AlienVault Longitudinal Study Part 4

(This article was first published on Data Driven Security, and kindly contributed to R-bloggers)

In Part 1 we looked at acquiring raw data, and wrangling it into a time series dataset. In Part 2 we looked at types of threats in the time series. In Part 3 we looked at countries. Now we will examine countries and types in combination in the AlienVault reputation database.

Just as we shaped our dataset for better understanding in previous posts, we will use the top 25 countries in the cctab table, and ignore address types with a compound description, i.e. with a semicolon in the type field. Here is the SQL to make our core dataset for this post:

-- Create table with countries and type summaries
DROP TABLE IF EXISTS avcctyp;
CREATE TABLE avcctyp (
    d         DATE,
    cc        CHAR(2),
    type      CHAR(50),
    avgrisk   DECIMAL(4,2),
    avgrel    DECIMAL(4,2),
    minrisk   INTEGER UNSIGNED,
    minrel    INTEGER UNSIGNED,
    maxrisk   INTEGER UNSIGNED,
    maxrel    INTEGER UNSIGNED,
    count     INTEGER UNSIGNED
);

-- Insert countries, types, and average, min, max scores 
INSERT INTO avcctyp
SELECT d, cc, type,
       AVG(risk) AS avgrisk, AVG(rel) AS avgrel,
       MIN(risk) AS minrisk, MIN(rel) AS minrel,
       MAX(risk) AS maxrisk, MAX(rel) AS maxrel,
       COUNT(*) AS count
  FROM avtrack, avip
 WHERE avtrack.iind=avip.iind
GROUP BY d, cc, type;

-- Delete compound types (with semicolon) and non-top-25 countries 
DELETE FROM avcctyp WHERE type LIKE '%;%';
DELETE FROM avcctyp WHERE cc NOT IN (SELECT * FROM cctab);

Let us now load a data frame in R with the country and type data:

library(RMySQL)

avdb = dbConnect(MySQL(), user='xxxxxx', password='zzzzzzzz', dbname='avrep')
rs <- dbSendQuery(avdb, statement = "SELECT * FROM avcctyp;")
avcctyp <- fetch(rs, n = -1)   # extract all rows
avcctyp$d = as.Date(avcctyp$d, "%Y-%m-%d")

We now have a data frame in R called avcctyp with the top 25 countries (from the first day of our series) and their types, averages and counts for each day of our time series.

Looking at one country, and one day, this SQL gives us a feel for what we can explore:

MariaDB [avrep]> SELECT * FROM avcctyp WHERE d='2013-10-01' AND cc='US';
+------------+------+----------------------+---------+--------+---------+--------+---------+--------+-------+
| d          | cc   | type                 | avgrisk | avgrel | minrisk | minrel | maxrisk | maxrel | count |
+------------+------+----------------------+---------+--------+---------+--------+---------+--------+-------+
| 2013-10-01 | US   | APT                  |    4.00 |   2.00 |       4 |      2 |       4 |      2 |     1 |
| 2013-10-01 | US   | C&C                  |    5.83 |   4.47 |       2 |      4 |      10 |      6 |   239 |
| 2013-10-01 | US   | Malicious Host       |    3.99 |   3.01 |       1 |      1 |       9 |      6 |   616 |
| 2013-10-01 | US   | Malware distribution |    4.00 |   3.00 |       4 |      3 |       4 |      3 |     1 |
| 2013-10-01 | US   | Malware Domain       |    5.81 |   2.05 |       1 |      1 |      10 |      4 |  2061 |
| 2013-10-01 | US   | Malware IP           |    5.11 |   3.02 |       1 |      1 |      10 |      5 |  1483 |
| 2013-10-01 | US   | Scanning Host        |    2.07 |   2.00 |       1 |      1 |       6 |      4 | 40228 |
| 2013-10-01 | US   | Spamming             |    6.11 |   2.39 |       1 |      2 |      10 |      5 |  1161 |
+------------+------+----------------------+---------+--------+---------+--------+---------+--------+-------+
8 rows in set (0.02 sec)

Here are scanning hosts by country:

ggplot(avcctyp[avcctyp$type=="Scanning Host",], aes(d, count)) + geom_point() + theme_bw() +
  scale_x_date(breaks=NULL) + xlab("Day") + ylab("Count") + facet_grid(. ~ cc)

And what we see:

Here are malware domains by country:

ggplot(avcctyp[avcctyp$type=="Malware Domain",], aes(d, count)) + geom_point() + theme_bw() +
  scale_x_date(breaks=NULL) + xlab("Day") + ylab("Count") + facet_grid(. ~ cc)

And what we see:

You can see that each type of address could be investigated this way to determine country based patterns in the various address types. We can also see the various types for a particular country:

ggplot(avcctyp[avcctyp$cc=="US",], aes(d, count)) + geom_point() + theme_bw() +
  scale_y_log10() + scale_x_date(breaks=NULL) + xlab("Day") + ylab("Count") + facet_grid(. ~ type)

And what we see:

This graphic shows us the various types of addresses as facets for United States entries.

There are many other relationships that could be explored in the Alien Vault threat feed. For example, our dataframe has average, minimum and maximum risk and reliability ratings. Countries and types could be more fully explored based on ratings, but these will have to be left as exercises for the reader. This four part series gave you a sense of how to start with a raw data feed and analyze it from a variety of perspectives. We hope you’ll try the code we’ve published here, and take it farther with your own exploration.

To leave a comment for the author, please follow the link and comment on his blog: Data Driven Security.

R-bloggers.com offers daily e-mail updates about R news and tutorials on topics such as: visualization (ggplot2, Boxplots, maps, animation), programming (RStudio, Sweave, LaTeX, SQL, Eclipse, git, hadoop, Web Scraping) statistics (regression, PCA, time series, trading) and more...



If you got this far, why not subscribe for updates from the site? Choose your flavor: e-mail, twitter, RSS, or facebook...

Comments are closed.