.ramdroid.appquarantineand another is named com.ramdroid.appquarantinepro.To verify the presence of these two packages, developers often use these package names as string, and thus the regular expression using quarantine matches.This explains why there is such a high number of apps containing quarantine in strings.We manually analyzed several dozens of apps and confirmed they are not real Covid-related apps (mainly music, education, news, shopping, and game apps).Regarding the corona keyword, it also appears in a substantial number of apps, especially in string feature.After manual investigation, we found that apps retrieved with this keyword use a framework called Corona developed by Coronalabs1 .These apps are mainly games, entertainment and personalization apps.Moreover, the word corona not only refers to Covid-19, but also has uses in several unrelated contexts (in architecture, beverages, books, music, movies, and games).
We decided to rule out both keywords corona and quarantine, as they revealed to mostly bring noise in our dataset.After filtering the apps gathered, i.e., not taking into account the corona and quarantine keywords, we obtained 4103 apps.
Hypothesis 2: Our second hypothesis is that since AndroZoo is known to contain successive versions of the same apps (Allix et al. 2016;Li et al. 2018), our dataset contains successive versions of the same Covid-related app.Therefore, for the subsequent analyses, we only keep the latest version of a given app.This step is performed by comparing apps version code-available in AndroZoo metadata-and keeping the highest value.After this step, our dataset contains 750 apps.
Hypothesis 3: Our third hypothesis is that official Covid-related apps are released only on the official Android market, i.e., the Google Play market.Figure 3 depicts the distributions of the apps per market from where they have been crawled.Most of the apps were released in the official Google Play market.By considering apps from Google Play only, we reach the number of 619 apps.

Quality Check #2:
To check the quality of our dataset, we inspect the date of the dex file in the APKs.The Covid-19 outbreak was not known until the very end of 2019, so we expect to find apps from 2020 only.Figure 4 presents the distribution of those apps by the date of the dex file in the APK.First, we observe that fewer apps were released during winter 2019-2020.Then we can see that starting from March 2020, a significant amount of apps were developed, which corresponds to the containment period.Finally, Figure 4 show that our set of Covid-related apps contains a significant amount of apps developed before the pandemic, which indicates the low quality of our dataset and that an additional layer of filtering is needed to only keep Covid-related apps.Indeed, our manual investigations of those apps revealed that they contained keywords that may be overly broad.For example, it appeared that keywords such as outbreak, pandemic, containment or lockdown were popular in games, most notably in the zombies and survival games genres. 2 0 1 9 -0 8 2 0 1 9 -0 9 2 0 1 9 -1 0 2 0 1 9 -1 1 2 0 1 9 -1 2 2 0 2 0 -0 1 2 0 2 0 -0 2 2 0 2 0 -0 3 2 0 2 0 -0 4 2 0 2 0 -0 5

Number of apps
Monthly distribution of our filtered set of 619 apps Hypothesis 4: Our fourth hypothesis is that the description of an app is a reliable source of information to check that an app is Covid-related.AndroZoo does not provide the app descriptions, so we queried Google Play to retrieve the description of the 619 apps.Note that it was possible to get the description for 537 apps, while 82 apps were not available anymore in Google Play at that time.Actually, AndroZoo uses a crawler to automatically download apps.It is possible that AndroZoo downloads an app at a time t 1 , but if Google Play decides to remove this app (or if the developer decides to remove it) at time t 2 , with t 1 < t 2 , it is not possible to access the description of the app anymore after time t 2 .We found that Google is actively deleting apps that are violating their policy with respect to Covid-19 (Carman 2020;Google 2020).
Manual investigations of the remaining 537 apps were conducted to qualify an app as Covid-related.We did so by analyzing the Google Play page of every app, reading the description, and looking at screenshots.We did not encounter any ambiguous case, hence it was straightforward to qualify an app as Covidrelated or not.With this method, we determined with high certainty that 44 apps taken from AndroZoo were Covid-related.
Note that the set of apps from our initial dataset for which the package name contained Covid-related keywords contained 131 elements and we only retrieved 44 apps.This is explained by the fact that among those 131 apps, there were, for almost one-fifth, different versions of the same app.In addition, we were not able to analyze the Google Play page of apps that were not available in the Google Play market.
Figure 5 gives two examples to show an app we discarded and an app that was considered as Covid-related.First, the game picture on the left clearly shows a game in which the user has to kill zombies, even though the game contains the string outbreak.Second, the Covid-related app on the right is explicit regarding the content delivered to the user.The title refers to Covid-19, the description gives clues about the content, i.e., information about Covid-19 and guidelines about Covid-19.Besides, the screenshots are also explicit by depicting what actions users can perform, here Covid-related actions, i.e., getting information about Covid-19, performing self-diagnosis, receiving guidelines, and news about Covid-19.It represents how unequivocally our decisions were to be made to qualify an app as Covid-related or not.
Hypothesis 5: As AndroZoo is not exhaustive and all the apps in the Google Play market are not available for download from every country (An-droZoo crawls from specific countries), we felt the need to expand our research.Consequently, manual investigations were conducted on the web to search for apps that would not be available in Google Play and/or that our empirical analysis on AndroZoo did not catch.We found 48 additional apps from diverse countries from different time span, e.g.we found apps that were first seen in June 2020.Thus, our set of COVID-applications reached a length of 92 apps.

Features extraction
In this section, we expose what features were extracted from the apps considered in this study and how we extracted it.Three ways of retrieving the needed information were considered: (1) Extracted from AndroZoo, (2) Automated analysis of the apps, and (3) Manual analysis, which we describe below.

AndroZoo metadata
Apps from AndroZoo are provided with additional metadata, i.e., a vector of length 11 representing different information.Those metadata elements are used to expose several properties of the apps under analysis.We consider the date of the dex files to verify when the Covid-related apps started to appear.
The package names and version codes were useful to have some insights into the versioning of the apps and to keep only the latest version of the app considered, i.e., one version of the app will be analyzed.When available, the number of AntiVirus products reporting an app as malicious (obtained from VirusTotal) was used for qualifying the maliciousness of the apps.Finally, AndroZoo metadata also indicate the source where an app was obtained from.

Automated analyses of apps
AndroZoo metadata being limited, we additionally leveraged existing tools and frameworks to analyze Android apps in order to obtain the information to acquire the information needed by our study.
-We developed a program relying on the Androguard software package2 .Thanks to this tool, we extract the permissions requested by the apps, as well as information about components in the applications (i.e., Activities, Services, etc.).We also automatically compute the complexity metrics described in Appendix A. -We used CogniCrypt (Krger et al. 2017) and its headless implementation CryptoAnalysis (CryptoAnalysis 2020) to check that they were written following best practices regarding cryptographic APIs.-As several Covid-related apps seek to acquire and process the location of users (considered a piece of sensitive information), we verify if those apps were subject to data leaks.Hence, we leverage FlowDroid-IccTA (Li et al. 2015), the state-of-the-art static data leak detector dedicated to Android apps.

Manual analyses of apps
While application information that can be readily obtained through automatic tools were necessary for our study, we went a step further and acquired qualitative data on the apps.By collecting and carefully reading the descriptions of apps 3 and by confirming the validity of our understanding by matching the reviews and app screenshots with the description, we were able to assemble highly-qualified data on the goals of app developers, on whether an app is developed by a state body or an individual person, etc.
Finally, after leveraging the automated tools described above, their results were manually confirmed for the set of 92).This allowed us to ensure those tools did not yield false-positives, and that their results were sound and consistent.

Results
In this section, we present our experimental results and we answer our research questions.

What are Covid-related apps used for?
Textual descriptions of apps on markets generally provide a wealth of information on the purpose and functionalities that developers advertise.We undertake to systematically examine the descriptions of all the apps under study.Unfortunately, since Google Play is actively moderating Covid-related apps, we have faced an issue with some apps that we were able to initially collect but which were no longer available on the market at the time of analysis.Eventually, our analysis of descriptions was performed on 78 apps.In other words, from the time we read the descriptions of the apps to curate our dataset (as explained in Section 3.1) and the time we perform this more in-depth study, i.e., collecting information related to the features of the apps, 14 apps (92−78) were not able anymore on GooglePlay.A Taxonomy of Covid-related apps.After a careful analysis of information available in Google Play, we summarize for each app its general goal, i.e., which aspects of the Covid-19 crisis the app is precisely intended to address.Eventually, we identified three main categories to which each app can be associated with possible overlap between categories, i.e., an app can be associated with several categories: 1. Information broadcast (top-down) -Apps in this category aim to provide users with various types of information, from general guidelines, infection statistics to general Covid-19 news.Although such apps are not always officially released by government bodies, they often relay official information from top (authorities) down (users).2. Upstream collection (bottom-up) -Apps in this category collect information from users and make it available to the developer and/or an official body, such as a country's health authorities.3. Tooling -Apps in this category serve as tools with functionalities that directly deal with daily aspects of the Covid-19 (e.g., generation of certificates).
[1] Information broadcast.From the collected dataset of Covid-related apps we identified several distinguishing scenarios in apps performing information broadcasting.Figure 7 overviews the related characteristics, notably based on the types of information that are made available to the user:  1. Guidelines on measures to take to minimize the risk of infection -Among such apps, some render maps highlighting high-risk areas.Other apps provide behavioral advice (e.g., how to wash hands), leveraging the whole spectrum of available media: (1) textual descriptions (for the majority of apps), (2) videos and, (3) audio clips.2. Continuously-updated Statistics on the pandemic evolution; 3. General information about Covid-19, such as about the typical symptoms.
We identified two different scenarios in the provision of general information: -Some apps present curated information, i.e., information that is somehow checked and filtered by the development team before it is shown to the public.Such information is often tagged in a way that allows interested people to find the source, and gauge its credibility.Sometimes, these apps are developed directly by an entity that itself carries credibility as a source of information, such as national healthcare authorities.An example is the MyHealth Sri Lanka app4 developed by the national ICT Agency, which presents to the user verified information on the current Covid-19 status.-A number of apps appear to provide unfiltered information regarding Covid-19.Their developers are not always themselves entities that would traditionally be assumed to have any specific credibility on the matter.For example, the DiagnoseMe app5 , which claims to provide the user with all the information on the virus, is proposed by an association with unrecorded expertise in health.
[2] Upstream collection.Most apps in our dataset perform data collection from users.This suggests that many app providers consider data to be key in the mitigation of the Covid-19 crisis.App providers indeed collect a variety of information, including user personal information (e.g., name, age, address, etc.), some medical information (e.g., whether a user is infected with Covid-19, the therapies that are used).Some apps are even used to keep a health diary (sharing information about symptoms every day), or to report the infection of people in the app user's acquaintances.Overall, we have identified three different ways in which apps collect user data, as summarized in Figure 8.Note that in the case of data collection and spread tracking apps, we did not try to qualify whether apps were as privacypreserving as their developers claimed they were (e.g., data is deleted after N days), nor to determine to what extent the collected data is shared with third parties.
Similarly, for this paper, we did not analyze the inner workings of contacttracing apps, and we did not evaluate the merit nor the opportunity of contacttracing, this having already been-and still being to this date-discussed by security researchers (Culnane 2020;Anderson 2020;Baumgrtner et al. 2020).Fig. 8: Upstream category Several apps take inputs from the users to offer diagnoses related to Covid-19.Such apps can provide a built-in questionnaire that users have to fill within the app, or leverage a virtual assistant or chatbot.In these cases, the diagnosis can be made automatically, with no interaction nor confirmation with a trained medical practitioner.
Other apps, however, provide a somewhat more traditional medical visit experience, by offering the facilities needed to remotely exchange (e.g., via instant text messages as well as voice and/or video calls) with a medical doctor.Such apps are used from home, since millions of people worldwide were confined, and were potentially reluctant or unable to visit a brick-and-mortar doctor's office.
Additionally, some apps are developed to track the spread of the virus by locating the users of the apps.While a few of those apps use simple geomonitoring with GPS information for tracking users, most apps do it automatically.Nevertheless, we found a few apps that request users to provide a-posteriori the locations they have visited on a given day.We also identified one app which uses QR code scanning at the entrance of public buildings to obtain precise location information, while still being fully under users' control.
With respect to tracing, a few apps promote social-distancing using the GPS location of users, the goal being to not approach other people too closely.
Furthermore, several apps implement contract-tracing, i.e., the ability to retrieve who a specific person has been in contact with, providing users a way to know if they have encountered someone infected, and potentially infectious.Contact-tracing apps mainly rely on three methods, (1) Using the GPS location of users, (2) Using the Bluetooth technology to detect proximity, and (3) Using a location diary that the users have to manually fill.
[3] Tooling The last category is the tooling category which includes several types of tools aimed at helping users deal with some consequences of the Covid-19 crisis (see Figure 9).A few apps allow users to auto-generate documents for their local authorities (e.g., travel authorization that had been made mandatory in several countries during containment).
Users can also install apps offering appointment-capabilities for medical purposes, or selling Covid-related products (e.g., masks, hand-sanitizers, etc.).
On the entertainment front, apps were released proposing games around the pandemic, or providing users with Covid-19-themed image filters, for example adding a virtual mask, or adding virtual decorative elements to an actual mask.
Lastly, apps were also made to cater to the newly-discovered needs of massive remote education.

Fig. 9: Tooling category
The interested reader can inspect Tables 2, 3 and 4 for more information about the mapping between categories and the apps for which we were able to retrieve the relevant information.
RQ1 Answer: Our empirical analyses show that overall, Covid-related apps are mainly developed for: 1. Providing users with information; 2. Collecting data from users; 3. Offering users Covid-related tools.
4.2 When did Covid-related apps start to appear in Google Play?
In this section, we consider the 92 collected Covid-related apks 6 .As the first release date of apps is not available on Google Play, we need to rely on other sources of information to try to find the first appearance date.Therefore, in order to visualize when Covid-related apps appeared, we selected, for each app (when available), the minimum date among the AndroZoo added date, the VirusTotal first seen date and dates available in third party datasets (i.e., Koodous 7 , APKCombo 8 , APKPure 9 ).The results can be seen in Figure 10.
We can see that those apps started to appear as early as February 2020, i.e., before the pandemic was officially recognized by the World Health Organization (i.e., March, 11 th 2020) (Ghebreyesus 2020).
We can also see in Figure 10 that since February 2020, the number of new apps seen has only increased.We note that for June 2020, the lower number is not significant: Our study, and app collection process, stops in early June 2020 10 .
RQ 2 Answer: Covid-related apps started to emerge on markets as early as February 2020, i.e., even before the World Health Organization officially declared Covid-19 as a pandemic.

Do Covid-related apps have specific characteristics?
In prior work, Tian et al. (2015) have shown that specific sets of apps can have similar characteristics (e.g., similar permissions, components, size, etc.).In this section, we investigate to what extent 92 apps form one coherent group that is significantly different than other apps.
To that end, for each app, we counted the number of different Android components (i.e., Activities, Broadcast Receivers, Services, and Content 6 We remind that we collected 92 apks but only 78 descriptions.So, if descriptions are required to perform a study, 78 apks are considered.If not, 92 apks are considered. 7 https://koodous.com 8https://apkcombo.com 9https://apkpure.com 10We remind the reader that, while our initial app dataset built from AndroZoo contains apps until end of May 2020, our dataset manually collected contains apps from June 2020 (See Section 3.1)  Providers), computed the size of the dex file, extracted the permissions needed as well as the libraries used.
Comparison Dataset: For comparing the characteristics of Covid-related apps with other apps characteristics, we randomly selected 5000 apps from our initial dataset of 184 563 apps.Those 5000 apps are sampled from the same time span (i.e., they are coming from the same initial dataset) to ensure that time is not a factor in potential differences.In the rest of this section, those 5000 apps are referred to standard apps, in opposition to Covid-related apps.
Android Components: Figure 11 depicts differences between standard apps and our set of Covid-related apps regarding the number of components included in the app.We notice that Covid-related apps tend to use fewer activities than standard apps and that their distributions are different.However, it is unclear whether or not there is a significant difference for the remaining components, i.e., services, broadcast receivers and content providers.Therefore, we apply the Mann-Whitney-Wilcoxon (MWW) test which is a non-parametric statistical test for assessing the significance of the difference between two distributions (Mann and Whitney 1947;Wilcoxon 1945).
We obtained the results available in Table 5.At a 0.05 significance level, we can see that the distribution of the number of activities in standard and Covid-related apps are significantly different, likewise for providers.However, regarding the distribution for services and receivers, the difference does not seem to be much significant.
Therefore, the difference between those apps seems to appear at the GUI level since Activities are the main UI building block in Android apps.This suggests that Covid-related may have less complex GUI than standard apps.Content Providers are slightly more used in Covid-related apps, which could hint at Covid-related apps being more data-centric than standard apps.Dex files size: Figure 12 shows the distributions of the dex sizes of both the Covid-related and standard apps.An MWW test confirms the statistical significance of the difference between both distributions.Figure 12 also shows that the majority of Covid-related apps are smaller than typical standard apps.However, the maximum dex size value is higher than standard apps, hinting at a higher diversity amongst Covid-related apps.
Permissions: In Table 6, we can compare the permissions used by the set of standard apps and the permissions of the Covid-related apps.For this purpose, we extracted for both sets of apps the top ten most requested permissions.First, a notable difference is that Covid-related apps tend to use more the wake lock permission.This permission is used for preventing the screen of the device from being turned off, and/or to ensure an app remains active.This feature is likely used for keeping the phone awake while locating the phone.In the same way, access fine location and access coarse location tend to be used more by Covid-related apps.
Figure 13 shows the distributions of the number of permissions requested by standard apps and Covid-related apps.We note that these distributions are significantly different, as confirmed by a p-value of 10 −9 for the MWW    Libraries: To compare the patterns of libraries inclusion, we measure the use of libraries by relying on a collection of well-known libraries.More specifically, we re-use two lists of libraries established in prior works (Li et al. 2019;Li et al. 2016): a list of 1 114 common libraries and a list of 240 advertisement libraries.Therefore, for both Covid-related and standard apps datasets, we computed the number of apps using at least one common library and one advertisement library.
Table 7 presents our results.First, we notice that almost all the apps (Covid-related and standard) use common libraries, which is not surprising since Android software development-just like non-mobile software-heavily relies on reusable libraries and frameworks.
However, the difference is significant regarding the advertisement libraries.Indeed, while advertisement libraries are used by more than 80% of standard apps, they only appear in less than 20% of Covid-related apps.Furthermore, only 3 out of 240 advertisement libraries are used in Covid-related apps, namely: (1) com.facebook, (2) com.startapp.androidand (3) com.flurry.This strongly suggests that the primary goal of Covid-related apps is not to obtain a financial gain from advertisement, in opposition to the vast majority of standard apps.4.4 Are Covid-related Android apps more complex than standard apps?
App complexity is an elusive concept.Yet, in the literature, there are various studies that propose metrics to measure some form of complexity and attempt to show its correlation with app quality and maintainability (Jošt et al. 2013;Gao et al. 2019b).We undertake to investigate our research question based on these common metrics from the literature (Chidamber and Kemerer 1994).
We provide in Appendix A the descriptions of the complexity metrics we use.
Since complexity comparison can be biased by the size of apps, we propose to compare the complexity metrics of Covid-related apps with 5000 randomly selected standard apps whose size is within the same size range (0 MB < size < 20 MB).
The different metrics attempt to capture the Lack of Cohesion in Methods (LCOM), the Weighted number of Methods per Class (WMC), the number of methods invoked per class, i.e., the Response For a Class (RFC), the Coupling Between Object classes (CBO) and the Number Of Children per class (NOC).
Figure 14 presents the distributions of metric values.LCOM, WMC, and CBO appear to present similar distributions across standard and Covid-related apps.However, MWW test revealed significant differences between the distributions of Covid-related apps and standard apps only for the following metrics: WMC (p = 0.0012), RFC (p < 10 −7 ), CBO (p < 10 −4 ) and NOC (p < 10 −4 ).
Overall, these results establish that Covid-related apps are, to some extent, less complex than standard apps.According to (Jošt et al. 2013), this result suggests that Covid-related apps may be more maintainable and of better quality.Additionally, we note that a lower complexity could also indicate that Covid-related apps have on average less functionalities and/or are focused on more specific goals, as was already hinted above in the permission usages comparison.
RQ 4 Answer: Our empirical study shows that Covid-related apps tend to be less complex than standard apps.We have seen in Section 4.1 that during our analyses, some Covid-related apps disappeared from the official Google Play in a matter of days.Google has announced in its policy regarding Covid-related apps (Google 2020) that specific checks are performed on Covid-related apps.In this section, we measure the impact of this policy on the removal of Covid-related apps from Google Play.Therefore, for each app that was initially identified at the beginning of our study (i.e., one of the 92 Covid-related apps), we queried the Google Play market, at the time of writing, to check if the app is still available.Around 15% of Covid-related apps (i.e., 14 apps) have been removed from Google Play.
In comparison, among 1675 standard apps taken randomly from our initial dataset (see Section 2), we found that 277 (i.e.16.54%) apps were removed from the Google Play market.
The removal rates of both app datasets are close.Actually, we expected a much higher removal rate for Covid-related apps.This relatively low ratio of removal for Covid-related apps could be explained in several ways: -Google either enforces its policy very quickly or pre-screens (i.e., before it is accepted on the market) each app that is potentially relevant to Covid-19; In that case, apps would either never make it to the market, or would be removed too quickly for AndroZoo crawlers to catch them; -App developers either rapidly adapted to Google's policy and/or very few developers proposed apps that conflict with Google's policy.
RQ 5 Answer: We are not able to reliably answer this research question.Indeed, even if, based on our study, we could conclude that Covid-related apps are removed from Google Play at a normal rate, we suspect that Google removes very quickly Covid-related apps that do not comply with the Google policy.To reliably conclude on this question, one would need access to key data about the stream of apps submitted to Google Play validation process.
On Google Play, in each web page of an app11 , there is a field developer that provides the name of the person or entity (e.g. a software company, a governmental institution, an OMG, etc.) who has released the app.After collecting this information, we detail in Table 2 (column Developer Type) the status (or the type) of the entity having released an app.Table 8 presents the number of released Covid-related apps for each type of entities.We can see that most of the app providers are governmental institutions.We indeed find Covid-related apps that are officially promoted by national governments (e.g.Government of Brazil12 or Government of France13 ).We also see apps released by more local governmental bodies (at the state or regional level).We have for instance apps from specific states of the USA  ).About 20% of the Covid-related apps (17 apps) are provided by companies.In order to understand why these apps have not been removed by Google, we further check the description of these apps and the descriptions of the companies.We found that: -Even if the developer is identified as a company, two apps have been developed on behalf of official bodies (Care19 16 is the official COVID-19 app for the states of South Dakota and North Dakota, COVID AP-HM 17 is an app developed for a hospital); -Seven apps are either endorsed by a ministry 18 , or working in close collaboration with medical/health actors 19 , or working in collaboration with renowned universities 20 .-Two apps are actually online shopping apps 21 .
-One app is not on the market anymore 22 .
-Finally, five apps related to social distancing 23 , or health 24 , or Covidrelated news 25 , have been released by companies without any explicit link to official organizations.We remind that the official Google Covid-19 policy (Google 2020) is that Covid-related apps with no explicit links with governmental bodies or health organizations cannot provide "health claims".We further check these 5 apps, and we confirm that they comply with the Google Covid-19 policy.
For the remaining nine Covid-related apps, we noticed that 3 apps have been provided by associations.More specifically, the DiagnoseMe26 app has been released by the Faso Civic association from Burkina Faso, the Self Shield App27 by the Commonwealth Medical Association (through the Commonwealth Centre for Digital Health organization) and the COVID Safe Paths28 app by a non-profit organization related to MIT.We also noticed that two apps have been developed by independent developers, and two other apps have been provided by researchers.One by a group of researchers from German Universities29 , one by researchers from the Aga Khan University in Pakistan30 .Finally, one app has been provided by an NGO (i.e., the Austria Red Cross), and one by a hospital (actually a group of hospitals in Paris, France).
We note that among all the Covid-related apps, 71% of them have been released by entities having multiple Android apps on Google Play.
Finally, we represent in the map of Figure 15 the geographical distribution of the apps over the world.We can see that Covid-related apps are provided world-wide (maybe less present in Africa).The countries in blue are the ones listed in Table 2.Note that we also identified 16 other apps from 16 countries that we were unable to obtain; These countries are represented in red.RQ 6 Answer: Most Covid-related apps that are seen on Google Play are developed with/for official government entities or public health organizations.We note that apps developed by non-official entities are not providing health claims, as per Google's policy.Covid-related apps have been released in countries from all continents.The majority (71%) of Covid-related developers have at least one other app available on Google Play.

Are Covid-related apps flagged by privacy & security scanners?
In contrast to a recent work (He et al. 2020), which focused on dissection Covidrelated malware, our aim in this work is not to perform an extensive security analysis of these apps.Nevertheless, we propose to leverage three practical security and privacy scanners on our set of 92 Covid-related apps in order to systematically evaluate three S&P aspects: (1) the presence of privacy leaks; (2) the number of apps flagged by VirusTotal; (3) the misuse of crypto-APIs.
[Privacy leaks] As we have seen in 4.1, most of the Covid-related apps are made for collecting personal and sensitive data, e.g., health data and/or the location of users.Therefore, the security and privacy aspects of these apps are crucial, and many people started to share concerns related to this topic (Page 2020;Parliament 2020;Stolton 2020).In order to assess the privacy of Covidrelated apps, we applied the state-of-the-art data leak detector FlowDroid-IccTA (Arzt et al. 2014;Li et al. 2015).Through static analysis, this tool is able to detect sensitive data leaks intra-component (e.g., inside an Activity) or inter-component (e.g., across Activities).
FlowDroid-IccTA was able to detect 24 intra-component data leaks in 2 different apps and found no inter-component leak.The app SODI31 contained only 1 potential leak, whereas the app Coronavirus -SUS 32 contained 23 potential leaks.Given that static analysis tools are subject to false-positives, we undertake to manually analyze every detected leak.
SODI an app promoting social-distancing.The app is not originating from a government.Our manual analysis concluded, however, that the reported leak is a false-positive alarm and does not constitute a real data leak.
Regarding Coronavirus -SUS, which is an official app of the government of Brazil, FlowDroid-IccTA flagged 24 potential sensitive leaks (i.e., there is a path between a source (that can access a sensitive data) to a sink (e.g.sendSMS)).We notice that four of these leaks allow the app to get the longitude and/or latitude (the sources) of the app to log it internally (the sink).However, this does not necessarily constitute a malicious behavior.
[AntiVirus detection] For each of the Covid-related apps, we have collected the detection reports from over 60 AntiVirus products, thanks to the VirusTotal API33 .None of Covid-related apps is flagged by any of the 60 anti-virus software available in VirusTotal at the time of writing.
[Crypto-API misuses] Finally, we leverage the state-of-the-art staticanalyzer CogniCrypt (Krger et al. 2017) through its headless implementation CryptoAnalysis (CryptoAnalysis 2020) for detecting cryptographic API misuses in Java programs.Such misuses could indeed indicate security issues.We found that 81 apps among our set of 92 Covid-related apps use JCA34 APIs.However, CogniCrypt did not report any cryptographic misuse.
In contrast, Gao et al. (2019a) have shown that in a dataset of more than 598 000 apks, 96% of apks using JCA exhibit dangerous misuses of cryptographic APIs.With 0%, Covid-related apps seem to be totally exempt from such misuses.
RQ 7 Answer: Covid-related apps do not seem to leak sensitive data; Covid-related apps are not flagged by AntiVirus engines; Covid-related apps are further free of common cryptography API misuses; Overall, the analysis results suggest that Covid-related apps do not present the same security problems as general apps.

Threat to validity
Our study bears a number of threats to validity related to the selection of apps, external factors that may impact our conclusions, and the limitations of the leveraged tools.

Covid-related apps retrieval
We may have missed some Covid-related apps.Our approach relies on simple heuristics to gather applications from AndroZoo, which helped to identify 44 unique apps.Therefore, we cannot guarantee that our app collection of Covid-related apps is exhaustive.However, we leveraged AndroZoo which is the largest and continuously-updated repository of Android applications available to the research community, which we further manually supplemented with other sources.Overall, it is unlikely that we missed a number of Covid-related apps that is large enough to significantly invalidate the conclusions presented in this paper.Actually, as we have revealed, we were able to catch in time some apps that were later removed from the market.
The enforcement of Google Policy on market apps may have biased our study on Covid-related apps.Our experiments were conducted until early June 2020, and we know that Google has been enforcing its policy regarding Covid-related apps since at least early April Google (2020).Given our vantage point without insider knowledge, our observations are limited to apps that were actually released on Google Play, and our study is blind to apps that were never let in Google Play.There is thus a possibility that some of our findings are consequences of Google's policy, and not a characterization of apps that were meant to be on Google Play.Nonetheless, our study reflects closely the landscape of apps that were made available to users.
On the other hand, we note that the removal of apps from the market may have been performed by either market maintainers or by developers themselves.Future studies may attempt to investigate closely the reasons of app removal.
Apk date is unreliable.In our collected dataset, the earliest date of appearance of Covid-related apps goes back to February 2020, but we relied on the earliest date between the AndroZoo added date, the first submission date to VirusTotal, and several other websites (included last update date in Google Play as discussed in Section 4.2).Since AndroZoo does not provide the release date of an app (which is different from the added date), it could be the case that the apps existed before.Unfortunately, we cannot rely on the dates of the files constituting the apps since they can easily be modified (Li et al. 2018).

Reliability of the app description
As discussed in Hypothesis 4 of Section 3.1, for classifying an app as Covidrelated or not, we manually analyzed the Google Play page of the apps.More specifically, we relied on the information available on the app description part of this page.There is a risk that this information could be dishonest, incomplete, or misleading.However, we note that no Covid-related app was flagged by antivirus products, which is an indication that Covid-related apps are unlikely to be malicious, leaving little incentives for apps publishers to provide deceptive information.Furthermore, we note that as long as this potential misinformation is only sporadic, it is very unlikely to significantly alter our taxonomy.Additionally, the fact that the app descriptions were still available on Google Play, suggests that Google did not detect any dishonest information.

Tools limitations
We leveraged several security scanners to evaluate different security characteristics of apps, and hence inherit their limitations.We cannot guarantee that these tools yielded accurate analyses.We mitigated the threats by first ensuring the selection of state of the art tools that are commonly used in the literature, and second by ensuring that we do not overclaim based on their results.

Related Work
Several prior works have conducted empirical studies on large sets of Android applications collected from app markets.Viennot et al. (2014) collected more than a million apps from Google Play and uncovered several interesting patterns in Android apps and the way they are developed.Also in 2014, Gorla et al. (2014) collected apps and their associated descriptions and automatically verified whether the descriptions actually matched the app behaviors while Qu et al. (2014) checked the descriptions against the permission usages.Other works focused on financial apps (Taylor and Martinovic 2017), on app maintenance and prices (Carbunar and Potharaju 2015), on malware (Zhou and Jiang 2012), or on the quality of apps descriptions (Jiang et al. 2014).
To the best of our knowledge, the academic literature has not yet reported on a systematic study on Covid-related Android apps.He et al. (2020) however have recently discussed the case of coronavirus-themed malicious apps.In this paper, the authors analyzed 277 malicious applications related to Covid-19 among an initial dataset of 2016 applications.In our paper, we did not identify any malicious Covid-related apps which seems to be in contradiction with He et al. (2020) who collected 277 malicious Covid-related apps.However, this discrepancy could be explained by at least two reasons: -We used a more selective filtering process to collect Covid-related apps, notably because we realized that the keywords we initially used (and that He et al. ( 2020) also used) were too broad, i.e., they catch many apps not related to Covid-19; -Among their initial dataset of 2016 apps, only 6 are coming from Google Play.Unfortunately, the paper does not precise whether the 6 apps from Google Play are part of their 277 identified malicious apps.Overall, their results show that the vast majority (and probably all, as per our results) of malicious Covid-related apps are coming from application sources that are outside Google Play.
Besides, note that in He et al. (2020) we do not have the information whether they considered different versions of a single app or not.Considering multiple versions could drastically lower the number of apps in their dataset.Their study further shows that there is a correlation between the number of Covid-19 cases in the world and the number of malicious Covid-related apps and that malicious developers did not repackage existing Covid-related apps for releasing malicious apps, contrary to a common malware practice (Zhou and Jiang 2012).He et al. (2020) uncovered 34 different malware families used in malicious Covid-related apps.Their results further suggest that malware developers do not target specific users but target a wide range of countries.
Since the emergence of the Covid-19 topic in the media, researchers are investigating its effect, not only on the medical front but also on our daily life.Related to the security perspective of mobile applications users, several security researchers and analysts publicly disclosed their findings, usually in blog posts or press articles, about the activity of malware developers trying to take advantage of the Covid-19 crisis (Doffman 2020;Buguroo 2020;Saleh 2020;Arsene 2020).
Similarly, from a privacy perspective, a common functionality of Covidrelated apps for fighting against the spread of Covid-19 is contact tracing.Tracing however carries several concerns with respect to user privacy.Indeed, Baumgrtner et al. (2020) show that although developers claim to respect privacy, it is possible to de-anonymize information about infected persons that are traced and even sabotage the tracing effort by injecting fake contacts.

Conclusion
In this paper, we provide a first systematic study of Covid-related Android applications.We collected from different channels 92 apps that are manually vetted as relevant.Then, based on the apps' described goals, our study yields a taxonomy of Covid-related Android apps as a contribution to the literature.Our empirical findings reveal that Covid-related apps have mainly three purposes: (1) inform users, (2) collect data, and (3) provide tooling capabilities for users.After exploring the inner characteristics (e.g., libraries, permissions, size) and the results of security and privacy scanners, we provide first insights into the nature of Covid-related apps.Overall, our empirical study constitutes a first milestone towards understanding who, what, and how Covid-related apps are built.We expect future work in the community to go in-depth into each of the dimensions that we have explored.

A Description of complexity metrics
Below we present five complexity metrics initially described by Chidamber and Kemerer (1994).

A.1 Weighted Methods per Class (WMC)
This metric represents the complexity of individual classes by counting the number of methods per class.Its purpose is to estimate how challenging the development and maintenance of a class are.Also, as the value for a given class grows, the impact on a class inheriting it will be greater as they will inherit all the methods.

A.2 Number Of Children (NOC)
This metric represents the number of immediate sub-classes of a class in the class hierarchy.It measures how many sub-classes will inherit the features (fields, methods, etc.) of the parent class.The greater the number of children, the greater the reuse of properties, meaning that it may require more testing.

A.3 Lack of Cohesion in Methods (LCOM)
This metric represents, for each data field in a class, the average percentage of the methods using that field.It measures the degree of cohesion between methods of a class and data fields of the given class.The lower the cohesion, the greater the complexity as it increases the probability of errors during the development process.

Fig. 3 :
Fig. 3: Distribution of the 750 supposedly Covid-related apps by markets where they were obtained from

Fig. 4 :
Fig. 4: Timeline representing the number of apps with Covid-related keywords in function of time

Fig. 5 :
Fig. 5: Comparison between two Google Play page of two different apps.Left app: Zombie game; right app: Covid-related app.

Fig. 6 :
Fig. 6: Process of our dataset curation.Numbers represent numbers of apps.Dotted boxes represent filters used to refine the dataset.

Fig. 7 :
Fig. 7: Information Broadcast category.The number below a leaf box indicates the number of related apps.

Fig. 10 :
Fig. 10: Number of Covid-related apps by appearance date.The month of June appears hatched to represent the fact that our study was conducted in early June, which means we do not have all apps released in June.

Fig. 12 :
Fig. 12: Size of applications: Comparison between standard apps and Covid-related apps

•
C apk the set of classes names in apk, • M apk the set of method names in apk, • F apk the set of file names in apk, • S apk the set of strings contained apk, We keep apk if at least one element of C apk ∪ M apk ∪ F apk ∪ S apk matches at least one regular expressions listed in Table 1.An app is kept if for instance, a class name contains the substring coronavirus, or if the app contains a string with pandemi as substring.

Table 1 :
Regular expressions used to filter out non Covid-related apps

Table 2 :
First part of Covid-related apps' characteristics retrieved from Google Play apps pages Table 3: Second part of Covid-related apps' characteristics retrieved from Google Play apps pages

Table 4 :
Third part of Covid-related apps' characteristics retrieved from Google Play apps pages

Table 5 :
p-value revealing the significance of the difference between standard apps' and Covid-related apps' number of components.

Table 6 :
Top ten most requested permissions in both standard apps and Covidrelated apps.Percentage indicates the ratio of apps using the permission.
statistical test.It is notable to see that the median value for Covid-related apps is half the median of standard apps, which clearly shows that Covid-related apps request fewer permissions than standard apps.This result suggests that Covid-related apps leverage fewer functionalities from the Android framework.

Table 7 :
Number of Covid-related/standard apps using libraries RQ 3 Answer: Covid-related apps are composed of less GUI-related components than standard apps, and are usually smaller.While overall they use fewer permissions, they tend to use more tracking-related permissions.With respect to libraries, Covid-related apps use common libraries just like any other apps, but significantly fewer advertisement libraries than standard apps.