Class Overview

 This class gave me the opportunity to try out some data analytics on real work data. We had some assignments which used the programming language R. Which I had not heard of but it seems to be mainly a language for data analysis and statistics. The language was very high level and unopinionated. Using it I was able to generate many charts using important statistics functions. I took AP Statistics in high school so it had been a while since I had used any of them before.

The Final

 I learned a lot in the class about data and the final put a focus on security. For the Final we were to be put into teams. Our task was to create an interesting analysis about data from a real world security scan. After being placed on a team of five we all gathered into a discussion on how to approach this problem. We were provided with a large JSON file which had a report of the security findings of the top downloaded app stores. From that my team was able to find there was several approaches we could take. We ended up following three different hypothesis.

• Do applications which reach out to more apps have less security?
• Do applications which ask for more permissions have less security?
• Are applications which are under a game category less secure?

 With these questions in mind I was able to write a Python script which pulled these relevant values out of the 300,000 line JSON files we were provided. Each JSON file typically contained security scans for two applications. Here is some example output from my Python script. I removed the name of the applications.

fb245f76-b90f-4a03-af9e-96bb7f457164.json
  APP_NAME_1
    30 permissions
    180 urls
    is not a game
    24.9 score
    vulnerabilites
      - api_resource_misconfiguration
  APP_NAME_2
    10 permissions
    174 urls
    is a game
    83.1 score
    vulnerabilites
      - writable_executable_files_private_check
ffb97e85-2df0-496d-9f14-b2b63233224b.json
  APP_NAME_1
    67 urls
    is not a game
    78.7 score
    vulnerabilites
      - leaked_logcat_data_build_fingerprint
  APP_NAME_1
    28 permissions
    52 urls
    is not a game
    81.3 score

correlation between urls and security 0.32574467189115913
correlation between permissions and security 0.17805601971994012
correlation between being a game and security -0.3589489598292062

 There is a print out of the correlation which summarizes our findings. To fully interpret the correlation data you could say that:

• Apps that request more URLs are weakly more secure
• Apps that ask for more permissions are very weakly more secure
• Apps that are games are weakly less secure

 None of the correlation data is strong but the whole dataset is just a collection of 22 applications. Since there is not a large sample size it is hard to get a strong conclusion. One of our applications was TikTok which happened to be a major outlier for our data.

The Results

 I found the results surprising. I would have expected that more requests and more permissions would mean more of a security attack vector. However, the results seem to suggest that the more they have of these the more secure. This lead me to question if the apps which had more requests and permissions were just more mature. In other words apps which gained a following and could support themselves eventually took on the responsibility to secure themselves as they grew. This would certainly explain our outlier which was by far the most popular and had the highest security score.

 snippet for finding correlation and then rendering a graph

print("correlation between being a game & security", df['score'].corr(df['game']))

plt.rcParams.update({'figure.figsize':(10,8), 'figure.dpi':100})
sns.lmplot(x='game', y='score', data=df)
plt.title("Are games less secure?");
plt.show()

Graphs

 I used MatPlotLib to generate graphs as part of the Python Script

 Amount of Requests vs Security (0.33 correlation)

---

 Amount of Permissions vs Security (0.18 correlation)

---

 1 means it is a Game, 0 means it is not a Game (-0.36 correlation)

Thanks for reading 👍