2.1 Android package structure
2.2 Dex format
2.3 Software birthmarking
2.4 Fuzzy hash – ssdeep
2.5 N-Gram tokenizer
3.1 Opcode sequence extraction
3.2 Fuzzy hashing based on opcode sequence
3.3 N-gram tokenizer sliding window search
4.1 7-Gram clustering result of Dex ssdeep vs. Dexofuzzy
4.2 ‘Operation Blackbird’ sample clustering
4.3 AlienVault OTX reports relationship
4.3.1 Relationship of samples containing similar opcode sequence
4.3.2 Relationship of identical samples in reports
4.3.3 Clustering malware samples using packer
4.3.4 Incorrect clustering caused by excessive usage of SDK
Appendix A: correlation of 74 AlienVault OTX reports
Recently, Android malware has shown a tendency towards large-scale distribution, which reduces the amount of time required for stealing data. Hackers implement such malicious code variants by reusing code. This paper proposes the use of the ‘Dalvik EXecutable Opcode Fuzzy’ (‘Dexofuzzy’) hash, which finds similar malware variants without the need for an analyst to have systematic or mathematical knowledge. Dexofuzzy is a method for generating similarity digests with software birthmarking of opcode sequences in Dex files based on ssdeep. The clustering results are obtained by N-gram tokenizing of the Dexofuzzy hash, which is generated from the malware samples. Observations and experimental results have demonstrated the effectiveness of Dexofuzzy’s similarity search in analysing the Android malware ‘Operation Blackbird’, which had previously been reported by ESTsecurity Security Response Center (ESRC). Finally, we have measured the similarity of the indicators of compromise (IOCs) in 74 reports published by AlienVault OTX and analysed the association between each report in order to attest the efficiency of Dexofuzzy.
Android malware has recently shown a tendency towards large-scale distribution, which reduces the amount of time required for stealing data. Hackers implement such malicious code variants by reusing code. According to a Kaspersky report , the number of incidences of Android malware in 2018 doubled compared with the previous year, despite the presence of security enhancements in the Android operating system [2, 3].
Many studies based on static, dynamic and hybrid analysis [4, 5] have been carried out in order to find ways to better protect Android devices and users from malware threats. Machine learning and similarity algorithm-based methods have been applied to the previous analysis data [6, 7, 8]. Machine learning-based detection methods enable the clustering and classifying of multiple malware samples, but there is a usage limit on machine learning: high entry barriers make it difficult for general analysts to use such methods with mathematical and statistical approaches and there are considerable costs for the preprocessing and learning of datasets. In contrast, detection methods based on similarity algorithms can evaluate the characteristics of malware samples without the need for additional learning processes. Numerous similarity algorithms have been proposed to improve performance . Although these algorithms require a hash index table for the similarity search, from the point of view of the analyst, the use of similarity algorithms is a much more straightforward method than machine learning.
In particular, the similarity digests algorithm is simple and fast and can be used to compare the similarity of form or objective among multiple samples. Searching for similar malicious codes requires minimum effort, which enables us to cluster and classify a large number of variant malicious codes. Although similarity digests have a limitation in inferring malignancy compared to existing analysis methods, it is simple and quick to cluster, classify and compare the large-scale samples. Because of these advantages, security researchers have introduced analytical methodologies using similarity digests for searching for similar samples – for example, Trend Micro proposed a similarity digest using the Locality Sensitive Hash (LSH), called TLSH , and JPCERT proposed Impfuzzy, which calculates an ssdeep hash of the import table of a given PE file .
In this paper, we propose the Dalvik EXecutable Opcode Fuzzy (Dexofuzzy) hashing method, which enables analysts to search and classify Android malware variants in a lightweight way. To do this, first we parse a Dex file in APK (which stands for Android Package and compressed by zip) to extract opcode. Then we calculate the ssdeep value of the opcode generated from the previous step, which is named Dexofuzzy. To search malware variants, we tokenize the hashes generated by Dexofuzzy using N-gram, and then index them for a fast search.
This paper is organized as follows: Section 2 explains the techniques that are used in Dexofuzzy. Section 3 describes a method to search similar malware samples using Dexofuzzy extracted from opcode. In section 4, we verify the performance of Dexofuzzy, in which the malware samples are properly clustered. Finally, section 5 summarizes the results from each part and concludes by highlighting the significance of the methodology that we have proposed in this paper.
Android files have an APK extension and utilize the ZIP file format. The most essential components of the APK are AndroidManifest.xml, classes.dex, and resources.arsc. AndroidManifest.xml is an encoded XML file containing information about attributes and permissions of the app, and resources.arsc is an encoded file containing resource information about the app. In order to run the app, classes.dex, known as the Dex file, is used, which contains opcode and is used to generate Dexofuzzy. Figure 1 illustrates the structure of the APK file.
The Dex file is the most important file that contains the compiled code that runs on Android. When decompiling the Dex file (a.k.a. Baksmaling), it returns bytecode called Smali. The structure of the Dex file is shown in Figure 2.
In the Dex file structure, Class Defs is a table that defines code information of Android files by Class. Class Data Offset is a set of bytecode offsets except for attribute information of class . We use a method to parse the Class Data Offset of the Dex file to extract the opcode sequence.
Software birthmarking is a promising technique used for the detection of software theft based on the use of unique characteristics to identify programs. A number of studies have been conducted to detect repackaged Android apps [13, 14, 15]. G. Myles et al. suggested a software birthmark using opcode-level N-gram . In this paper, we use the software birthmark method on the opcode of Dex files to detect Android malware variants that reuse existing malicious code.
The Context Triggered Piecewise Hashes (CTPH) algorithm, proposed by J. Komblum, is a method that generates a similarity digest by dividing a file into segments using the Rolling Hash method, unlike the existing hashing method that guarantees integrity . CTPH is designed to identify the similarity of data, and is currently used for analysing malicious code .
N-gram is a contiguous sequence of n items from a given sample of text. The method is widely used for measuring the similarity of sentences in search engines, big data, and computer forensics. B. Wallace presented the results of clustering malware variants by hashing the malware samples to ssdeep and tokenizing them with 7-gram, and proposed the use of the N-Gram tokenizer for searching N-gram as one of the ways to improve the performance of ssdeep . Elasticsearch is a search engine that provides a distributed, multitenant-capable, full-text search. Because it supports N-gram tokenizer, it is highly effective for similarity searches .
This section describes how Dexofuzzy works to measure the similarity between Android malware families. Dexofuzzy parses the Dex file inside the Android file to extract the opcode, and the extracted features are used to generate the similarity digests using ssdeep. The features are also used to search for similar malicious code by tokenizing the generated Dexofuzzy to N-gram size and using the sliding window technique. Figure 4 is a schematic illustration of how Dexofuzzy is created.
The classes.dex file, which is the core file required to run an Android application, is decompiled into Dalvik bytecode using APKTool . We will use the opcode defined in Dalvik bytecode for the similarity search. In this paper, opcode has been selected because the opcode sequence will remain unchanged in spite of overall changes in the Dalvik bytecode, even in the case of changing the Package Name, Class Name and Method Name and adding one or more variables in reusable codes. Figure 5 is a representative example to show that Dalvik bytecodes are changed (except for the opcode sequence) in two identical malware samples by the addition of variables in the Method and by changing the Package Name. Therefore, opcode is the most powerful characteristic in the regard that the original remains unchanged despite changes in other elements when measuring the similarity of malware variants implemented by reusing codes.
Figure 6 displays the process of extracting the opcode sequence from each method into Hex values by parsing the classes.dex file. The extracted opcode sequence is bundled into a single string that stores the opcode sequence of the Dex file in list form. Since such an opcode sequence string has unique characteristics of each method, the malware variants reusing the code contain a large number of opcode sequence strings that match in the list, compared to malware which is not implemented by reusing code.
As shown in Figure 7, the extracted opcode sequence string list performs the first stage of fuzzy hashing for each string. Next, the fuzzy-hashed Methodfuzzy list is fuzzy-hashed again in the second stage to generate Dexofuzzy.
The ssdeep format is composed of BlockSize:Signature1:Signature2, where BlockSize is the minimum block size for rolling hashing. Signature1 is a signature generated by rolling hashing based on previously calculated block size, and Signature2 is a signature generated by rolling hashing with twice the block size. Dexofuzzy is a similarity digest generated based on ssdeep, which enables us simply to compare the signatures using the function ssdeep compare. Figure 8 shows the result of comparing two Android malware variants with Dexofuzzy.
>>> import Ssdeep
>>> Ssdeep.compare("48:U7uPrEMc0HZj0/zeGnD2KmUCNc2FuGgy9fY:UHMHZ4/zeGD2+Cap3y9Q", "48:U7uPrEMc0HZj0/zeGnV2KmDmUCNc9tzFuGgLM/xBDyw9V4XIU0pt:UHMHZ4/zeGV2lCa43LM5B/H7U0pt")
B. Wallace has confirmed that Signature1 is more efficient than Signature2 for searching malware by performing a test to compare the two signatures with 7-gram Slice with ssdeep . Accordingly, we use Signature1 to search for similar malware variants in this paper.
We use Elasticsearch to search for Dexofuzzy by N-gram. Elasticsearch basically provides an N-gram tokenizer and indexes all generated N-length tokens. Figure 9 describes the process of tokenizing Dexofuzzy into 7-grams to detect the indexed seven‑character strings with the sliding window technique.
In addition, Elasticsearch is useful for calculating the optimal thresholding value to increase the true positive rate and to reduce the false negative rate of each sample since the method enables the number of M to be specified, where the N-gram strings are matched, as a parameter when searching. Figures 10 and 11 illustrate how to search for a similar Dexofuzzy by setting N-gram and Minimum Should Match (from Elasticsearch settings) in a single Dexofuzzy sample.
This chapter presents the results of clustering and performance measurements on malware variants that are active in the wild, with the generated Dexofuzzy. We collected 212,955 malware samples and created a list of the ssdeep and Dexofuzzy hash values of a Dex file of the APK respectively, then searched samples using the Elasticsearch 7-gram tokenizer and clustered the result values in order to assess the performance of Dexofuzzy.
In this section, we will also demonstrate the effectiveness of the Dexofuzzy similarity search using the visualization tool Gephi  to analyse the malware ‘Operation Blackbird’, which has previously been reported. Finally, we collected the malware samples based on indicators of compromise (IOCs) in AlienVault OTX reports published by multiple vendors and then created Dexofuzzy and clustered it in order to explain the relationship between the reports. Table 1 shows the configuration of the environment for the experiment.
|CPU||Intel(R) Core(TM) i5-7500 CPU @ 3.40GHz|
|OS||Ubuntu 18.08 LTS|
|Search engine||Elasticsearch 7.2|
We clustered the similarity digest hashes of the Dex files, which were computed using Dex ssdeep and Dexofuzzy respectively, with 431 samples detected as Trojan.Android.KRBanker, to evaluate the performance of the two similarity clustering algorithms. Figure 12 shows the clustering results of Dex ssdeep and Dexofuzzy by using the 7-gram search.
The experimental results show that Dex ssdeep hashes are clustered into 144 groups and 21 groups are clustered by Dexofuzzy – the clustering size of Dexofuzzy is seven times smaller than that of ssdeep, which means that Dexofuzzy is more effective than ssdeep. Next, we conducted an experiment to search for multiple Dexofuzzy hashes simultaneously to measure the search performance of the Elasticsearch tokenizer. Figure 13 shows the operation time of Dexofuzzy when searching 212,955 samples at the same time.
In the following experiment, we searched for similar variants using 7-gram tokens of Dexofuzzy generated in 10 malware samples from the IOCs  of ‘Operation Blackbird’ reported by ESTsecurity. Figure 14 shows the clustering result of the additionally found malware variants among 212,955 samples.
A total of eight variants (excluding the original malware) were found from the 10 malware samples from the IOCs  of the ‘Operation Blackbird’ report. Table 2 presents the MD5s of the searched malware samples and the detected Dexofuzzy 7-gram tokens, which are highlighted in bold.
We have analysed the samples in Clustering D based on the IOC (MD5: 20a274cbc057bd2035961af97724b70c), which is indicated in the report ‘Operation Blackbird’, to identify the similarity of the samples from the clustering results. Table 3 shows the MD5 and Dexofuzzy hashes which are created from samples in ‘Operation Blackbird’ and Clustering D, and Figure 15 describes the class structure of three malware samples.
|Operation Blackbird IOC||20a274cbc057bd2035961af97724b70c||12:H7Oe0yovJkUTSX9Xh75L7xZ2PeLLLLLLL0NEN/
Analysing three malware samples revealed that the same source code was partially reused in ‘20a274cbc057bd2035961af97724b70c’ and ‘c919f72a8a0a64edd6a68dfe20e6bb36’, and that ‘aecf1472bd8a061fd0fdd0722b841ee0’ was created by adding the logic to specify the malicious behaviours. Figure 16 illustrates the logic added in malware samples.
Finally, we have investigated the correlation using Dexofuzzy based on 2,494 IOCs extracted from 74 AlienVault OTX reports released by multiple vendors. Appendix A represents the correlation of 74 AlienVault OTX reports. We have selected 465 IOCs from 15 of the AlienVault OTX reports and clustered the samples to analyse the correlation between the reports in this experiment. Figure 17 shows that there are four types of relationship in the results of analysing the 15 reports, which were clustered into six groups.
The first type is the relationship, in which opcode sequence is similar, clustering nine reports into four reports, as shown in Figure 18.
Tables 4 to 7 show clustering results (MD5, Dexofuzzy) of the set of malware samples containing a similar opcode sequence. The samples in each report are clustered by similar logic, and the clustered reports are correlated according to the themes.
|Android banking malware masquerades as Flash Player||78c2444fe15a8e58c629076781d9442a||96:VsMRXFTmw+cFcpF1THYb6Hi4W493J05zRo/+
|MARCHER GETS CLOSE TO USERS BY TARGETING MOBILE BANKING - ANDROID APPS - SOCIAL MEDIA - AND EMAIL||140687aa4d4fc70175c7df1d737d5515||96:5yMzj+xFTmXrcupF1THvb6Hi47493J05zMFom
|Monero-Mining HiddenMiner Android Malware Can Potentially Cause Device Failure||530bd6c95c3a79c04f49880a44c348db||384:H3Ddbm8whYHVxIiKXB+O51BVapknL+
|CoinMiner and other malicious cryptominers targeting Android||766055b991805fe8ef0a1c96643a98a1||384:H3Ddbm8whYHVxIiKXB+O51BVapknL
|Operation C-Major actors used Mobile Spyware Against Targets||11ba93d968bd96e9e9c9418ea1fdcbbc||384:qK9ydh5ix2PaQdyUrrrr6CpXX5asWMou
|Group5: Syria and the Iranian Connection||8ebeb3f91cda8e985a9c61beb8cdde9d||384:qK9ydh5ix2PaQdyUrrrr6CpXX5asWMou
|Investigating a Libyan Cyber Espionage Campaign Targeting High-Profile Influentials||93ebc337c5fe4794d33df155986a284d||384:qK9ydh5ix2PaQdyUrrrr9CpXRgasWMou
|KevDroid: New Android Malware||56b1f4800fa0e083caf0526c3de26059||3072:KEKQgvB+ZNXB/ZTUKvcb0crwIbRo
|Reaper Group Updated Mobile Arsenal||d6abaa07f7e525153116c98412115b2e||3072:EEKQgvB+ZNXB/ZTUKvcb0crwIbRkT
The second is the group of samples including the same MD5, which clusters similar samples, leading to the result that there is a relationship among the analysed reports. Figure 19 shows the relationship between the reports, in which similar malware samples are clustered.
Table 8 describes two reports where similar Telegram malware was examined, enabling us to clearly identify if those samples are similar using Dexofuzzy.
|TeleRAT: Another Android Trojan Leveraging Telegrams Bot API to Target Iranian Users||9d23f7688a82d487a8bb87df19cb2426||768:L6T++XMpGVznLwCtyaipLbV8wyNN7HN
|Android Trojan controlled via Telegram spies on Iranian users||34be73f9fdccc152530f2d6cc26cc640||413,917|
Table 9 shows that there are eight malware samples containing the same SHA-256 in the reports ‘Kemoge: Another Mobile Malicious Adware Infecting Android’ and ‘GOOLIGAN MORE THAN A MILLION GOOGLE ACCOUNTS BREACHED’. Additionally, three more similar samples were found in the IOCs of the report ‘GOOLIGAN MORE THAN A MILLION GOOGLE ACCOUNTS BREACHED’ based on the identified samples.
|Kemoge: Another Mobile Malicious Adware Infecting Android
GOOLIGAN MORE THAN A MILLION GOOGLE ACCOUNTS BREACHED
|GOOLIGAN MORE THAN A MILLION GOOGLE ACCOUNTS BREACHED||5b446ec92f1cf0a2a06fbe66a95a6c89||3072:2+y4CDTN0Qw7Nw968CcGpunZZUWj
The third type clusters malware samples that are packed using the same logic. In this case, however, there is no correlation between the reports. Figure 20 shows the clustering results of the malware samples that are packed by similar packers Qihoo and Tencent.
Figure 21 displays the clustered malware samples using the packer Qihoo, and Table 10 indicates the details of the five clustered malware samples from the reports from ‘GOOLIGAN MORE THAN A MILLION GOOGLE ACCOUNTS BREACHED’ and ‘A Whale of a Tale: HummingBad Returns.’
|GOOLIGAN MORE THAN A MILLION GOOGLE ACCOUNTS BREACHED||eda506a6c01c3c7e149ebaebcf929c40||6:z9m3KnKA4fdQV/EughUtDK1LAYDK4HSM
|A Whale of a Tale: HummingBad Returns||b5103298638ec324923422559d3ace55||6:zLMXv56Pikl7KnKztsk4fdQV/PfwVrUtDsqxhY
The 28 samples in the reports ‘A Whale of a Tale: HummingBad Returns’ and ‘New Rootnik Variant’ were clustered by the similarity of the logic used in the Tencent packer. Figure 22 presents the result of analysing the class structure of four representative malicious codes from each cluster based on Dexofuzzy, and Table 11 shows a list of malware samples packed by Tencent using 7-gram search of Dexofuzzy.
|A Whale of a Tale: HummingBad Returns||5f512bf1f51141d4201dcfe819dc2165||84:9IIiCtwXLNRtwkqrEEDjyqJP21bS8FhYLZ
|New Rootnik Variant||fc2b5e892ce00df128545247ddd9d104||384:QIIdVCtwXLNRtwkqrEEDjyqJP21bS8FhYL
The fourth cluster of malware samples contains more SDK opcode sequence than malicious opcode sequence due to excessive usage of SDK, as shown in Figure 23.
Figure 24 shows an example of two samples which are incorrectly clustered due to the reason mentioned above, and Table 12 details the two samples that are detected by the 7-gram search of Dexofuzzy. As the 7-gram search of Dexofuzzy is based on the similarity of opcode sequence, malware samples are clustered according to the SDK usage rather than the logic similarity in case the proportion of malicious logic is relatively lower than that of SDK usage.
|GOOLIGAN MORE THAN A MILLION GOOGLE ACCOUNTS BREACHED||1634b1fb3b353019e9d3b7b3d21507ab||768:XgwscmCnqs+13uXFAF5UPgoj8iSfU5Ngk5
|A Whale of a Tale: HummingBad Returns||9e099645a13a339f83af08941db40056||768:BJlKnhSX7RF96rXlOaLQWmjI0T21utTj8i0
As Android malware has shown a tendency toward large-scale distribution, users need more efficient and reliable analysis information to cope with the rapidly changing security environment. This paper proposes the methodology ‘Dexofuzzy’ to overcome the limitations of hardware resources or analysis environment. Dexofuzzy measures the similarity among a large number of Android malware variants by using opcode of the Dex file. In addition, Dexofuzzy utilizes Elasticsearch’s N-gram tokenizer, making it easier to perform a quick analysis of similarities.
In this paper, we have used Dexofuzzy to figure out the correlation based on 2,494 IOCs from 74 AlienVault OTX reports written by various vendors (see Appendix) and identified the correlation by mainly utilizing 15 reports and 465 samples among those reports. In addition, the eight more unknown samples were found among the 212,955 samples by utilizing the similarity search ‘Dexofuzzy’ with the IOCs of the report ‘Operation Blackbird.’ However, Dexofuzzy has limitations in that samples are incorrectly clustered when they are packed with the same packer or the proportion of SDK usage is excessive. It is recommended to use existing static and dynamic analysis methods along with Dexofuzzy to overcome these limitations, which helps us to more effectively respond to the large and rapidly evolving sets of Android-based malware. This tool is an open-source project that can be downloaded from the GitHub repository  and is also provided in the Pypi repository .
 Mobile malware evolution 2018. https://securelist.com/mobile-malware-evolution-2018/89689/. Cited May, 2019.
 Cyber attacks on Android devices on the rise. https://www.gdatasoftware.com/blog/2018/11/31255-cyber-attacks-on-android-devices-on-the-rise. Cited May, 2019.
 2018 Mobile Threat Landscape. https://www.trendmicro.com/vinfo/au/security/research-and-analysis/threat-reports/roundup/2018-mobile-threat-landscape. Cited May, 2019.
 Faruki, P.; Bharmal, A.; Laxmi, V.; Ganmoor, V.; Gaur, M. S.; Conti, M.; Rajarajan, M. Android security: A survey of issues, malware penetration, and defenses. IEEE Commun. Surveys Tutorials, vol. 17, no. 2, pp. 998–1022, Mar.–Apr. 2015.
 Yu, B.; Fang, Y.; Yang, Q.; Tang, Y.; Liu, L. et al. A survey of malware behavior description and analysis. Frontiers of Information Technology and Electronic Engineering, Volume 19(5), 583-603, 2018.
 Classifying Malware using Import API and Fuzzy Hashing – impfuzzy. https://blogs.jpcert.or.jp/en/2016/05/classifying-mal-a988.html. Cited May, 2019.
 Android Runtime (ART) and Dalvik. https://source.android.com/devices/tech/dalvik. Cited May, 2019.
 Zhou, W.; Zhou, Y.; Jiang, X.; Ning, P. Detecting Repackaged Smartphone Applications in Third-Party Android Marketplaces. In Proceedings of the 2nd ACM Conference on Data and Application Security and Privacy (CODASPY), 2012.
 Ko, J.; Shim, H.; Kim, D.; Jeong, Y.-S.; Cho, S.-j.; Park, M.; Han, S.; Kim, S. B. Measuring similarity of Android applications via reversing and k-gram birthmarking. In Proc. of the 2013 Research in Adaptive and Convergent Systems (RACS’13), Montreal, Quebec, Canada. ACM, October 2013, pp.336–341.
 Zhang, F.; Huang, H.; Zhu, S.; Wu, D.; Liu, P. Viewdroid: Towards obfuscation-resilient mobile application repackaging detection. In Proc. 7th ACM Conf. Secur. Privacy Wireless Mobile Netw., 2014, pp.25–36.
 Intezer Community Tip: How to Optimize ssdeep Comparisons with ElasticSearch. https://www.intezer.com/intezer-community-tip-ssdeep-comparisons-with-elasticsearch. Cited May, 2019.
 Apktool. https://ibotpeaches.github.io/Apktool. Cited Jul., 2019.
 ‘Operation Blackbird’. https://blog.alyac.co.kr/2035. Cited Jul., 2019.
 Open Source Project – ‘Dexofuzzy: Dalvik EXecutable Opcode Fuzzyhash’. https://github.com/ESTsecurity/Dexofuzzy.
 Python Package Index – ‘Dexofuzzy: Dalvik EXecutable Opcode Fuzzyhash’. https://pypi.org/project/dexofuzzy.
 StealJob: New Android Malware Used by Donot APT Group. https://ti.360.net/blog/articles/stealjob-new-android-malware-used-by-donot-apt-group-en/. Cited Jul., 2019.
 Gustuff banking botnet targets Australia. https://blog.talosintelligence.com/2019/04/gustuff-targets-australia.html. Cited Jul., 2019.
 Exodus: New Android Spyware Made in Italy. https://securitywithoutborders.org/blog/2019/03/29/exodus.html. Cited Jul., 2019.
 KBuster: Fake Bank App in South Korean. https://ti.360.net/blog/articles/kbuster-fake-bank-app-in-south-korean-en/. Cited Jul., 2019.
 MalBus: Popular South Korean Bus App Series in Google Play Found Dropping Malware After 5 Years of Development. https://securingtomorrow.mcafee.com/other-blogs/mcafee-labs/malbus-popular-south-korean-bus-app-series-in-google-play-found-dropping-malware-after-5-years-of-development/. Cited Jul., 2019.
 Google Play Apps Drop Anubis Banking Malware, Use Motion-based Evasion Tactics. https://blog.trendmicro.com/trendlabs-security-intelligence/google-play-apps-drop-anubis-banking-malware-use-motion-based-evasion-tactics/. Cited Jul., 2019.
 Android Trojan steals money from PayPal accounts even with 2FA on. https://www.welivesecurity.com/2018/12/11/android-trojan-steals-money-paypal-accounts-2fa/. Cited Jul., 2019.
 Gplayed’s younger brother is a banker and its after Russian banks. https://blog.talosintelligence.com/2018/10/gplayerbanker.html. Cited Jul., 2019.
 Roaming Mantis, part III: iOS crypto-mining and spreading via malicious content delivery system. https://securelist.com/roaming-mantis-part-3/88071/. Cited Jul., 2019.
 BusyGasper - the unfriendly spy. https://securelist.com/busygasper-the-unfriendly-spy/87627/. Cited Jul., 2019.
 Android/BondPath: a Mature Spyware. https://www.fortinet.com/blog/threat-research/android-bondpath--a-mature-spyware.html. Cited Jul., 2019.
 Anubis is Back: Are You Prepared? https://news.sophos.com/en-us/2018/08/14/anubis-is-back-are-you-prepared/. Cited Jul., 2019.
 DOKKAEBI: Documents of Korean and Evil Binary. https://www.virusbulletin.com/conference/vb2018/abstracts/dokkaebi-documents-korean-and-evil-binary/. Cited Jul., 2019.
 Infrastructure and Samples of Hamas’ Android Malware Targeting Israeli Soldiers. https://www.clearskysec.com/glancelove/. Cited Jul., 2019.
 CoinMiner and other malicious cryptominers targeting Android. https://www.sophos.com/en-us/medialibrary/PDFs/technical-papers/sophos-coinminer-and-other-malicious-cryptominers-tpna.aspx. Cited Jul., 2019.
 Reaper Group’s Updated Mobile Arsenal. https://researchcenter.paloaltonetworks.com/2018/04/unit42-reaper-groups-updated-mobile-arsenal/. Cited Jul., 2019.
 Pocket cryptofarms: Investigating mobile apps for hidden mining. https://securelist.com/pocket-cryptofarms/85137/. Cited Jul., 2019.
 Fake AV Investigation Unearths KevDroid. New Android Malware. https://blog.talosintelligence.com/2018/04/fake-av-investigation-unearths-kevdroid.html. Cited Jul., 2019.
 Monero-Mining HiddenMiner Android Malware Can Potentially Cause Device Failure. https://blog.trendmicro.com/trendlabs-security-intelligence/monero-mining-hiddenminer-android-malware-can-potentially-cause-device-failure/. Cited Jul., 2019.
 TeleRAT: Another Android Trojan Leveraging Telegrams Bot API to Target Iranian Users. https://researchcenter.paloaltonetworks.com/2018/03/unit42-telerat-another-android-trojan-leveraging-telegrams-bot-api-to-target-iranian-users/. Cited Jul., 2019.
 Downloaders on Google Play spreading malware to steal Facebook login details. https://blog.avast.com/downloaders-on-google-play-spreading-malware-to-steal-facebook-login-details. Cited Jul., 2019.
 New version of mobile malware Catelites possibly linked to Cron cyber gang. https://blog.avast.com/new-version-of-mobile-malware-catelites-possibly-linked-to-cron-cyber-gang. Cited Jul., 2019.
 BankBot Found on Google Play and Targets Ten New UAE Banking Apps. http://blog.trendmicro.com/trendlabs-security-intelligence/bankbot-found-google-play-targets-ten-new-uae-banking-apps/. Cited Jul., 2019.
 WAP-billing Trojan-Clickers on rise. https://securelist.com/wap-billing-trojan-clickers-on-rise/81576/. Cited Jul., 2019.
 New WannaCry-Mimicking SLocker Abuses QQ Services. http://blog.trendmicro.com/trendlabs-security-intelligence/new-wannacry-mimicking-slocker-abuses-qq-services/?linkId=40489549. Cited Jul., 2019.
 Unmasking Android Malware: A Deep Dive into a New Rootnik Variant, Part III. http://blog.fortinet.com/2017/07/09/unmasking-android-malware-a-deep-dive-into-a-new-rootnik-variant-part-iii. Cited Jul., 2019.
 LeakerLocker: Mobile Ransomware Acts Without Encryption. https://securingtomorrow.mcafee.com/other-blogs/mcafee-labs/leakerlocker-mobile-ransomware-acts-without-encryption/. Cited Jul., 2019.
 SpyDealer: Android Trojan Spying on More Than 40 Apps. https://unit42.paloaltonetworks.com/unit42-spydealer-android-trojan-spying-40-apps/. Cited Jul., 2019.
 CopyCay: AN IN-DEPTH ANALYSIS OF THE COPYCAT ANDROID MALWARE CAMPAIGN. https://www.checkpoint.com/downloads/resources/copycat-research-report.pdf. Cited Jul., 2019.
 SLocker Mobile Ransomware Starts Mimicking WannaCry. http://blog.trendmicro.com/trendlabs-security-intelligence/slocker-mobile-ransomware-starts-mimicking-wannacry/. Cited Jul., 2019.
 Koler Android Ransomware Targets the US with Fake PornHub Apps. https://www.bleepingcomputer.com/news/security/koler-android-ransomware-targets-the-us-with-fake-pornhub-apps/. Cited Jul., 2019.
 Spyware targets Iranian Android users by abusing messaging app Telegram’s Bot API. https://blog.avast.com/spyware-targets-iranian-android-users-by-abusing-messaging-app-telegram-bot-api#hs_cos_wrapper_blog_comments. Cited Jul., 2019.
 Analyzing Xavier: An InformationStealing Ad Library on Android. https://documents.trendmicro.com/assets/appendix--analyzing-xavier-an-information-stealing-ad-library-on-android.pdf. Cited Jul., 2019.
 Dvmap: the first Android malware with code injection. https://securelist.com/78648/dvmap-the-first-android-malware-with-code-injection/. Cited Jul., 2019.
 Marcher Gets Close to Users by Targeting Mobile Banking, Android Apps, Social Media, and Email. https://f5.com/labs/articles/threat-intelligence/malware/marcher-gets-close-to-users-by-targeting-mobile-banking-android-apps-social-media-and-email-26004. Cited Jul., 2019.
 DressCode Android Malware Finds Apparent Successor in MilkyDoor. http://blog.trendmicro.com/trendlabs-security-intelligence/dresscode-android-malware-finds-successor-milkydoor/. Cited Jul., 2019.
 Swearing Trojan Continues to Rage, Even After Authors’ Arrest. https://blog.checkpoint.com/2017/03/21/swearing-trojan-continues-rage-even-authors-arrest/. Cited Jul., 2019.
 Google Play Apps Infected with Malicious IFrames. http://researchcenter.paloaltonetworks.com/2017/03/unit42-google-play-apps-infected-malicious-iframes/. Cited Jul., 2019.
 Breaking The Weakest Link Of The Strongest Chain. https://securelist.com/blog/incidents/77562/breaking-the-weakest-link-of-the-strongest-chain/. Cited Jul., 2019.
 Deep Analysis of Android Rootnik Malware Using Advanced Anti-Debug and Anti-Hook, Part II: Analysis of The Scope of Java. http://blog.fortinet.com/2017/01/26/deep-analysis-of-android-rootnik-malware-using-advanced-anti-debug-and-anti-hook-part-ii-analysis-of-the-scope-of-java. Cited Jul., 2019.
 A Whale of a Tale: HummingBad Returns. http://blog.checkpoint.com/2017/01/23/hummingbad-returns/. Cited Jul., 2019.
 Switcher: Android joins the ‘attack-the-router’ club. https://securelist.com/switcher-android-joins-the-attack-the-router-club/76969/. Cited Jul., 2019.
 Fake Apps Take Advantage of Super Mario Run Release. http://blog.trendmicro.com/trendlabs-security-intelligence/fake-apps-take-advantage-mario-run-release/. Cited Jul., 2019.
 Comodo Threat Research Labs Warns Android Users of ‘Tordow v2.0’ outbreak. https://blog.comodo.com/comodo-news/comodo-warns-android-users-of-tordow-v2-0-outbreak/. Cited Jul., 2019.
 GOOLIGAN: MORE THAN A MILLION GOOGLE ACCOUNTS BREACHED. http://blog.checkpoint.com/wp-content/uploads/2016/12/Gooligan-Research-Report.pdf. Cited Jul., 2019.
 PluginPhantom: New Android Trojan Abuses ‘DroidPlugin’ Framework. http://researchcenter.paloaltonetworks.com/2016/11/unit42-pluginphantom-new-android-trojan-abuses-droidplugin-framework/. Cited Jul., 2019.
 HackingTeam back for your Androids, now extra insecure! http://rednaga.io/2016/11/14/hackingteam_back_for_your_androids/. Cited Jul., 2019.
 Exaspy - Commodity Android Spyware Targeting High-level Executives. https://www.symantec.com/connect/blogs/exaspy-commodity-android-spyware-targeting-high-level-executives. Cited Jul., 2019.
 Android banking malware masquerades as Flash Player, targeting large banks and popular social media apps. https://blog.fortinet.com/2016/11/01/android-banking-malware-masquerades-as-flash-player-targeting-large-banks-and-popular-social-media-apps. Cited Jul., 2019.
 BITTER: a targeted attack against Pakistan. https://www.forcepoint.com/blog/security-labs/bitter-targeted-attack-against-pakistan. Cited Jul., 2019.
 DressCode and its Potential Impact for Enterprises. http://blog.trendmicro.com/trendlabs-security-intelligence/dresscode-potential-impact-enterprises/. Cited Jul., 2019.
 Investigating a Libyan Cyber Espionage Campaign Targeting High-Profile Influentials. https://cyberkov.com/wp-content/uploads/2016/09/Hunting-Libyan-Scorpions-EN.pdf. Cited Jul., 2019.
 Four spyware apps removed from Google Play. https://blog.lookout.com/blog/2016/09/16/embassy-spyware-google-play/. Cited Jul., 2019.
 First Twitter-controlled Android botnet discovered. http://www.welivesecurity.com/2016/08/24/first-twitter-controlled-android-botnet-discovered/. Cited Jul., 2019.
 Group5: Syria and the Iranian Connection. https://citizenlab.ca/2016/08/group5-syria/. Cited Jul., 2019.
 SpyNote Android Trojan Builder Leaked. http://researchcenter.paloaltonetworks.com/2016/07/unit42-spynote-android-trojan-builder-leaked/. Cited Jul., 2019.
 DroidJack Uses Side-Load…It’s Super Effective! Backdoored Pokemon GO Android App Found. https://www.proofpoint.com/us/threat-insight/post/droidjack-uses-side-load-backdoored-pokemon-go-android-app. Cited Jul., 2019.
 From HummingBad to Worse. http://blog.checkpoint.com/wp-content/uploads/2016/07/HummingBad-Research-report_FINAL-62916.pdf. Cited Jul., 2019.
 Android Malware Clicker.G!Gen Found on Google Play. https://blogs.mcafee.com/mcafee-labs/android-malware-clicker-dgen-found-google-play/. Cited Jul., 2019.
 ‘Operation C-Major’ Actors Also Used Android, BlackBerry Mobile Spyware Against Targets. http://blog.trendmicro.com/trendlabs-security-intelligence/operation-c-major-actors-also-used-android-blackberry-mobile-spyware-targets/. Cited Jul., 2019.
 New Android Trojan ‘Xbot’ Phishes Credit Cards and Bank Accounts, Encrypts Devices for Ransom. http://researchcenter.paloaltonetworks.com/2016/02/new-android-trojan-xbot-phishes-credit-cards-and-bank-accounts-encrypts-devices-for-ransom/. Cited Jul., 2019.
 Android.Bankosy: All ears on voice call-based 2FA. http://www.symantec.com/connect/blogs/androidbankosy-all-ears-voice-call-based-2fa. Cited Jul., 2019.
 Android.ZBot banking Trojan uses ‘web injections’ to steal confidential data. http://news.drweb.com/show/?i=9754&lng=en&c=14. Cited Jul., 2019.
 Pornographic-themed Malware Hits Android Users in China, Taiwan, Japan. http://blog.trendmicro.com/trendlabs-security-intelligence/pornographic-themed-malware-hits-android-users-in-china-taiwan-japan/. Cited Jul., 2019.
 Chinese Taomike Monetization Library Steals SMS Messages. http://researchcenter.paloaltonetworks.com/2015/10/chinese-taomike-monetization-library-steals-sms-messages/. Cited Jul., 2019.
 The Postal Group. https://www.cert.pl/wp-content/uploads/2015/12/The_Postal_Group.pdf. Cited Jul., 2019.
 Kemoge: Another Mobile Malicious Adware Infecting Over 20 Countries. https://www.fireeye.com/blog/threat-research/2015/10/kemoge_another_mobi.html. Cited Jul., 2019.
 Two Games Released in Google Play Can Root Android Devices. http://blog.trendmicro.com/trendlabs-security-intelligence/two-games-released-in-google-play-can-root-android-devices/. Cited Jul., 2019.
 New ‘Ghost Push’ Variants Sport Guard Code; Malware Creator Published Over 600 Bad Android Apps. http://blog.trendmicro.com/trendlabs-security-intelligence/new-ghost-push-variants-sport-guard-code-malware-creator-published-over-600-bad-android-apps/. Cited Jul., 2019.
 Android trojan drops in, despite Google Bouncer. http://www.welivesecurity.com/2015/09/22/android-trojan-drops-in-despite-googles-bouncer/. Cited Jul., 2019.
 Locker: an Android ransomware full of surprises. https://www.fortinet.com/blog/threat-research/locker-an-android-ransomware-full-of-surprises.html. Cited Jul., 2019.
 Porn clicker keeps infecting apps on Google Play. http://www.welivesecurity.com/2015/07/23/porn-clicker-keeps-infecting-apps-on-google-play/. Cited Jul., 2019.
 Attack of the 90s Kids: Chinese Teens Take On the Mobile Ransomware Trade. http://blog.trendmicro.com/trendlabs-security-intelligence/attack-of-the-90s-kids-chinese-teens-take-on-the-mobile-ransomware-trade/. Cited Jul., 2019.
 COOLREAPER: The Coolpad Backdoor. https://www.paloaltonetworks.com/content/dam/paloaltonetworks-com/en_US/assets/pdf/reports/Unit_42/unit42-cool-reaper.pdf. Cited Jul., 2019.