Web applications are attacked one out of three days

A typical web application is attacked at least once every three days, according to a report released by data security firm Imperva.

The third edition of Imperva’s semi-annual Web Application Attack Report (WAAR), released on Tuesday, is based on an analysis of internet traffic collected from 50 publicly available web applications between December 2011 and May 2012.

Imperva determined that a typical web application experienced 59 “battle days” – days in which at least one attack incident occurred – during the six-month period.

Many of the monitored applications differed in size and purpose, and most of them were hosted in the U.S. and the European Union, said Amichai Shulman, Imperva’s chief technology officer.

Imperva found that for a typical web application the median number of attack incidents recorded during a six-month period was 137.

An attack incident was defined by the company as a burst of malicious traffic that exceeded a rate of 30 attack requests per five minutes.

This method of counting attacks was significantly different than the one used by the company for its previous WAAR reports, which focused on the total number of attack requests.

The worst case seen by the company involved an application that experienced 1,383 attack incidents spanning 141 battle days, or 80 percent of the days in the six-month period.

The typical attack incident had a magnitude of 195 requests and lasted almost 8 minutes, Imperva said in its report. However, the worst incident lasted 10 times longer than that and involved 8,790 attack requests.

The new methodology of interpreting data revealed that SQL injection (SQLi) was the attack technique most commonly used. The median number of SQLi attacks experienced by a typical web application was 17.5 and in the worst case it was 320.

This is a significant change, because previous WAAR reports placed cross-site scripting (XSS) and directory traversal attacks ahead of SQLi attacks in terms of frequency.

The new methodology allowed the company’s researchers to see things in a different way, Shulman said. “While the number of individual requests for cross-site scripting and directory traversal is higher than for SQL injection, in reality, the number of attacks in which SQL injection is involved is higher.”

It’s possible that SQL injection, while the most popular attack technique, is not the most successful one, Shulman said. However, “I find it hard to believe that attackers are wasting so much energy over SQL injection if it’s not proving to be successful,” he said.

Another interesting finding was that the highest number of SQL injection requests originated in France and not the U.S., which is the primary source of other types of attacks like remote file inclusion, directory traversal or local file inclusion.

Some attack types have a well-known geographic bias, Shulman said. For example, many email scraping attacks originate from African countries and comment spam attacks are commonly launched from Eastern Europe and Russia.

Imperva tried to use a number of statistical methods to find patterns in the timing of the attacks and actually concluded that they can’t be predicted, Shulman said. “The fact that you’ve been attacked today doesn’t say anything about what is going to happen tomorrow.”

Companies need to be prepared to protect their Web applications at all times and should be prepared to do so against the worst attacks, not just the average ones, he said.