A big change is happening in the world of data centers. The software-defined data center (SDDC) is coming, and the biggest data center operators are leading the way – cloud providers, telecoms and social media companies. The change is all about service agility and driving down cost. In short, it is about maximizing data center ROI.
Security is an essential enabler of this change. It has to be embedded in the fabric of the SDDC. What are the key factors for achieving this, and how is Palo Alto Networks able to deliver on those requirements better than any other partner?
1. Zero Trust Security
The amount of machine-to-machine, east-west traffic far eclipses the volume of traffic flowing into and out of the data center. Once malware gains a foothold in an infrastructure, it will move laterally in order to compromise more hosts, gain access to data, and steal or damage assets. With customers and business processes sharing the same physical infrastructure, there is a need for strict security isolation of inter-VM traffic, based on those business processes, rather than on ports, protocols and IP addresses. The Palo Alto Networks® VM-Series security platform provides these fine-grained traffic controls needed to achieve strong risk mitigation in shared environments. By enforcing a Zero Trust model of communication within each business process, the integrity of each process is well secured.
Figure 1. Zero Trust security with Palo Alto Networks VM-Series for KVM/OpenStack®
2. Positive Security Controls
A positive security control model allows only what is explicitly identified as good (“whitelisting”). A negative control model (“blacklisting”) only blocks what is known to be bad. By definition, a negative security model cannot stop unknown threats. Traditional, port-based firewalls implement positive controls, but these controls do not extend to the application layer. To address this gap, legacy firewall vendors have added intrusion prevention system (IPS) functionality to port-based firewalling. However, IPS is still a blacklist approach — anything not known to be bad is assumed to be good. This is how hackers exploit traditional defenses. Our solution implements a positive control model through Layer 7, giving security architects the power to create policies that limit application and data traffic flows to the specific requirements of a business process. No other security solution does this.
Figure 2. Application/User/Content-based policy engine enables positive security control model
3. Security Orchestration
A positive, Zero Trust-based security architecture is of no value if it cannot be implemented and maintained in the software-defined environment. Traditional firewalls are designed to work in relatively static environments, where security policies are tightly coupled with network layer parameters (e.g., IP addresses, ports, protocols). This approach is not adapted to environments where the network topology dynamically adjusts in response to changing demand for services.
The Palo Alto Networks solution architecture is fundamentally different. Policy definition is based on business process-based parameters: the application, the user and the content. Security policy is associated with the logically defined resources (e.g., SQL servers supporting credit card transactions) that implement the business process — regardless of where those resources reside at the network layer. This architectural difference enables security controls to be instantiated dynamically and in concert with the compute, storage and network resources that are instantiated in response to changing demand. Equally important, when resources are removed, the security instances and policies associated with those resources are also removed. Figure 2 illustrates this capability (Dynamic Address Groups).
Figure 3. Security Orchestration with Dynamic Address Groups
4. People and Partnerships
As security and networking become software-defined, service providers are facing a growing skills shortage. Their staffs need to acquire new skills in scripting languages, writing to APIs and combining software packages into full solutions. In short, networking and security engineering is also becoming software-defined. The skills issue is possibly the biggest barrier to meeting the aggressive targets large DC operators have set for themselves.
At Palo Alto Networks, we have invested heavily in developing our people and the industry partnerships needed to help service providers take SDDC from the drawing board to implementation. Our platforms are tested to work with the virtualized platforms and orchestration stacks favored by service providers and enterprises. Our engineers and consultants have the real-world experience working with service providers to implement orchestrated, software-defined security solutions. This strong set of skills, experience and proven solutions are critical success factors for bringing SDDC projects to life on time and on budget.
Conclusion
Security is not a just a perimeter issue for the SDDC. There is an essential need to monitor and control east-west traffic, even within a single, physical server. Positive security controls must provide full visibility, reduce the attack surface to a minimum, and stop unknown threats. The security architecture must be business process-oriented; otherwise, the rules for enforcing policy cannot be orchestrated in tandem with the system as a whole. Finally, it only makes sense to work with a vendor that, not only offers a sound technology foundation, but also has proven solutions, a partner ecosystem, and the people to help you be successful.
As part of our PAN-OS 7.0 release, we—on the Technical Publications team—are happy to provide you with solution-based documentation that helps you use our features to solve real-world network security problems.
What’s New in PAN-OS 7.0
PAN-OS® 7.0.0 Release Notes— Provides information about the PAN-OS 7.0 release, including instructions for upgrading to this release, known issues, software compatibility information, and addressed issues.
(New!) PAN-OS CLI® Quick Start—Provides quick start information and cheat sheets to help you get started with the PAN-OS CLI, including how to find a command and how to get help on how to use a command. This guide supersedes the PAN-OS CLI Reference Guide.
Panorama™ Administrator’s Guide Version 7.0—Provides comprehensive information to help you manage your Palo Alto Networks devices using the Panorama virtual appliance or an M-Series appliance, including instructions for setting up distributed log collection.
VM-Series Deployment Guide Version 7.0—Provides information about the process of deploying the virtualized next-generation firewall to secure east/west or north/south traffic in your network. It includes the deployment topologies supported for public and private cloud.
GlobalProtect™ Administrator’s Guide Version 7.0—Provides comprehensive information for setting up your GlobalProtect infrastructure to extend policy to all users, regardless of location or device.
WildFire™ Administrator’s Guide Version 7.0—Provides instructions on configuring WildFire to identify, analyze and automatically generate protection for previously unknown malware.
Happy reading!
Your friendly Technical Publications team
Palo Alto Networks recently won the Next Generation Firewall award category for NetworkWorld Asia (NWA) 4th annual Information Management Awards 2015. The award was announced on June 4 in Singapore.
Launched in 2012, the NetworkWorld Asia Information Management Awards recognize Asia’s leaders in Information Security, Storage and Data Management. These are the only regional Editors’ Choice Awards for the region, backed by leading Asian publications and portals such as NetworkWorld Asia, Networks Asia, Security Asia and Storage Asia.
A team of editors and CIO advisers shortlisted the finalists for nearly 30 award categories, and then submitted the list for online voting. One hundred regional CIOs and IT heads of end-user organizations were invited to vote, and the entry with the most votes in each category emerged as the award winner.
Winning vendors were judged based on reputation in the marketplace, features and USPs of the solution, widespread acceptance of technology and/or brand, users’ feedback on the solution/deployment, and durability, scalability and quality of service.
KP Unnikrishan, Marketing Director, Asia Pacific & Japan for Palo Alto Networks was present at the Award ceremony. Check out some of the photos from the event below!
KP Unnikrishan (left) with Victor Ng, Editor-in-Chief of NetworkWorld Asia
On May 12, 2015, Unit 42 observed an apparent watering hole attack, also known as a strategic website compromise (SWC), involving the President of Myanmar’s website. Visiting the main page hosted at “www.president-office.gov[.]mm” triggered the malicious content, as the threat actors injected an inline frame (IFRAME) into a JavaScript file used by Drupal for the site’s theme.
Unit 42 believes threat actors chose this website to set up a watering hole in order to target and gather information on individuals in Myanmar, individuals involved in political relations with the country and/or organizations doing business in Myanmar. Unit 42 has evidence to suggest the threat actors have had access to the website since November 2014 if not earlier.
Shortly after we reported the infection to the operators of the website, they took it offline. A new website containing the same content is hosted at “www.myanmarpresidentoffice.info”, which has several artifacts and references to the original content hosted at “president-office.gov.mm” but does not contain the exploit code. We believe the use of the new domain may be part of their remediation process.
This blog discusses the known details of the watering hole, interesting characteristics of the delivered Evilgrab sample (AKA Vidgrab) and the threat infrastructure associated with the attack.
Chain of Compromise
The main page previously hosted at “www.president-office.gov.mm” was powered by Drupal, which loaded several Javascript files that applied a Drupal theme. One of these Javascript files loaded by the main page, named “script.js” and seen in Figure 1, was responsible for several of the website’s features, including the cycling slides of content on the main page.
Figure 1. External JavaScript Used to Load Drupal Theme
The “script.js” file also contained an IFRAME (Figure 2), which Unit 42 believes threat actors injected to exploit the browsers of visitors to the website. We analyzed the content in “script.js”, as well as the HTTP response received from the web server. One interesting thing to note is that the web server, specifically Drupal version 7, used HTTP responses that contain the “Last-Modified” field for caching purposes. We checked the response for the “script.js” file that contained the injected IFRAME and found a “Last-Modified” date of “Wed, 24 Dec 2014 02:38:58 GMT”, which may suggest that the threat actor injected the IFRAME on December 24, 2014.
Figure 2. IFRAME Injected into Drupal JavaScript
Unfortunately, we do not have access to the content that was hosted at this location and requests for access currently result in an HTTP 404 Not Found error. Unit 42 cannot determine which vulnerability this code may have exploited without access to the content. But regardless of the vulnerability exploited, our WildFire system detected the payload in transit and classified the file as malware.
Unit 42 is aware of another malicious script hosted on the President of Myanmar’s website in November 2014, a month prior to when the IFRAME described in this blog appears to have been injected. VirusTotal captured the contents hosted at the following URL[1], which hosted a VBScript[2] that exploited CVE-2014-6332 to install a downloader Trojan:
This suggests threat actors, who may or may not be the same ones who injected the malicious IFRAME, have displayed a consistent interest in compromising visitors to this website since at least November 2014.
Payload Installation
On May 12, 2015, a globally recognized organization in the oil and gas industry visited the following URL that hosted the watering hole on the President of Myanmar’s website:
Visiting this URL resulted in the download of a variant of the Evilgrab Trojan that has been used in past cyber espionage campaigns[3][4]. During our malware analysis efforts, we found some interesting features within this Evilgrab sample, which is denoted as version ‘V2014-v05’ that has the following attributes:
This Evilgrab sample attempts to detect certain antivirus products on an infected system and will only run if it does not detect the presence of Kaspersky, TrendMicro, Symantec’s Norton, ESET, or AVG antivirus products. The initial Evilgrab payload has two embedded dynamic link libraries (DLL): it uses one DLL to load the second DLL that contains the functional code. The initial payload carries out an installation process by storing both of these DLLs, as well as the path to the initial payload, in the Windows registry in encrypted form to the following registry keys:
Software\rar\data – Functional Code DLL
Software\rar\s – Loader DLL
Software\rar\e – Path to Initial Payload
While previous Evilgrab versions also installed their functional code to these registry locations, the installation process itself within the initial Evilgrab payload includes an interesting anti-analysis technique that relies on the structured exception handler (SEH) to call important functions.
Let’s take a step back and first describe the structured event handler, which is built into an application that includes code to handle exceptions. The SEH allows a developer to catch exceptions that occur during the execution of the application and run specific code to handle the exception instead of crashing the application. Exceptions can occur for a variety of reasons, such as attempting to divide a value by zero or attempting to access a memory segment without the proper permissions.
The initial Evilgrab payload uses the SEH to carry out the installation process, by setting up the SEH to call specific functions in the event of an exception and including code that purposefully causes an exception. Evilgrab uses the SEH and forced exceptions as an anti-analysis technique to add a level of difficulty to the malware analysis process. For example, Evilgrab uses the assembly code in Figure 3 that shows a call to a function that we named ‘divBy0_invokeExceptionToCallXor58′.
Figure 3. Assembly Code To Call Function that Forces an Exception
The call to the ‘divBy0_invokeExceptionToCallXor58′ function has a pointer to a buffer that contains cipher text (buf_LoaderDLLInCipherText), as well as a pointer to a DWORD (dd_LoaderDLLLength) that contains the length of the buffer. In the ‘divBy0_invokeExceptionToCallXor58′ function, the assembly instructions in Figure 4 cause an exception by attempting to divide a value by zero by setting the value in ‘ecx’ to zero (xor ecx, ecx instruction) and attempting to divide the value in ‘eax’ with ‘ecx’ (idiv ecx instruction):
Figure 4. Assembly Code To Force an Exception by Dividing by Zero
This division by zero exception invokes the SEH to call a specific function to handle the exception. The exception is handled by the exception handler in Figure 5.
Figure 5. Evilgrab’s Exception Handler Invoked After Forcing an Exception
The exception handler was created to handle the division by zero exception by running the function that Unit 42 named ‘xorBufferBy58′. The purpose of forcing this exception is to call the ‘xorBufferBy58′, using the previously mentioned ‘buf_codeInCipherText’ and ‘dd_codeLength’ values as arguments.
The sample uses this technique to call functions we’ve named ‘createWinlogonProcessAndInjectCode’ and ‘launchInjectedCode’. The ‘createWinlogonProcessAndInjectCode’ function creates a suspended process (CREATE_SUSPENDED flag) using the %SYSTEM%\winlogon.exe executable. It then allocates several memory sections within the winlogon.exe process using VirtualAllocEx and it writes data to these sections using WriteProcessMemory, including the compressed payload that was decrypted using the ‘xorBufferBy58′ function. It also writes a block of shellcode to the entry point of the winlogon.exe process to load the EvilGrab loader DLL, which is responsible for obtaining the Evilgrab functional code from the registry and executing it. When the last exception has been triggered in the initial Evilgrab payload, the SEH calls the ‘launchInjectedCode’ function to resume the suspended ‘winlogon.exe’ process to launch the Evilgrab functional code.
Evilgrab Functionality
Evilgrab is a fully functional remote administration tool (RAT) that allows threat actors to interact with compromised systems to exfiltrate data. The method in which this Evilgrab payload communicates with its C2 server is rather interesting. Previously publically discussed Evilgrab samples sent a beacon of “\x01\x00\x00\x00\x33″ to the C2 server; however, this payload issues a fake HTTP request to the C2 server in place of this beacon. It uses raw sockets to send data to and receive data from its C2 server, which allows the payload to construct custom packets. The fake HTTP request used as a beacon is as follows:
The first four bytes (\xdd\x00\x00\x00) are anomalous, as the HTTP protocol requires the HTTP verb (GET, POST, etc.) to be at the very beginning of the packet. The first four bytes in this packet specify the length of the following data and the remaining bytes are data sent to the C2 server. Evilgrab will use this packet structure for all correspondence with the C2 server. In addition to the anomaly in the first four bytes, the HTTP Host field in the Evilgrab request is also anomalous as it contains a full URL instead of just the hostname of the web server. The malware author put the full URL to a Windows update page in the Host field instead of including the URL portion (/windowsupdate/v6/default.aspx?ln=zh-cn) after the HTTP verb and the domain (update.microsoft.com) in the Host field. The malware author chose this particular Windows update URL in an attempt to make the HTTP request look legitimate.
After the Evilgrab payload sends out this fake HTTP request beacon, it receives the C2 server’s response and checks for a specific response to confirm that the payload communicated with an Evilgrab C2 server. The payload checks the C2 server’s response for the following:
The response shown above is also an anomalous HTTP response for several reasons. First, the Location field does not have a space before the location. Second, “Bad Request (Invalid Verb)” is used in an HTTP 400 Bad Request error message not an HTTP 301 message. The HTTP 400 Bad Request error would make sense, as a web server would expect the HTTP request to start with an HTTP verb but it begins with four bytes for the data length as previously mentioned. Mila at ContagioDump observed the same C2 response to Evilgrab in a delivery document exploiting CVE-2012-0158 in August 2013[5], but that sample did not use the fake HTTP request as a beacon as seen here.
Immediately after receiving the appropriate C2 response to its beacon, Evilgrab sends a 4096-byte packet to the C2 server that contains the following:
Again, the first four bytes is the length of the following data, followed by a static response identifier (0xA0) and a pipe-delimited (‘|’) string of data gathered from the compromised system. Table 1 shows each field and the description of its contents.
Description
Data Type
Example Value
Campaign ID
String
2015-05-13
System IP Address
String
(192.168.180.47)
TCP Port from System
Decimal
49157
Operating System Version
String
Windows7
First Letter of Hostname
Character
J
First Letter of Username
Character
A
Video Capture Device Connected
String
No
System Idle Time
String
0天0小时0分28秒
Removable Drive Connected
String
No
Evilgrab Version
String
V2014-v05
Evilgrab Process ID
Decimal
2052
Static Zero
Decimal
0
Random Value based initial value of 0x50FB125B repeatedly XOR by GetTickCount
Hexadecimal
50fb78a5
Boolean value if the keylogger is running
Hexadecimal
0
Boolean value that the operator sets via the 0x7e command for unknown reason.
Character
0
Table 1. Each element of the system data sent from Evilgrab to the C2 server
The functional Evilgrab code contains a fully featured command handler that allows an operator to interact with the infected system to carry out remote administration activities and data exfiltration. Table 2 contains a comprehensive list of the commands available within the command handler.
Command
Description
0x78
Turns on the QQ Memory Scraper and Keylogger
0x79
Kills the QQ Memory Scraper and Keylogger functionality
0x7a
Sets flags within the class. One of the flags is the hexadecimal value in the initial data sent from the host, specifically the 13th element of the pipe-delimited string
0x7b
Uploads a specified file from the system to the C2 server
0x7c
Creates a file with a specified name.
0x7d
Sends the flags that indicate whether the QQ Memory Scraper and Keylogger are running
0x7e
Sets a boolean value within the ActiveSettings. Unknown reason, but operators may use it to note if they have been there or not.
0x82
Enumerate mounted volumes of storage and their type. The drive type prefixes the volume label, and the drive type prefixes sent within the response to the C2 are: Removable F-Fixed N-remote (network) C-cdrom D-ramdisk
0x83
List contents of a folder, or file, along with each files last modification time, filename and file attributes
0x84
Check to see if a specific file exists.
0x85
Receive a file from the C2 and Execute it
0x86
Creates a file and sets the file pointer
0x87
Close handles to files created in command 0x85
0x88
Loads a DLL using ShellExecuteW using the “open” verb.
0x89
Creates a directory with a specified name
0x8a
Delete a specified file
0x8b
Delete a directory and its contents.
0x8c
Obtains the creation, modification and access times of a file and sends them to the C2
0x8e
Executes a file using Explorer’s token or runs a DLL using ShellExecuteW and the open verb.
0x8f
Move a specified file to a specified location
0x90
Steal credentials from Window’s Protected Storage (PStore)
0x92
Create a reverse shell
0x93
Write string to file for an unknown purpose.
0x94
Sets flag v2 + 0x19
0x98
Enumerates visible Windows and reports the process names to the C2
0x99
Sends the WM_DESTROY message to a specific Window to close it
0x9a
Show a specified Window and set it as the foreground
0x9b
Show a specified Window
0x9c
Set the title of a Window
0x9d
Interact with open window by issuing keystrokes.
0x9f
Issue keystroke
0xb0
Compares the length of v2 + 0xB2 with the specified value.
0xb1
Set a specified registry value, and responds with “\xa6打开子健失败” (Open Zijian failure) if it fails.
0xb2
Delete a specified registry value, and responds with “\xa6删除子健失败” (Remove Zijian failure) if it fails or “\xa5删除子健成功” (Remove Zijian success) if successful.
0xb3
Enumerates the values within a specified registry key, and responds with “\xa5获取目标信息失败” or “\xa5Failed to obtain key information” if it is unsuccessful.
0xb4
Rename a specific registry key to another value, and responds with “\xa6重命名子健失败” or “\xa6Rename Zijian failure” if it is unsuccessful.
0xb5
Create a specific registry key, and responds with “\a7新建项成功” or “\a7New item successful” if it is successful.
0xb7
Deletes a specified key, and responds with “\xaa删除Key失败” (Delete key failure) if it is unsuccessful or “\xab删除Key成功” (Delete Key Success) if successful.
0xb8
Echoes the message 0xb8 back to the C2
0xb9
List services and each service’s status and boot method
0xba
Start or stop a service.
0xbb
Modify the configuration of a service.
0xbc
Creates a service using specified name, description and binary path, and responds with “创建服务 <name> 成功” (Create a service <name> success) if successful.
0xbd
Determines available network locations (TCP and UDP) by calling the GetExtendedTcpTable and GetExtendedUdpTable API functions
0xbe
List running processes.
0xbf
Terminate a specified process
0xc0
Gathers system information, such as operating system version, CPU name and speed, physical memory and amount available, current process ID, as well as data saved to the clipboard.
0xc1
Uninstall Evilgrab.
0xc2
Stop Evilgrab’s main thread, effectively killing Evilgrab until next reboot
0xc3
Same as 0xc2 command
0xc5
Create a temporary file.
0xe0
Closes an open TCP connection that matches a specified network location. This command uses the SetTcpEntry function to close a connection. This command responds “关闭连接成功” (Close the connection is successful).
0xe1
Take a single screenshot
0xe2
Take a single screenshot
0xe3
Starts video capture using single screenshots.
0xe4
Echoes the message 0xe4 back to the C2
0xe5
Starts video capture using single screenshots.
0xe6
List contents of a folder.
0xe9
Sets up proxy communication point between the C2 and another specified network location over a specified TCP port.
0xea
Closes the thread responsible for the proxy communications set up in the 0xe9 command
0xec
Sets up the VideoInputDeviceCategory class for video capture
0xed
Closes a Window, appears to stop the video capture using VideoInputDeviceCategory
0xee
Starts video capture using the VideoInputDeviceCategory class
0xf0
Starts audio capture that it sends directly to the C2
0xf1
Appears to stop the audio capture
0xf2
Search for specific files and exfiltrate their contents.
0xf5
Stops the thread that was created in command 0xf2 to exfiltrate files by setting a specific flag (mainDataStructure[800])
Table 2. Commands available in Evilgrab command handler
In addition to the command handler, Evilgrab’s functional code also contains the following supplemental functionality:
Plugin Support – Evilgrab enumerates the %USERPROFILE%\\WindowsPlugin folder and runs all files with a “.exe” file extension.
QQ Monitoring – Evilgrab monitors for windows associated with Tencent’s QQ messaging program and will scrape memory for strings to steal messages.
Keylogging – Logs keystrokes to ‘%USERPROFILE%\users.bin’.
Unit 42 created a ChopShop module to parse packet captures containing communications between Evilgrab and its C2 server.
Infrastructure Analysis
The Evilgrab payload delivered by the watering hole had the following hardcoded domains that it uses as C2 servers:
dns.websecexp[.]com
ns.websecexp[.]com
appeur.gnway[.]cc
Unit 42 discovered additional infrastructure related these three domains, as seen in the chart in Figure 6.
Figure 6. Infrastructure related to Evilgrab C2 Servers
Unit 42 is aware of the following additional subdomains hosted on the domain websecexp[.]com:
usafi.websecexp[.]com
usacia.websecexp[.]com
webhttps.websecexp[.]com
usagovdns.websecexp[.]com
The domain dns.websecexp[.]com had also been used as a C2 server for a sample of the 9002 Trojan, which is another tool used in cyber espionage campaigns. This domain resolved to the IP address 59.188.16[.]130 as far back as December 2013, which also hosted the following domains. In contrast to websecexp[.] com, this second level was registered using a service to hide the registrant information:
ceshi.mailpseonfz[.]com
dns.mailpseonfz[.]com
Unit 42 is aware of the ceshi.mailpseonfz[.]com domain hosting C2 services for another Evilgrab sample, as well as a sample of the 9002 Trojan. The time frame that the infrastructure above has hosted Evilgrab and 9002 C2 server spans from 2013 to 2015, which suggests the same group is reusing the same infrastructure over a period of years.
Conclusion
Threat actors compromised the President of Myanmar’s website to create a watering hole to infect visitors to the website. Based on data collected in our threat intelligence cloud, the watering hole was active and delivering a malicious payload during May 2015. Open source intelligence suggests that the site may have been a watering hole containing an exploit for CVE-2014-6332 in November 2014 as well. Setting up a watering hole on this site suggests the threat actors, possibly comprising more than one group, are looking to collect information on individuals in Myanmar, individuals involved in political relations with the country and/or organizations doing business in Myanmar.
The May 2015 watering hole delivered a variant of the Evilgrab Trojan to visitors via an unknown vulnerability. The Evilgrab payload itself uses an interesting anti-analysis technique to increase the complexity required to analyze the Trojan. In addition, the Evilgrab payload delivered by this watering hole shares infrastructure that has hosted C2 servers for other Evilgrab payloads, as well as samples of the 9002 Trojan. The threat actors have used this infrastructure in attacks since at least 2013.
This watering hole attack shows threat groups’ continued adoption of this attack vector, as it is much more difficult to analyze and detect than the typical spear-phishing attacks. Once a threat actor has control over the web server hosting the watering hole, the actor can control when to start and stop the delivery of the malicious content, which requires constant monitoring of traffic to the website to determine if and when the attack occurs. However, in this case the threat actors reused old infrastructure to host the C2 servers for the delivered payload, which made detection and attribution easier.
PAN-OS 7.0 is here, with enhancements that help organizations reduce response times, discover and prevent unknown threats, and streamline management of their security infrastructure.
Watch below as our product experts review what’s new in the Palo Alto Networks Security Platform:
