DNS-Query-NX.r

# Library Loading
library("RPostgreSQL");
library("car");

# Connect to Database
pgDrv <- dbDriver("PostgreSQL")
dbh <- dbConnect(pgDrv, host="localhost", dbname="dnsmonitor", user="dnsmon", password="tooEasy")

# Retrieve Statistics from DB
stats <- dbGetQuery(dbh, "select client.id, client.ip, sum(queries) as queries, sum(nx) as nx, sum(answers) as answers, sum(errors) as errors, count(distinct day) as days_active
from client
	inner join client_stats on client.id = client_stats.client_id
	where ip BETWEEN '10.1.0.0' and '10.1.0.255' 
	group by client.id, client.ip having count(distinct day) > 20")
  
# Close the Database Connection and free variables
dbDisconnect(dbh)
rm(dbh)
rm(pgDrv)

# Queries will be "x"
x <- stats$queries
x.minusmean = x - mean(x)
x.minusmeansq = x.minusmean ^ 2

# NX Records will be "y"
y <- stats$nx
y.minusmean = y - mean(y)
y.minusmeansq = y.minusmean ^ 2

# Standard Deviations
sd.x = sd(x);
sd.y = sd(y);

# Build the Table as in #5
dns <- data.frame( x, y, row.names = stats$ip )
dns$x_mean <- x.minusmean
dns$x_meansq <- x.minusmeansq
dns$y_mean <- y.minusmean
dns$y_meansq <- y.minusmeansq
# Calculate Cor. Coef. Numerator.
dns$product = x.minusmean * y.minusmean
# Z-Scores
dns$x_zscore = abs(x.minusmean) / sd.x
dns$y_zscore = abs(y.minusmean) / sd.y

# Removing Outliers:
normal <- subset(dns, y_zscore <= 3, select = c( x, y ))

# Scatter Plot
plot( x, y, xlab="Queries", ylab="NX Responses",
      main="DNS Queries and NX Responses")

# Regression Line
regression <- lm( y ~ x )
regLine( regression, col="red" )

# Regression line without outliers:
regression_normal <- lm( normal$y ~ normal$x )
regLine( regression_normal, col = "green" )