Web Application Security
                    Guillermo Pérez
                  bisho@tuenti.com
            BE arch & Security Lead Eng.

                   bcndevcon 2011
Things to deal with...
     in web app security
Web App security
●   Anonymous attackers
●   Worldwide access
●   Shared environment for all users
●   Easy distribution, profitable
●   On top of all other components security:
    ○ Network security
    ○ OS security
    ○ Server software security
    ○ Social Engineering
    ○ Even more! browsers, plugins, virus, user computer
       security, shared computers, open wifis...
How to achieve it?
Web App security


  Humans (developers) are the bigger risk

   Give tools, frameworks & policies so no
developer has to ever think how to secure up
things. Should be clear and the easiest path.

      But there is no perfect security...
Top risks?
Top 10 security issues in webapps
From OWASP (risks != frequency)
  1. Injection
  2. XSS
  3. Broken auth, session management
  4. Insecure direct object references
  5. CSRF
  6. Security misconfiguration
  7. Failure to restrict URL access
  8. Unvalidated redirects
  9. Insecure crypto storage
 10. Insufficient transport layer protection
Top 10 security issues in webapps
 1.   Injection
 2.   XSS
 3.   Broken auth, session management
 4.   Insecure direct object references
 5.   CSRF
 6.   Security misconfiguration
 7.   Failure to restrict URL access
 8.   Unvalidated redirects
 9.   Insecure crypto storage
10.   Insufficient transport layer protection
1. Injection flaws

Trick services to execute unintended
commands to gain control or access
unauthorized data.

● Several types:
  ○   SQL
  ○   OS execution
  ○   LDAP
  ○   XPath
  ○   NoSQL
  ○   uploads
1. Injection flaws
●   Explotability: EASY
●   Prevalence: COMMON
●   Detectability: AVERAGE
●   Impact: SEVERE
● Prevention:
    ○ Keep untrusted data separate from commands
● How:
    ○ Use safe, parametrized apis vs writting code to be
      executed by interpreter.
    ○ Escape special chars depending on interpreter.
    ○ Data cast, whitelist input validation.
1. Injection flaws: SQL
● http://example.com/?id=' or '1'='1
● Explicit cast, escaping IN-PLACE
  ○ mysqli_escape_string()
  ○ ...
● Use prepared statements
  ○ Provides data separation
  ○ Client-side implementations (PDO)
  ○ SELECT * FROM table where id=?
● Use safe apis for query generation
  ○ $mysqlService->select($table, $pk, $fields,
    $where...)
● Safe ORM framework
  ○ $storage->read($keys);
1. Injection flaws: OS




● Don't use OS execution :)
● Escape
  ○ escapeshellarg
1. Injection flaws: uploads

● Don't put them on public folder
● Don't use user-provided data for names
● Whitelist extensions
● Validate content
● Store separately from app (DB, separate
  servers)
● Ensure write permissions are the minimum
  possible
Top 10 security issues in webapps
 1.   Injection
 2.   XSS
 3.   Broken auth, session management
 4.   Insecure direct object references
 5.   CSRF
 6.   Security misconfiguration
 7.   Failure to restrict URL access
 8.   Unvalidated redirects
 9.   Insecure crypto storage
10.   Insufficient transport layer protection
2. XSS


Trick services to return browser-executable
code to user/s.

● Several classifications:
   ○ Breaking context vs sub-context
   ○ Persistant vs non-persistent
   ○ Traditional vs DOM
2. XSS
●   Explotability: AVERAGE
●   Prevalence: WIDESPREAD
●   Detectability: EASY
●   Impact: MODERATE
● Prevention:
    ○ Escape untrusted data depending on context
    ○ HTTP-Only Cookie mitigation is useless
● How:
    ○   Escape everything (even safe vars)
    ○   Escape in TEMPLATES (context aware)
    ○   Other (URL params) in specialized safe apis
    ○   Unit test
2. XSS: Classification by context

● Breaking context:
  ○ <a href="?id<?=$_GET['id']?>">
  ○ "<script> ...
  ○ Easy to detect & test
    ■ Unit-test templates with all injections for all vars
       and validate html
● Non breaking context:
  ○ <a href="<?=$_GET['url']?>">
  ○ javascript: ...
  ○ HARD TO DETECT
2. XSS: Classification by persistance
● Persistant
  ○ Data gets stored in DB
  ○ Users will be hit by regular navigation
  ○ Easier to test (templates)
● Non persistant
  ○ A request with some params returns XSS
  ○ Users need to be trick to navigate into the malicious
      link
   ○ More frequent (No results for 'blah')
   ○ Somewhat harder to test (cover error messages,
      non-template based responses)
2. XSS: Classification by mode

● Traditional
  ○ Just by exploiting browser parsing
  ○ Easy to test
● DOM
  ○ Cheating on JS
    ■ data from server injected in DOM
       ● Use innerText
       ● Do not compose html in JS
    ■ parsing data from uri, forms as 'safe'
  ○ Pretty hard to test. Avoid missuse, provide safe apis.
2. XSS


@tuenti
● Escape on templates
● Escape everything, even what doesn't need
  to be escaped:
    ○ <?=View::escape_unsafe($html)?>
● Link generation framework
● Tests for templates, controllers
2. XSS: HTML
● Never put untrusted data in:
  ○ <script> contents
  ○ HTML comments
  ○ tag/attribute names
  ○ <style> contents
● Contexts: Content, attributes, url params,
  urls, js...
● Rich formating
  ○ Use alternative markup lang
     ■ Markdown
     ■ Textile
  ○ Filter HTML (white listing, carefull!!!)
2. XSS: JS


● Encode with xNNN (" might break HTML
  that is parsed before)
● Prefer reading values from dom
● URL pieces are not safe
● Beware of double context: setInterval('...'),
  eval()
2. XSS: JSON
● Easy to escape (single context)
● Can put the load on the browser (harder to
  test)
● Avoid mixing contexts (json on html, or json
  with/for html)
● Eval json as js can trigger js execution
  ○ Safe, full json encoding in server (never use half-
    baked json templates!!!)
  ○ Use the slow json-parse.js vs json.js regexp
    validation
● Be aware of context. content-type!
Top 10 security issues in webapps
 1.   Injection
 2.   XSS
 3.   Broken auth, session management
 4.   Insecure direct object references
 5.   CSRF
 6.   Security misconfiguration
 7.   Failure to restrict URL access
 8.   Unvalidated redirects
 9.   Insecure crypto storage
10.   Insufficient transport layer protection
3. Auth & Session



Attack authentication and sessions to gain
control over an account.

● Passwords
● Session issues
3. Auth & Session

●   Explotability: AVERAGE
●   Prevalence: COMMON
●   Detectability: AVERAGE
●   Impact: SEVERE
● Prevention:
    ○ SSL, good session handling, detect auth brute force,
      avoid plain text passwords, strong password
      recovery, user sessions control (logout, history,
      close all), detect anomalous login patterns...
3. Auth & Session
● Passwords
    ○ Use SSL or digest auth
    ○ Enforce good passwords, rotation
    ○ Store passwords securely (constant time salted
      hashes)
    ○ fight phising (easy URL, educate users)
●   Authentication
    ○ Don't make distintions between bad login / password
    ○ Reset to hide real logins, time-limited tokens, old
      password invalidate resets
    ○ Detect brute force, lock accounts
    ○ Watch misconfigurations
    ○ Specially on admin, secondary platforms
3. Auth & Session

● Sessions
  ○ Random Ids, >= 128 bit
  ○ Use SSL
  ○ Use secure=yes, httponly for cookies
  ○ No session fixation
  ○ No session ids in URLs
  ○ Change session id on priviledge scalation or switch
    between http->https
  ○ Expiration
  ○ Offer logout, history, close all
  ○ Do not send cookies to CDNs, non-principal sites
Top 10 security issues in webapps
 1.   Injection
 2.   XSS
 3.   Broken auth, session management
 4.   Insecure direct object references
 5.   CSRF
 6.   Security misconfiguration
 7.   Failure to restrict URL access
 8.   Unvalidated redirects
 9.   Insecure crypto storage
10.   Insufficient transport layer protection
4. Direct object references

Apps usually map backend objects to URLs.

An attacker might bypass privacy and
authentication by accessing directly to
resources if don't do the appropiate checks.

● Trusted params
● Images
4. Direct object references

●   Explotability: EASY
●   Prevalence: COMMON
●   Detectability: EASY
●   Impact: MODERATE
● Prevention:
    ○   Properly check privacy on all objects
    ○   Good policy on where to put the privacy check
    ○   Do not trust params. Sign params is an option
    ○   Hide real db keys (show pos X in search Y, /me)
    ○   Make urls hard to guess
4. Direct object references




Sounds stupid...
    ...but happens!
4. Direct object references

● Never check privacy on controllers
● Never check privacy on storage layer
● Privacy in backend api methods
  ○ With entry point documentation
  ○ Clear responsibility for privacy!
  ○ Most of the time implicit with good api design
  ○ Good performance
  ○ Easy to use privacy framework
4. Direct object references


● Documentation
/*
 * @epoint-changes-state YES
 * @epoint-privacy-control IMPLICIT
 * - Only deletes current user tag if exists.
 * @epoint-summary Deletes the current user's tag on a photo
 *...
 */
public function deleteMyTag($photoKey) {
    $userId = CurrentUser::getId();
    ...
4. Direct object references

● Privacy framework api
   ○ TPrivacy::hasAnyOf / hasAllOf
   ○ + Privacy providers

if (TPrivacy::hasAnyOf(
    CurrentUser::getId() == $photoOwner,
    array('TagApi', TagApi::IS_TAGGED, $photoKey),
    array('...
    ...
  )) {
Top 10 security issues in webapps
 1.   Injection
 2.   XSS
 3.   Broken auth, session management
 4.   Insecure direct object references
 5.   CSRF
 6.   Security misconfiguration
 7.   Failure to restrict URL access
 8.   Unvalidated redirects
 9.   Insecure crypto storage
10.   Insufficient transport layer protection
5. CSRF

Cross site request forgery [CSFR in tuenti :)]
Trick a authenticated user to submit requests to
a service and do actions without consent. The
browser will send the cookies and the request
might look legit.

● Image tags (get)
● Forms (post)
● ...
5. CSRF

●   Explotability: AVERAGE
●   Prevalence: WIDESPREAD
●   Detectability: EASY
●   Impact: MODERATE
● Prevention:
    ○ Require a non-predictable token param on all
      actions that modify state
    ○ Use POST for all actions that modify state
    ○ Use custom header in ajax requests
    ○ Check Origin header when available!!!
5. CSRF


@tuenti:
● Before was check when using a post param
  ○ Default values caused us issues
● Now explicit annotation on controllers
  ○ @ChangesState
● Evangelize developers
Top 10 security issues in webapps
 1.   Injection
 2.   XSS
 3.   Broken auth, session management
 4.   Insecure direct object references
 5.   CSRF
 6.   Security misconfiguration
 7.   Failure to restrict URL access
 8.   Unvalidated redirects
 9.   Insecure crypto storage
10.   Insufficient transport layer protection
6. Security misconfiguration




Missing security updates, open proxies, open
ports, default accounts, directory listing,
forgotten hardening...
6. Security misconfiguration

●   Explotability: EASY
●   Prevalence: COMMON
●   Detectability: EASY
●   Impact: MODERATE
● Prevention:
    ○   Develop install & configuration procedures
    ○   Document services and subscribe to updates
    ○   Hide services versions when possible
    ○   Separate components to minimize risks
6. Security misconfiguration

@tuenti:
● we are big >1k servers
    ○ + possibilities for some issue
●   But...
    ○   We use config management (pupet)
    ○   Good deployment procedures, documentation
    ○   Very isolated services
    ○   Few generic web components
    ○   Good systems team
Top 10 security issues in webapps
 1.   Injection
 2.   XSS
 3.   Broken auth, session management
 4.   Insecure direct object references
 5.   CSRF
 6.   Security misconfiguration
 7.   Failure to restrict URL access
 8.   Unvalidated redirects
 9.   Insecure crypto storage
10.   Insufficient transport layer protection
7. URL access




Attacker guesses URLs that lead to
functionality, information.
7. URL access


●   Explotability: EASY
●   Prevalence: UNCOMMON
●   Detectability: AVERAGE
●   Impact: MODERATE
● Prevention:
    ○ Deny by default
    ○ Deploy by selection
7. URL access



@tuenti
● Good deploy system
● Splited environments for production and dev
● Most non-public services restricted to vpn +
  centralized auth
Top 10 security issues in webapps
 1.   Injection
 2.   XSS
 3.   Broken auth, session management
 4.   Insecure direct object references
 5.   CSRF
 6.   Security misconfiguration
 7.   Failure to restrict URL access
 8.   Unvalidated redirects
 9.   Insecure crypto storage
10.   Insufficient transport layer protection
8. Unvalidated redirects




Use a service redirect to trick users into clicking
on a link (belongs to valid service) and achieve
more effective phising/virus
downloads/revenue.
8. Unvalidated redirects

●   Explotability: AVERAGE
●   Prevalence: UNCOMMON
●   Detectability: EASY
●   Impact: MODERATE
● Prevention:
    ○ Don't expose destination URL as param, use
      references to a white list
    ○ Ensure end URLs are safe (Safe search, user
      reporting tools...)
Top 10 security issues in webapps
 1.   Injection
 2.   XSS
 3.   Broken auth, session management
 4.   Insecure direct object references
 5.   CSRF
 6.   Security misconfiguration
 7.   Failure to restrict URL access
 8.   Unvalidated redirects
 9.   Insecure crypto storage
10.   Insufficient transport layer protection
9. Insecure crypto storage


Some data is sensible enought to require being
stored encripted/hashed, to protect it from
being stolen.

Unsalted hashes might be exploitable, backups
might contain keys or cleartext, services might
expose decrypt mecanisms, internal attacks
might have access to keys.
9. Insecure crypto storage

●   Explotability: DIFFICULT
●   Prevalence: UNCOMMON
●   Detectability: DIFFICULT
●   Impact: SEVERE
● Prevention:
    ○ Keep backups encripted, don't store keys on same
      place.
    ○ Use salted hashes and constant time hashes
    ○ Ensure keys are protected
    ○ Don't offer full info (credit card XXXX 1234)
Top 10 security issues in webapps
 1.   Injection
 2.   XSS
 3.   Broken auth, session management
 4.   Insecure direct object references
 5.   CSRF
 6.   Security misconfiguration
 7.   Failure to restrict URL access
 8.   Unvalidated redirects
 9.   Insecure crypto storage
10.   Insufficient transport layer protection
10. Transport layer protection




An attacker might sniff traffic of your users to
steal sessions to retrieve data, do spam...
10. Transport layer protection

●   Explotability: DIFFICULT
●   Prevalence: COMMON
●   Detectability: EASY
●   Impact: MODERATE
● Prevention:
    ○ Ensure to use SSL on all requests & resources
      loaded.
    ○ Change session ids when switching to https.
    ○ If optional, try to detect shared IPs and auto-enable
      on those.
Top 10 security issues in webapps
 1.   Injection
 2.   XSS
 3.   Broken auth, session management
 4.   Insecure direct object references
 5.   CSRF
 6.   Security misconfiguration
 7.   Failure to restrict URL access
 8.   Unvalidated redirects
 9.   Insecure crypto storage
10.   Insufficient transport layer protection
11.   Extras
11. Extras: Cross domain data leak

Ajax is changing the web apps, with js-rich
clients that request data.
Beware of exposing JS / JSON user data
through GET requests without CSRF
tokens/headers! <script> tag is not Cross
Domain safe!
● Require custom header (needs
   XMLhttpRequest) keep using GET
● Check origin header
11. Extras: Clickjacking

Trick users to click/copy content on your page
(by-passing CSRF) by using a hidden frame
● Use Frame-options
● Some anti-frame JS (hard)
   ○ top.location might not be accesible, cause JS error
   ○ redirections might be cancelled
   ○ Best (not pretty):
      blank page with link target _blank, if top.
      location == self.location, add content
11. Extras: Unicode



● Filter special unicode that can break design
● UTF encoding might bypass your XSS-filters
● UTF url encoding might bypass directory
  checks...
● NULL code %00 might bypass suffixes
11. Extras: HTTP/Mail Headers



● Are subject to CR/LF injection, leading to
   ○   XSS
   ○   Spam
   ○   redirection
   ○   ...
● Use safe api
11. Extras: People

● People is always the weakest link
● Phising
   ○   Educate
   ○   Good urls
   ○   Design
   ○   Referer analysis
   ○   React
● Self-inflicted JS injection
   ○ Educate
   ○ Filter content, be aware of surges
No input validation?
No input validation?
Minimize malformed data, make it match
business needs.
NOT as primary method to avoid XSS,
injection...

● Rules:
  ○   SERVER SIDE
  ○   Apply to all (form, url params, cookies, http headers)
  ○   Define whitelists of valid chars
  ○   Define length
  ○   Business on top of that
No input validation?


● Even thought, tuenti has a good validation
  system:
  ○ Based on annotations on controllers.
  ○ At data layer (storage definition)
● Makes exploits harder
● Good practice, clean code
● Explicit args in controllers
Other important aspects
Logging, stats, counters


● Very important for security
● Stats:
   ○ Detect issues, patterns to take measures.
● Logs
  ○ Analize issues.
● Counters
  ○ Detect & react to malicious activity
Error handling


● Sanitize error messages, use same
  templating system
● Do not provide information to users
● Control debug mode
● Dangers:
  ○ Log review tools (XSS)
  ○ As payload upload mecanism
Community
● Take care of community!
  ○ Thank security researchers
  ○ Reply fast
  ○ <24h fix policy
  ○ Tipically <2h!
  ○ Hall of FAME!!!
● How to report
  ○ Standard box security@tuenti.com + dns entries
  ○ Regular user support
   ○ Researchers know us
Web Security Future?
Browser XSS protections



● Reflexion XSS protection
  ○ Different implementations IE8, chrome
  ○ Adds issues, new problems
  ○ Non perfect, might improve?
Client side templates



● Data only requests are easier to escape
● It's harder to inject data into client-side
  templates (only persistent XSS)
● Templates might work in DOM mode
● SLOW in non recent browsers
Better JS




● More secure mashups
  ○ Google Caja...
● More enterprise JS
  ○ Dart, GWT, Closure, CoffeeScript
Plugins ... Apis ... Browsers



● Flash plugin will die!

● But new HTML5 apis will bring more issues

● Browsers extensions nightmare
Avoid cookies


Using XMLHttpRequest with sid as param, from
rich JS apps. Destination domain that does not
have cookies.

● Decreases attack vectors on:
  ○ CSRF
  ○ Click jacking
SSL improvements



● HSTS
    ○ Force SSL
    ○ Certificate pinning
●   False start
    ○ -30% handshake latency
Content Security Policy (Mozilla)
● Restricts a lot of attacking vectors
   ○ Forbids inline javascript
   ○ Forbids dynamic js code: eval, setTimeout(<string>)
   ○ Restricts inline data source (can be reverted for
     images for example)
   ○ Whitelist sources for each type of content (js, css,
     images, ajax...)
   ○ Configures frame permissions better
● Hard to implement in complex sites
   ○ twitter mobile is using it
   ○ Reports issues (to detect attacks, debug/testing
     phase)
?
bisho@tuenti.com

                  We are hiring!
          http://jobs.tuenti.com

Tuenti: Web Application Security

  • 1.
    Web Application Security Guillermo Pérez bisho@tuenti.com BE arch & Security Lead Eng. bcndevcon 2011
  • 2.
    Things to dealwith... in web app security
  • 3.
    Web App security ● Anonymous attackers ● Worldwide access ● Shared environment for all users ● Easy distribution, profitable ● On top of all other components security: ○ Network security ○ OS security ○ Server software security ○ Social Engineering ○ Even more! browsers, plugins, virus, user computer security, shared computers, open wifis...
  • 4.
  • 5.
    Web App security Humans (developers) are the bigger risk Give tools, frameworks & policies so no developer has to ever think how to secure up things. Should be clear and the easiest path. But there is no perfect security...
  • 6.
  • 7.
    Top 10 securityissues in webapps From OWASP (risks != frequency) 1. Injection 2. XSS 3. Broken auth, session management 4. Insecure direct object references 5. CSRF 6. Security misconfiguration 7. Failure to restrict URL access 8. Unvalidated redirects 9. Insecure crypto storage 10. Insufficient transport layer protection
  • 8.
    Top 10 securityissues in webapps 1. Injection 2. XSS 3. Broken auth, session management 4. Insecure direct object references 5. CSRF 6. Security misconfiguration 7. Failure to restrict URL access 8. Unvalidated redirects 9. Insecure crypto storage 10. Insufficient transport layer protection
  • 9.
    1. Injection flaws Trickservices to execute unintended commands to gain control or access unauthorized data. ● Several types: ○ SQL ○ OS execution ○ LDAP ○ XPath ○ NoSQL ○ uploads
  • 10.
    1. Injection flaws ● Explotability: EASY ● Prevalence: COMMON ● Detectability: AVERAGE ● Impact: SEVERE ● Prevention: ○ Keep untrusted data separate from commands ● How: ○ Use safe, parametrized apis vs writting code to be executed by interpreter. ○ Escape special chars depending on interpreter. ○ Data cast, whitelist input validation.
  • 11.
    1. Injection flaws:SQL ● http://example.com/?id=' or '1'='1 ● Explicit cast, escaping IN-PLACE ○ mysqli_escape_string() ○ ... ● Use prepared statements ○ Provides data separation ○ Client-side implementations (PDO) ○ SELECT * FROM table where id=? ● Use safe apis for query generation ○ $mysqlService->select($table, $pk, $fields, $where...) ● Safe ORM framework ○ $storage->read($keys);
  • 12.
    1. Injection flaws:OS ● Don't use OS execution :) ● Escape ○ escapeshellarg
  • 13.
    1. Injection flaws:uploads ● Don't put them on public folder ● Don't use user-provided data for names ● Whitelist extensions ● Validate content ● Store separately from app (DB, separate servers) ● Ensure write permissions are the minimum possible
  • 14.
    Top 10 securityissues in webapps 1. Injection 2. XSS 3. Broken auth, session management 4. Insecure direct object references 5. CSRF 6. Security misconfiguration 7. Failure to restrict URL access 8. Unvalidated redirects 9. Insecure crypto storage 10. Insufficient transport layer protection
  • 15.
    2. XSS Trick servicesto return browser-executable code to user/s. ● Several classifications: ○ Breaking context vs sub-context ○ Persistant vs non-persistent ○ Traditional vs DOM
  • 16.
    2. XSS ● Explotability: AVERAGE ● Prevalence: WIDESPREAD ● Detectability: EASY ● Impact: MODERATE ● Prevention: ○ Escape untrusted data depending on context ○ HTTP-Only Cookie mitigation is useless ● How: ○ Escape everything (even safe vars) ○ Escape in TEMPLATES (context aware) ○ Other (URL params) in specialized safe apis ○ Unit test
  • 17.
    2. XSS: Classificationby context ● Breaking context: ○ <a href="?id<?=$_GET['id']?>"> ○ "<script> ... ○ Easy to detect & test ■ Unit-test templates with all injections for all vars and validate html ● Non breaking context: ○ <a href="<?=$_GET['url']?>"> ○ javascript: ... ○ HARD TO DETECT
  • 18.
    2. XSS: Classificationby persistance ● Persistant ○ Data gets stored in DB ○ Users will be hit by regular navigation ○ Easier to test (templates) ● Non persistant ○ A request with some params returns XSS ○ Users need to be trick to navigate into the malicious link ○ More frequent (No results for 'blah') ○ Somewhat harder to test (cover error messages, non-template based responses)
  • 19.
    2. XSS: Classificationby mode ● Traditional ○ Just by exploiting browser parsing ○ Easy to test ● DOM ○ Cheating on JS ■ data from server injected in DOM ● Use innerText ● Do not compose html in JS ■ parsing data from uri, forms as 'safe' ○ Pretty hard to test. Avoid missuse, provide safe apis.
  • 20.
    2. XSS @tuenti ● Escapeon templates ● Escape everything, even what doesn't need to be escaped: ○ <?=View::escape_unsafe($html)?> ● Link generation framework ● Tests for templates, controllers
  • 21.
    2. XSS: HTML ●Never put untrusted data in: ○ <script> contents ○ HTML comments ○ tag/attribute names ○ <style> contents ● Contexts: Content, attributes, url params, urls, js... ● Rich formating ○ Use alternative markup lang ■ Markdown ■ Textile ○ Filter HTML (white listing, carefull!!!)
  • 22.
    2. XSS: JS ●Encode with xNNN (" might break HTML that is parsed before) ● Prefer reading values from dom ● URL pieces are not safe ● Beware of double context: setInterval('...'), eval()
  • 23.
    2. XSS: JSON ●Easy to escape (single context) ● Can put the load on the browser (harder to test) ● Avoid mixing contexts (json on html, or json with/for html) ● Eval json as js can trigger js execution ○ Safe, full json encoding in server (never use half- baked json templates!!!) ○ Use the slow json-parse.js vs json.js regexp validation ● Be aware of context. content-type!
  • 24.
    Top 10 securityissues in webapps 1. Injection 2. XSS 3. Broken auth, session management 4. Insecure direct object references 5. CSRF 6. Security misconfiguration 7. Failure to restrict URL access 8. Unvalidated redirects 9. Insecure crypto storage 10. Insufficient transport layer protection
  • 25.
    3. Auth &Session Attack authentication and sessions to gain control over an account. ● Passwords ● Session issues
  • 26.
    3. Auth &Session ● Explotability: AVERAGE ● Prevalence: COMMON ● Detectability: AVERAGE ● Impact: SEVERE ● Prevention: ○ SSL, good session handling, detect auth brute force, avoid plain text passwords, strong password recovery, user sessions control (logout, history, close all), detect anomalous login patterns...
  • 27.
    3. Auth &Session ● Passwords ○ Use SSL or digest auth ○ Enforce good passwords, rotation ○ Store passwords securely (constant time salted hashes) ○ fight phising (easy URL, educate users) ● Authentication ○ Don't make distintions between bad login / password ○ Reset to hide real logins, time-limited tokens, old password invalidate resets ○ Detect brute force, lock accounts ○ Watch misconfigurations ○ Specially on admin, secondary platforms
  • 28.
    3. Auth &Session ● Sessions ○ Random Ids, >= 128 bit ○ Use SSL ○ Use secure=yes, httponly for cookies ○ No session fixation ○ No session ids in URLs ○ Change session id on priviledge scalation or switch between http->https ○ Expiration ○ Offer logout, history, close all ○ Do not send cookies to CDNs, non-principal sites
  • 29.
    Top 10 securityissues in webapps 1. Injection 2. XSS 3. Broken auth, session management 4. Insecure direct object references 5. CSRF 6. Security misconfiguration 7. Failure to restrict URL access 8. Unvalidated redirects 9. Insecure crypto storage 10. Insufficient transport layer protection
  • 30.
    4. Direct objectreferences Apps usually map backend objects to URLs. An attacker might bypass privacy and authentication by accessing directly to resources if don't do the appropiate checks. ● Trusted params ● Images
  • 31.
    4. Direct objectreferences ● Explotability: EASY ● Prevalence: COMMON ● Detectability: EASY ● Impact: MODERATE ● Prevention: ○ Properly check privacy on all objects ○ Good policy on where to put the privacy check ○ Do not trust params. Sign params is an option ○ Hide real db keys (show pos X in search Y, /me) ○ Make urls hard to guess
  • 32.
    4. Direct objectreferences Sounds stupid... ...but happens!
  • 33.
    4. Direct objectreferences ● Never check privacy on controllers ● Never check privacy on storage layer ● Privacy in backend api methods ○ With entry point documentation ○ Clear responsibility for privacy! ○ Most of the time implicit with good api design ○ Good performance ○ Easy to use privacy framework
  • 34.
    4. Direct objectreferences ● Documentation /* * @epoint-changes-state YES * @epoint-privacy-control IMPLICIT * - Only deletes current user tag if exists. * @epoint-summary Deletes the current user's tag on a photo *... */ public function deleteMyTag($photoKey) { $userId = CurrentUser::getId(); ...
  • 35.
    4. Direct objectreferences ● Privacy framework api ○ TPrivacy::hasAnyOf / hasAllOf ○ + Privacy providers if (TPrivacy::hasAnyOf( CurrentUser::getId() == $photoOwner, array('TagApi', TagApi::IS_TAGGED, $photoKey), array('... ... )) {
  • 36.
    Top 10 securityissues in webapps 1. Injection 2. XSS 3. Broken auth, session management 4. Insecure direct object references 5. CSRF 6. Security misconfiguration 7. Failure to restrict URL access 8. Unvalidated redirects 9. Insecure crypto storage 10. Insufficient transport layer protection
  • 37.
    5. CSRF Cross siterequest forgery [CSFR in tuenti :)] Trick a authenticated user to submit requests to a service and do actions without consent. The browser will send the cookies and the request might look legit. ● Image tags (get) ● Forms (post) ● ...
  • 38.
    5. CSRF ● Explotability: AVERAGE ● Prevalence: WIDESPREAD ● Detectability: EASY ● Impact: MODERATE ● Prevention: ○ Require a non-predictable token param on all actions that modify state ○ Use POST for all actions that modify state ○ Use custom header in ajax requests ○ Check Origin header when available!!!
  • 39.
    5. CSRF @tuenti: ● Beforewas check when using a post param ○ Default values caused us issues ● Now explicit annotation on controllers ○ @ChangesState ● Evangelize developers
  • 40.
    Top 10 securityissues in webapps 1. Injection 2. XSS 3. Broken auth, session management 4. Insecure direct object references 5. CSRF 6. Security misconfiguration 7. Failure to restrict URL access 8. Unvalidated redirects 9. Insecure crypto storage 10. Insufficient transport layer protection
  • 41.
    6. Security misconfiguration Missingsecurity updates, open proxies, open ports, default accounts, directory listing, forgotten hardening...
  • 42.
    6. Security misconfiguration ● Explotability: EASY ● Prevalence: COMMON ● Detectability: EASY ● Impact: MODERATE ● Prevention: ○ Develop install & configuration procedures ○ Document services and subscribe to updates ○ Hide services versions when possible ○ Separate components to minimize risks
  • 43.
    6. Security misconfiguration @tuenti: ●we are big >1k servers ○ + possibilities for some issue ● But... ○ We use config management (pupet) ○ Good deployment procedures, documentation ○ Very isolated services ○ Few generic web components ○ Good systems team
  • 44.
    Top 10 securityissues in webapps 1. Injection 2. XSS 3. Broken auth, session management 4. Insecure direct object references 5. CSRF 6. Security misconfiguration 7. Failure to restrict URL access 8. Unvalidated redirects 9. Insecure crypto storage 10. Insufficient transport layer protection
  • 45.
    7. URL access Attackerguesses URLs that lead to functionality, information.
  • 46.
    7. URL access ● Explotability: EASY ● Prevalence: UNCOMMON ● Detectability: AVERAGE ● Impact: MODERATE ● Prevention: ○ Deny by default ○ Deploy by selection
  • 47.
    7. URL access @tuenti ●Good deploy system ● Splited environments for production and dev ● Most non-public services restricted to vpn + centralized auth
  • 48.
    Top 10 securityissues in webapps 1. Injection 2. XSS 3. Broken auth, session management 4. Insecure direct object references 5. CSRF 6. Security misconfiguration 7. Failure to restrict URL access 8. Unvalidated redirects 9. Insecure crypto storage 10. Insufficient transport layer protection
  • 49.
    8. Unvalidated redirects Usea service redirect to trick users into clicking on a link (belongs to valid service) and achieve more effective phising/virus downloads/revenue.
  • 50.
    8. Unvalidated redirects ● Explotability: AVERAGE ● Prevalence: UNCOMMON ● Detectability: EASY ● Impact: MODERATE ● Prevention: ○ Don't expose destination URL as param, use references to a white list ○ Ensure end URLs are safe (Safe search, user reporting tools...)
  • 51.
    Top 10 securityissues in webapps 1. Injection 2. XSS 3. Broken auth, session management 4. Insecure direct object references 5. CSRF 6. Security misconfiguration 7. Failure to restrict URL access 8. Unvalidated redirects 9. Insecure crypto storage 10. Insufficient transport layer protection
  • 52.
    9. Insecure cryptostorage Some data is sensible enought to require being stored encripted/hashed, to protect it from being stolen. Unsalted hashes might be exploitable, backups might contain keys or cleartext, services might expose decrypt mecanisms, internal attacks might have access to keys.
  • 53.
    9. Insecure cryptostorage ● Explotability: DIFFICULT ● Prevalence: UNCOMMON ● Detectability: DIFFICULT ● Impact: SEVERE ● Prevention: ○ Keep backups encripted, don't store keys on same place. ○ Use salted hashes and constant time hashes ○ Ensure keys are protected ○ Don't offer full info (credit card XXXX 1234)
  • 54.
    Top 10 securityissues in webapps 1. Injection 2. XSS 3. Broken auth, session management 4. Insecure direct object references 5. CSRF 6. Security misconfiguration 7. Failure to restrict URL access 8. Unvalidated redirects 9. Insecure crypto storage 10. Insufficient transport layer protection
  • 55.
    10. Transport layerprotection An attacker might sniff traffic of your users to steal sessions to retrieve data, do spam...
  • 56.
    10. Transport layerprotection ● Explotability: DIFFICULT ● Prevalence: COMMON ● Detectability: EASY ● Impact: MODERATE ● Prevention: ○ Ensure to use SSL on all requests & resources loaded. ○ Change session ids when switching to https. ○ If optional, try to detect shared IPs and auto-enable on those.
  • 57.
    Top 10 securityissues in webapps 1. Injection 2. XSS 3. Broken auth, session management 4. Insecure direct object references 5. CSRF 6. Security misconfiguration 7. Failure to restrict URL access 8. Unvalidated redirects 9. Insecure crypto storage 10. Insufficient transport layer protection 11. Extras
  • 58.
    11. Extras: Crossdomain data leak Ajax is changing the web apps, with js-rich clients that request data. Beware of exposing JS / JSON user data through GET requests without CSRF tokens/headers! <script> tag is not Cross Domain safe! ● Require custom header (needs XMLhttpRequest) keep using GET ● Check origin header
  • 59.
    11. Extras: Clickjacking Trickusers to click/copy content on your page (by-passing CSRF) by using a hidden frame ● Use Frame-options ● Some anti-frame JS (hard) ○ top.location might not be accesible, cause JS error ○ redirections might be cancelled ○ Best (not pretty): blank page with link target _blank, if top. location == self.location, add content
  • 60.
    11. Extras: Unicode ●Filter special unicode that can break design ● UTF encoding might bypass your XSS-filters ● UTF url encoding might bypass directory checks... ● NULL code %00 might bypass suffixes
  • 61.
    11. Extras: HTTP/MailHeaders ● Are subject to CR/LF injection, leading to ○ XSS ○ Spam ○ redirection ○ ... ● Use safe api
  • 62.
    11. Extras: People ●People is always the weakest link ● Phising ○ Educate ○ Good urls ○ Design ○ Referer analysis ○ React ● Self-inflicted JS injection ○ Educate ○ Filter content, be aware of surges
  • 63.
  • 64.
    No input validation? Minimizemalformed data, make it match business needs. NOT as primary method to avoid XSS, injection... ● Rules: ○ SERVER SIDE ○ Apply to all (form, url params, cookies, http headers) ○ Define whitelists of valid chars ○ Define length ○ Business on top of that
  • 65.
    No input validation? ●Even thought, tuenti has a good validation system: ○ Based on annotations on controllers. ○ At data layer (storage definition) ● Makes exploits harder ● Good practice, clean code ● Explicit args in controllers
  • 66.
  • 67.
    Logging, stats, counters ●Very important for security ● Stats: ○ Detect issues, patterns to take measures. ● Logs ○ Analize issues. ● Counters ○ Detect & react to malicious activity
  • 68.
    Error handling ● Sanitizeerror messages, use same templating system ● Do not provide information to users ● Control debug mode ● Dangers: ○ Log review tools (XSS) ○ As payload upload mecanism
  • 69.
    Community ● Take careof community! ○ Thank security researchers ○ Reply fast ○ <24h fix policy ○ Tipically <2h! ○ Hall of FAME!!! ● How to report ○ Standard box security@tuenti.com + dns entries ○ Regular user support ○ Researchers know us
  • 70.
  • 71.
    Browser XSS protections ●Reflexion XSS protection ○ Different implementations IE8, chrome ○ Adds issues, new problems ○ Non perfect, might improve?
  • 72.
    Client side templates ●Data only requests are easier to escape ● It's harder to inject data into client-side templates (only persistent XSS) ● Templates might work in DOM mode ● SLOW in non recent browsers
  • 73.
    Better JS ● Moresecure mashups ○ Google Caja... ● More enterprise JS ○ Dart, GWT, Closure, CoffeeScript
  • 74.
    Plugins ... Apis... Browsers ● Flash plugin will die! ● But new HTML5 apis will bring more issues ● Browsers extensions nightmare
  • 75.
    Avoid cookies Using XMLHttpRequestwith sid as param, from rich JS apps. Destination domain that does not have cookies. ● Decreases attack vectors on: ○ CSRF ○ Click jacking
  • 76.
    SSL improvements ● HSTS ○ Force SSL ○ Certificate pinning ● False start ○ -30% handshake latency
  • 77.
    Content Security Policy(Mozilla) ● Restricts a lot of attacking vectors ○ Forbids inline javascript ○ Forbids dynamic js code: eval, setTimeout(<string>) ○ Restricts inline data source (can be reverted for images for example) ○ Whitelist sources for each type of content (js, css, images, ajax...) ○ Configures frame permissions better ● Hard to implement in complex sites ○ twitter mobile is using it ○ Reports issues (to detect attacks, debug/testing phase)
  • 78.
    ? bisho@tuenti.com We are hiring! http://jobs.tuenti.com