Web Application Defense with Bayesian Attack Analysis (OWASP AppSecDC 2012)

This presentation highlights a Bayesian analysis proof of concept integration of OSBF-Lua and the ModSecurity open source WAF.

Transcript

1. © 2012 Presented by: Web Application Defense with Bayesian Attack Analysis Ryan Barnett Senior Security Researcher OWASP ModSecurity CRS Leader
2. © 2012 Ryan Barnett - Background • Trustwave – Senior Security Researcher – Member of SpiderLabs Research – Surveillance Team Lead •  IDS/IPS •  MailMax •  WAF – Web Application Defense – ModSecurity Project Leader • Author – “Preventing Web Attacks with Apache” •  Pearson Publishing - 2006 – “The Web Application Defenders’ Cookbook” •  Wiley Publishing – (Due end of 2012)
3. © 2012 Agenda •  Attack Resistance Testing –  Blacklist Filter Evasions –  ModSecurity SQL Injection Challenge Result Example •  Evasion Analysis –  Time-to-Hack Metrics –  Common Evasion Methodology •  Using Bayesian Analysis for Attack Detection –  OSBF-Lua within ModSecurity –  Ham/SPAM Training –  Attack Detection Examples •  Conclusion –  Development Plans –  Call for participation
4. © 2012 Target Audience: Defender Community https://www.owasp.org/index.php/Defenders
5. © 2012 Defending Live Web Applications http://www.swsec.com/resources/touchpoints/
6. © 2012
7. © 2012 © 2012 Attack Resistance Testing: Blacklist Filter Evasions
8. © 2012 OWASP ModSecurity Core Rule Set Project http://www.owasp.org/index.php/Category:OWASP_ModSecurity_Core_Rule_Set_Project
9. © 2012 http://blog.spiderlabs.com/2011/06/announcing-the-modsecurity-sql-injection-challenge.html
10. © 2012 DB DB DB DB SQL Injection Challenge Architecture IBM Cenzic HP HTTP/HTTPS Acunetix ModSecurity site
11. © 2012 Two Challenge Levels •  Level I – Speed Hacking –  Find an SQLi attack vector –  Exploit the SQLi vulnerability –  Enumerate the required DB data –  Submit the data to us for review •  Level II – Blacklist Filter Evasion –  Same as Level I, however you must evade the OWASP ModSecurity CRS Blacklist Filters
12. © 2012 Level II – Filter Evasions
13. © 2012 Challenge Participation •  More than 650 participants (in 18 Countries) http://www.modsecurity.org/demo/challenge.html
14. © 2012 Challenge Winners •  Winners received the following: –  Recognition - Name(s) listed on the Challenge website –  Shwag - ModSecurity t-shirt •  Everyone is Happy  •  Well, almost everyone…
15. © 2012 © 2012 Level II Filter Evasion: Example
16. © 2012 Attacking the RegEx Logic SecRule REQUEST_FILENAME|ARGS_NAMES|ARGS|XML:/* \ "\bunion\b.{1,100}?\bselect\b" \ "phase: 2,rev:'2.2.0',capture,t:none,t:urlDecodeUni,t:htmlEntityDe code,t:lowercase, t:replaceComments,t:compressWhiteSpace,ctl:auditLogParts= +E,block,msg:'SQL Injection Attack',id:'959047',tag:'WEB_ATTACK/ SQL_INJECTION',tag:'WASCTC/WASC-19',tag:'OWASP_TOP_10/ A1',tag:'OWASP_AppSensor/CIE1',tag:'PCI/6.5.2',logdata:'% {TX.0}',severity:'2',setvar:'tx.msg=% {rule.msg}',setvar:tx.sql_injection_score=+% {tx.critical_anomaly_score},setvar:tx.anomaly_score=+% {tx.critical_anomaly_score},setvar:tx.%{rule.id}- WEB_ATTACK/SQL_INJECTION-%{matched_var_name}=%{tx.0}" Regex allows up to 100 characters between “union” and “select”
17. © 2012 http://dev.mysql.com/doc/refman/4.1/en/comments.html
18. © 2012 Excessive Comment Text •  %40%40new%20union %23sqlmapsqlmapsqlmapsqlmapsqlmapsqlmap sqlmapsqlmapsqlmapsqlmapsqlmapsqlmapsql mapsqlmapsqlmapsqlmapsql%0Aselect %201,2,database%23sqlmap%0A%28%29 •  URL Decoded •  @@new union#sqlmapsqlmapsqlmapsqlmapsqlmapsql mapsqlmapsqlmapsqlmapsqlmapsqlmapsqlmap sqlmapsqlmapsqlmapsqlmapsql\nselect 1,2,database#sqlmap\n() 103 chars of random text bypasses the regex rule logic
19. © 2012 © 2012 Evasion Analysis
20. © 2012 Common Methodology •  Automation to identify injection points –  NetSparker –  Arachni –  Sqlmap –  Havij •  Manual testing to develop working SQLi payloads –  An iterative process of trial and error 1.  Send initial payloads and observe DB responses 2.  Use obfuscation tactics (comments, encodings, etc…) 3.  Send payload and observe DB response 4.  Repeat steps 2 - 3
21. © 2012 Iterative Testing Example div 1 union%23%0Aselect 1,2,current_user div 1 union%23foo*/*bar%0Aselect 1,2,current_user div 1 union%23foofoofoofoo*/*bar%0Aselect 1,2,current_user div 1 union%23foofoofoofoofoofoofoofoofoofoo*/*bar %0Aselect 1,2,current_user … div 1 union %23foofoofoofoofoofoofoofoofoofoofoofoofoofo ofoofoofoofoofoofoofoofoofoofoofoofoofoofoofoo foofoofoofoo*/*bar%0Aselect 1,2,current_user
22. © 2012 Time-to-Hack Metrics Time-to-Hack Metric Speed Hacking Filter Evasion Avg. # of Requests 170 433 Avg. Duration (Time) 5 hrs 23 mins 72 hrs Shortest # of Requests 36 118 Shortest Duration (Time) 46 mins 10 hrs
23. © 2012 Filter Evasion Conclusions •  Blacklist filtering will only slow down determined attackers •  Attackers need to try many permutations to identify a working filter evasion •  The OWASP ModSecurity Core Rules Set’s blacklists SQLi signatures caught several hundred attempts before an evasion was found Questions •  How can we use this methodology to our advantage? •  What detection technique can we use other than regular expressions?
24. © 2012 Application Intrusion Detection •  Positive/Whitelist Security Model Input Validation –  Allowed characters –  Length –  WAF Traffic Profiling •  Response Time Latency Tracking –  Deviations of response data due to blind SQLi queries (waitfor delay, benchmark() or pg_sleep) •  Response Page Fingerprint Deviations –  Changes to the page construction (title, size, etc…) –  Deviation in the amount of sensitive records returned https://www.owasp.org/index.php/Category:OWASP_AppSensor_Project
25. © 2012 © 2012 Using Bayesian Analysis for Attack Detection
26. © 2012 Bayesian Analysis for HTTP •  RegEx detection is binary –  The operator either matched or it didn’t –  Need a method of detecting attack probability •  Bayesian analysis has achieved great results in Anti-SPAM efforts for email •  Can’t we use the same detection logic for HTTP data? –  Data Source •  Email – OS level text files •  HTTP – text taken directly from HTTP transaction –  Data Format •  Email – Mime headers + Email body •  HTTP – URI + Request Headers + Parameters –  Data Classification •  Non-malicious HTTP request = HAM •  HTTP Attack payloads = SPAM
27. © 2012 OSBF-Lua •  OSBF-Lua by Fidelis Assis –  Orthogonal Sparse Bigrams with Confidence Factor (OSBF) –  Uses space characters for tokenization (which means that it factors in meta-characters) –  Very fast –  Accurate classifiers –  http://osbf-lua.luaforge.net/ •  Moonfilter by Christian Siefkes –  Wrapper script for OSBF –  http://www.siefkes.net/software/moonfilter/ •  Integrate with ModSecurity’s Lua API
28. © 2012 Training the OSBF Classifiers Attack Detected? (Using the OWASP ModSecurity CRS) No Train as HAM Yes Train as SPAM
29. © 2012 Theory of Operation - HAM 1.  Non-malicious user data does not trigger any blacklist rules 2.  Lua script trains OSBF classifier that payloads are HAM Lua: Executing script: /etc/httpd/modsecurity.d/bayes_train_ham.lua Arg Name: ARGS:txtFirstName and Arg Value: Bob. Arg Name: ARGS:txtLastName and Arg Value: Smith. Arg Name: ARGS:txtSocialScurityNo and Arg Value: 123-12-9045. Arg Name: ARGS:txtDOB and Arg Value: 1958-12-12. Arg Name: ARGS:txtAddress and Arg Value: 123 Someplace Dr.. Arg Name: ARGS:txtCity and Arg Value: Fairfax. Arg Name: ARGS:drpState and Arg Value: VA. Arg Name: ARGS:txtTelephoneNo and Arg Value: 703-794-2222. Arg Name: ARGS:txtEmail and Arg Value: bob.smith@mail.com. Arg Name: ARGS:txtAnnualIncome and Arg Value: $90,000. Arg Name: ARGS:drpLoanType and Arg Value: Car. Arg Name: ARGS:sendbutton1 and Arg Value: Submit. Low Bayesian Score: . Training payloads as non-malicious.
30. © 2012 Theory of Operation - SPAM 1.  Attacker sends malicious payloads during initial testing phase 2.  Payloads are caught by our blacklist rules 3.  Lua script trains OSBF classifier that payloads are SPAM [Thu Nov 03 15:21:08 2011] [error] [client 72.192.214.223] ModSecurity: Warning. Pattern match ".*" at TX:981231-WEB_ATTACK/SQL_INJECTION- ARGS:artist. [file "/etc/httpd/modsecurity.d/crs/ base_rules/modsecurity_crs_48_bayes_analysis.conf"] [line "1"] [data "Completed Bayesian Training on SQLi Payload: @@new union#sqlmapsqlmapsqlmapsqlmapsqlmapsqlmapsqlmapsqlmap sqlmapsqlmapsqlmapsqlmapsqlmapsqlmapsqlmapsqlmapsql\ \x0aselect 1,2,database#sqlmap\\x0a()."] [hostname "www.modsecurity.org"] [uri "/testphp.vulnweb.com/ artists.php"] [unique_id "VCqlxsCo8AoAADYJV3kAAAAH"]
31. © 2012 Theory of Operation - Unknown •  Previous evasion payload is now caught [Thu Nov 03 15:28:18 2011] [error] [client 72.192.214.223] ModSecurity: Warning. Bayesian Analysis Alert for ARGS:artist with payload: "@@new union#sqlmapsqlmapsqlmapsqlmapsqlmapsqlmapsqlmapsqlmapsqlmap sqlmapsqlmapsqlmapsqlmapsqlmapsqlmapsqlmapsql\nselect 1,2,database#sqlmap\n()" [file "/etc/httpd/modsecurity.d/ crs/base_rules/modsecurity_crs_48_bayes_analysis.conf"] [line "3"] [msg "Bayesian Analysis Detects Probable SQLi Attack."] [data "Score: {prob=0.99999999965698,probs= {0.99999999965698,3.4301898614548e-10},class=\\x22/var/log/ httpd/spam\\x22,pR=5.5841622861233,reinforce=true}"] [severity "CRITICAL"] [tag "WEB_ATTACK/SQL_INJECTION"] [tag "WASCTC/WASC-19"] [tag "OWASP_TOP_10/A1"] [tag "OWASP_AppSensor/CIE1"] [tag "PCI/6.5.2"] [hostname "www.modsecurity.org"] [uri "/testphp.vulnweb.com/ artists.php"] [unique_id "bcjElMCo8AoAADYlSXMAAAAI"]
32. © 2012 Bayesian Alert for Evasion Payload
33. © 2012 © 2012 Conclusion
34. © 2012 Development Plans/Call for Assistance •  This proof of concept will eventually be put into the OWASP ModSecurity CRS –  Other projects may consider using it too (AppSensor, ESAPI, etc…) •  Need to include HTTP Header data in training –  For accurate Bayesian classification, more data is better. –  Including HTTP Header data may also help to identify non-browser/tool attacks •  Need more testing –  If you would like to help with testing, please contact me and I will provide you access to the Lua scripts.
35. © 2012 ModSecurity T-Shirt Giveaway •  What was the shortest “Time-to-Evasion” from Level II? •  10 hrs.
36. © 2012 Contact/Resources •  Email –  OWASP: ryan.barnett@owasp.org –  Trustwave: rbarnett@trustwave.com •  Twitter –  @ryancbarnett –  @ModSecurity –  @SpiderLabs •  Blog –  http://tacticalwebappsec.blogspot.com –  http://blog.spiderlabs.com