A Crawlerbased Study of Spyware in the Web

1
A Crawler-based Study of Spyware in the Web

• Alex Moshchuk, Tanya Bragin,
• Steve Gribble, Hank Levy

2
Why measure spyware?

• Understand the problem before defending against
  it
• Many unanswered questions
• Whats the spyware density on the web?
• Where do people get spyware?
• How many spyware variants are out there?
• What kinds of threats does spyware pose?
• New ideas and tools
• Detection
• Prevention

3
Approach

• Large-scale study of spyware on the Web
• Crawl interesting portions of the web
• Download content
• Determine if it is malicious
• Use virtual machines
• Two strategies
• Executable study
• Find executables with known spyware
• Drive-by download study
• Find web pages with drive-by downloads

4
Analyzing Executables

• Web crawler collects a pool of executabes
• For each
• Clone a clean virtual machine
• 10-node VM cluster, 4 VMs per node
• Automatically install executable
• Run analysis to see what changed
• Currently, an anti-spyware tool (Ad-Aware)
• Average analysis time 90 sec. per executable

5
Analyzing Drive-by Downloads

• Evaluate the safety of browsing the web
• Automatic virtual browsing
• Render pages in a real browser inside clean VM
• Internet Explorer
• Define triggers for suspicious browsing activity
• Process creation
• Files written outside browser temp. folders
• Suspicious registry modifications
• Run anti-spyware check only when trigger fires

6
Executable Study Results

• Crawled 32 million pages in 10000 domains
• Downloaded 26,000 executables
• Found spyware in 13.5 of them
• 6 installed three or more spyware variants
• 142 unique spyware threats
• Only 29 found more than 20 times

7
Infection of Executables

• Visit a site and download a program
• Whats the chance that you got spyware?

8
Drive-by Download Results

• 5.5 of pages we examined carried drive-by
  downloads
• 1.4 exploit browser vulnerabilities

9
Types of spyware

• Quantify the kinds of threats posed by spyware
• Consider five spyware functions
• Whats the chance a spyware program contains each
  function?

10
Summary

• Lots of bad stuff on the web
• 1 in 8 programs is infected with spyware
• 1 in 18 web pages has a spyware drive-by download
• Most of it is just annoying (Adware)
• But a significant fraction poses big risks
• Spyware companies target specific popular content
• Few spyware variants are encountered in practice
• More details
• A Crawler-based Study of Spyware in the Web
  NDSS06