If all lines will always have the same number of fields, this will work.

use warnings;
use strict;

open my $wfh, '>', 'out.csv' or die $!;

my $cols = "Protocol, Source IP Address, Source Port, Data Size\n";
print $wfh $cols;

while (<DATA>){
    if (/
          (?:.*?\s){3}  # get rid of the time
          (.*?)         # capture the proto ($1)
          \s+           # skip the next whitespace    
          (.*?):(\d+)   # separate IP and port, capture both ($2, $3)
          .*?\(         # skip everything until an opening parens
          (\d+)         # capture bytes ($4)
        /x
       ){
        print $wfh "$1, $2, $3, $4\n";
    }
}


__DATA__
2016-04-17 10:12:27:682011 GMT tcp 115.239.248.245:1751 -> 192.168.0.1
+7:8080 52976f9f34d5c286ecf70cac6fba4506 04159c6111bca4f83d7d606a617ac
+c5d6a58328d3a631adf3795f66a5d6265f4d1ec99977a5ae8cb2f3133c9503e5086a5
+f2ac92be196bb0c9a9f653f9669495 (312 bytes)
[download]

output file:

# cat out.csv
Protocol, Source IP Address, Source Port, Data Size
tcp, 115.239.248.245, 1751, 312
[download]

Update: original regex before expanding and explanation, and before not capturing the word 'bytes'

/(?:.*?\s){3}(.*?)\s+(.*?):(\d+).*?\((.*?)\)$/
[download]