ATT&CK explorer

An adversary may create a new instance or virtual machine (VM) within the compute service of a cloud account to evade defenses. Creating a new instance may allow an adversary to bypass firewall rules and permissions that exist on instances currently residing within an account. An adversary may Create Snapshot of one or more volumes in an account, create a new instance, mount the snapshots, and then apply a less restrictive security policy to collect Data from Local System or for Remote Data Staging.(Citation: Mandiant M-Trends 2020) Creating a new instance may also allow an adversary to carry out malicious activity within an environment without affecting the execution of current running instances.

An adversary may delete a cloud instance after they have performed malicious activities in an attempt to evade detection and remove evidence of their presence. Deleting an instance or virtual machine can remove valuable forensic artifacts and other evidence of suspicious behavior if the instance is not recoverable. An adversary may also Create Cloud Instance and later terminate the instance after achieving their objectives.(Citation: Mandiant M-Trends 2020)

An adversary may revert changes made to a cloud instance after they have performed malicious activities in attempt to evade detection and remove evidence of their presence. In highly virtualized environments, such as cloud-based infrastructure, this may be accomplished by restoring virtual machine (VM) or data storage snapshots through the cloud management dashboard or cloud APIs. Another variation of this technique is to utilize temporary storage attached to the compute instance. Most cloud providers provide various types of storage including persistent, local, and/or ephemeral, with the ephemeral types often reset upon stop/restart of the VM.(Citation: Tech Republic - Restore AWS Snapshots)(Citation: Google - Restore Cloud Snapshot)

Adversaries may modify settings that directly affect the size, locations, and resources available to cloud compute infrastructure in order to evade defenses. These settings may include service quotas, subscription associations, tenant-wide policies, or other configurations that impact available compute. Such modifications may allow adversaries to abuse the victim’s compute resources to achieve their goals, potentially without affecting the execution of running instances and/or revealing their activities to the victim. For example, cloud providers often limit customer usage of compute resources via quotas. Customers may request adjustments to these quotas to support increased computing needs, though these adjustments may require approval from the cloud provider. Adversaries who compromise a cloud environment may similarly request quota adjustments in order to support their activities, such as enabling additional Resource Hijacking without raising suspicion by using up a victim’s entire quota.(Citation: Microsoft Cryptojacking 2023) Adversaries may also increase allowed resource usage by modifying any tenant-wide policies that limit the sizes of deployed virtual machines.(Citation: Microsoft Azure Policy) Adversaries may also modify settings that affect where cloud resources can be deployed, such as enabling Unused/Unsupported Cloud Regions.

Adversaries may bridge network boundaries by compromising perimeter network devices or internal devices responsible for network segmentation. Breaching these devices may enable an adversary to bypass restrictions on traffic routing that otherwise separate trusted and untrusted networks. Devices such as routers and firewalls can be used to create boundaries between trusted and untrusted networks. They achieve this by restricting traffic types to enforce organizational policy in an attempt to reduce the risk inherent in such connections. Restriction of traffic can be achieved by prohibiting IP addresses, layer 4 protocol ports, or through deep packet inspection to identify applications. To participate with the rest of the network, these devices can be directly addressable or transparent, but their mode of operation has no bearing on how the adversary can bypass them when compromised. When an adversary takes control of such a boundary device, they can bypass its policy enforcement to pass normally prohibited traffic across the trust boundary between the two separated networks without hinderance. By achieving sufficient rights on the device, an adversary can reconfigure the device to allow the traffic they want, allowing them to then further achieve goals such as command and control via Multi-hop Proxy or exfiltration of data via Traffic Duplication. Adversaries may also target internal devices responsible for network segmentation and abuse these in conjunction with Internal Proxy to achieve the same goals.(Citation: Kaspersky ThreatNeedle Feb 2021) In the cases where a border device separates two separate organizations, the adversary can also facilitate lateral movement into new victim environments.

Adversaries may bridge network boundaries by modifying a network device’s Network Address Translation (NAT) configuration. Malicious modifications to NAT may enable an adversary to bypass restrictions on traffic routing that otherwise separate trusted and untrusted networks. Network devices such as routers and firewalls that connect multiple networks together may implement NAT during the process of passing packets between networks. When performing NAT, the network device will rewrite the source and/or destination addresses of the IP address header. Some network designs require NAT for the packets to cross the border device. A typical example of this is environments where internal networks make use of non-Internet routable addresses.(Citation: RFC1918) When an adversary gains control of a network boundary device, they may modify NAT configurations to send traffic between two separated networks, or to obscure their activities. In network designs that require NAT to function, such modifications enable the adversary to overcome inherent routing limitations that would normally prevent them from accessing protected systems behind the border device. In network designs that do not require NAT, adversaries may use address translation to further obscure their activities, as changing the addresses of packets that traverse a network boundary device can make monitoring data transmissions more challenging for defenders. Adversaries may use Patch System Image to change the operating system of a network device, implementing their own custom NAT mechanisms to further obscure their activities.

Adversaries may compromise a network device’s encryption capability in order to bypass encryption that would otherwise protect data communications. (Citation: Cisco Synful Knock Evolution) Encryption can be used to protect transmitted network traffic to maintain its confidentiality (protect against unauthorized disclosure) and integrity (protect against unauthorized changes). Encryption ciphers are used to convert a plaintext message to ciphertext and can be computationally intensive to decipher without the associated decryption key. Typically, longer keys increase the cost of cryptanalysis, or decryption without the key. Adversaries can compromise and manipulate devices that perform encryption of network traffic. For example, through behaviors such as Modify System Image, Reduce Key Space, and Disable Crypto Hardware, an adversary can negatively effect and/or eliminate a device’s ability to securely encrypt network traffic. This poses a greater risk of unauthorized disclosure and may help facilitate data manipulation, Credential Access, or Collection efforts. (Citation: Cisco Blog Legacy Device Attacks)

Adversaries may reduce the level of effort required to decrypt data transmitted over the network by reducing the cipher strength of encrypted communications.(Citation: Cisco Synful Knock Evolution) Adversaries can weaken the encryption software on a compromised network device by reducing the key size used by the software to convert plaintext to ciphertext (e.g., from hundreds or thousands of bytes to just a couple of bytes). As a result, adversaries dramatically reduce the amount of effort needed to decrypt the protected information without the key. Adversaries may modify the key size used and other encryption parameters using specialized commands in a Network Device CLI introduced to the system through Modify System Image to change the configuration of the device. (Citation: Cisco Blog Legacy Device Attacks)

Adversaries disable a network device’s dedicated hardware encryption, which may enable them to leverage weaknesses in software encryption in order to reduce the effort involved in collecting, manipulating, and exfiltrating transmitted data. Many network devices such as routers, switches, and firewalls, perform encryption on network traffic to secure transmission across networks. Often, these devices are equipped with special, dedicated encryption hardware to greatly increase the speed of the encryption process as well as to prevent malicious tampering. When an adversary takes control of such a device, they may disable the dedicated hardware, for example, through use of Modify System Image, forcing the use of software to perform encryption on general processors. This is typically used in conjunction with attacks to weaken the strength of the cipher in software (e.g., Reduce Key Space). (Citation: Cisco Blog Legacy Device Attacks)

Adversaries may make changes to the operating system of embedded network devices to weaken defenses and provide new capabilities for themselves. On such devices, the operating systems are typically monolithic and most of the device functionality and capabilities are contained within a single file. To change the operating system, the adversary typically only needs to affect this one file, replacing or modifying it. This can either be done live in memory during system runtime for immediate effect, or in storage to implement the change on the next boot of the network device.

Adversaries may modify the operating system of a network device to introduce new capabilities or weaken existing defenses.(Citation: Killing the myth of Cisco IOS rootkits) (Citation: Killing IOS diversity myth) (Citation: Cisco IOS Shellcode) (Citation: Cisco IOS Forensics Developments) (Citation: Juniper Netscreen of the Dead) Some network devices are built with a monolithic architecture, where the entire operating system and most of the functionality of the device is contained within a single file. Adversaries may change this file in storage, to be loaded in a future boot, or in memory during runtime. To change the operating system in storage, the adversary will typically use the standard procedures available to device operators. This may involve downloading a new file via typical protocols used on network devices, such as TFTP, FTP, SCP, or a console connection. The original file may be overwritten, or a new file may be written alongside of it and the device reconfigured to boot to the compromised image. To change the operating system in memory, the adversary typically can use one of two methods. In the first, the adversary would make use of native debug commands in the original, unaltered running operating system that allow them to directly modify the relevant memory addresses containing the running operating system. This method typically requires administrative level access to the device. In the second method for changing the operating system in memory, the adversary would make use of the boot loader. The boot loader is the first piece of software that loads when the device starts that, in turn, will launch the operating system. Adversaries may use malicious code previously implanted in the boot loader, such as through the ROMMONkit method, to directly manipulate running operating system code in memory. This malicious code in the bootloader provides the capability of direct memory manipulation to the adversary, allowing them to patch the live operating system during runtime. By modifying the instructions stored in the system image file, adversaries may either weaken existing defenses or provision new capabilities that the device did not have before. Examples of existing defenses that can be impeded include encryption, via Weaken Encryption, authentication, via Network Device Authentication, and perimeter defenses, via Network Boundary Bridging. Adding new capabilities for the adversary’s purpose include Keylogging, Multi-hop Proxy, and Port Knocking. Adversaries may also compromise existing commands in the operating system to produce false output to mislead defenders. When this method is used in conjunction with Downgrade System Image, one example of a compromised system command may include changing the output of the command that shows the version of the currently running operating system. By patching the operating system, the adversary can change this command to instead display the original, higher revision number that they replaced through the system downgrade. When the operating system is patched in storage, this can be achieved in either the resident storage (typically a form of flash memory, which is non-volatile) or via TFTP Boot. When the technique is performed on the running operating system in memory and not on the stored copy, this technique will not survive across reboots. However, live memory modification of the operating system can be combined with ROMMONkit to achieve persistence.

Adversaries may install an older version of the operating system of a network device to weaken security. Older operating system versions on network devices often have weaker encryption ciphers and, in general, fewer/less updated defensive features. (Citation: Cisco Synful Knock Evolution) On embedded devices, downgrading the version typically only requires replacing the operating system file in storage. With most embedded devices, this can be achieved by downloading a copy of the desired version of the operating system file and reconfiguring the device to boot from that file on next system restart. The adversary could then restart the device to implement the change immediately or they could wait until the next time the system restarts. Downgrading the system image to an older versions may allow an adversary to evade defenses by enabling behaviors such as Weaken Encryption. Downgrading of a system image can be done on its own, or it can be used in conjunction with Patch System Image.

Adversaries may deploy a container into an environment to facilitate execution or evade defenses. In some cases, adversaries may deploy a new container to execute processes associated with a particular image or deployment, such as processes that execute or download malware. In others, an adversary may deploy a new container configured without network rules, user limitations, etc. to bypass existing defenses within the environment. In Kubernetes environments, an adversary may attempt to deploy a privileged or vulnerable container into a specific node in order to Escape to Host and access other containers running on the node. (Citation: AppSecco Kubernetes Namespace Breakout 2020) Containers can be deployed by various means, such as via Docker's <code>create</code> and <code>start</code> APIs or via a web application such as the Kubernetes dashboard or Kubeflow. (Citation: Docker Containers API)(Citation: Kubernetes Dashboard)(Citation: Kubeflow Pipelines) In Kubernetes environments, containers may be deployed through workloads such as ReplicaSets or DaemonSets, which can allow containers to be deployed across multiple nodes.(Citation: Kubernetes Workload Management) Adversaries may deploy containers based on retrieved or built malicious images or from benign images that download and execute malicious payloads at runtime.(Citation: Aqua Build Images on Hosts)

Adversaries may build a container image directly on a host to bypass defenses that monitor for the retrieval of malicious images from a public registry. A remote <code>build</code> request may be sent to the Docker API that includes a Dockerfile that pulls a vanilla base image, such as alpine, from a public or local registry and then builds a custom image upon it.(Citation: Docker Build Image) An adversary may take advantage of that <code>build</code> API to build a custom image on the host that includes malware downloaded from their C2 server, and then they may utilize Deploy Container using that custom image.(Citation: Aqua Build Images on Hosts)(Citation: Aqua Security Cloud Native Threat Report June 2021) If the base image is pulled from a public registry, defenses will likely not detect the image as malicious since it’s a vanilla image. If the base image already resides in a local registry, the pull may be considered even less suspicious since the image is already in the environment.

Adversaries may reflectively load code into a process in order to conceal the execution of malicious payloads. Reflective loading involves allocating then executing payloads directly within the memory of the process, vice creating a thread or process backed by a file path on disk (e.g., Shared Modules). Reflectively loaded payloads may be compiled binaries, anonymous files (only present in RAM), or just snubs of fileless executable code (ex: position-independent shellcode).(Citation: Introducing Donut)(Citation: S1 Custom Shellcode Tool)(Citation: Stuart ELF Memory)(Citation: 00sec Droppers)(Citation: Mandiant BYOL) For example, the `Assembly.Load()` method executed by PowerShell may be abused to load raw code into the running process.(Citation: Microsoft AssemblyLoad) Reflective code injection is very similar to Process Injection except that the “injection” loads code into the processes’ own memory instead of that of a separate process. Reflective loading may evade process-based detections since the execution of the arbitrary code may be masked within a legitimate or otherwise benign process. Reflectively loading payloads directly into memory may also avoid creating files or other artifacts on disk, while also enabling malware to keep these payloads encrypted (or otherwise obfuscated) until execution.(Citation: Stuart ELF Memory)(Citation: 00sec Droppers)(Citation: Intezer ACBackdoor)(Citation: S1 Old Rat New Tricks)

Adversaries may employ various means to detect and avoid debuggers. Debuggers are typically used by defenders to trace and/or analyze the execution of potential malware payloads.(Citation: ProcessHacker Github) Debugger evasion may include changing behaviors based on the results of the checks for the presence of artifacts indicative of a debugged environment. Similar to Virtualization/Sandbox Evasion, if the adversary detects a debugger, they may alter their malware to disengage from the victim or conceal the core functions of the implant. They may also search for debugger artifacts before dropping secondary or additional payloads. Specific checks will vary based on the target and/or adversary. On Windows, this may involve Native API function calls such as <code>IsDebuggerPresent()</code> and <code> NtQueryInformationProcess()</code>, or manually checking the <code>BeingDebugged</code> flag of the Process Environment Block (PEB). On Linux, this may involve querying `/proc/self/status` for the `TracerPID` field, which indicates whether or not the process is being traced by dynamic analysis tools.(Citation: Cado Security P2PInfect 2023)(Citation: Positive Technologies Hellhounds 2023) Other checks for debugging artifacts may also seek to enumerate hardware breakpoints, interrupt assembly opcodes, time checks, or measurements if exceptions are raised in the current process (assuming a present debugger would “swallow” or handle the potential error).(Citation: hasherezade debug)(Citation: AlKhaser Debug)(Citation: vxunderground debug) Malware may also leverage Structured Exception Handling (SEH) to detect debuggers by throwing an exception and detecting whether the process is suspended. SEH handles both hardware and software expectations, providing control over the exceptions including support for debugging. If a debugger is present, the program’s control will be transferred to the debugger, and the execution of the code will be suspended. If the debugger is not present, control will be transferred to the SEH handler, which will automatically handle the exception and allow the program’s execution to continue.(Citation: Apriorit) Adversaries may use the information learned from these debugger checks during automated discovery to shape follow-on behaviors. Debuggers can also be evaded by detaching the process or flooding debug logs with meaningless data via messages produced by looping Native API function calls such as <code>OutputDebugStringW()</code>.(Citation: wardle evilquest partii)(Citation: Checkpoint Dridex Jan 2021)

Adversaries may modify property list files (plist files) to enable other malicious activity, while also potentially evading and bypassing system defenses. macOS applications use plist files, such as the <code>info.plist</code> file, to store properties and configuration settings that inform the operating system how to handle the application at runtime. Plist files are structured metadata in key-value pairs formatted in XML based on Apple's Core Foundation DTD. Plist files can be saved in text or binary format.(Citation: fileinfo plist file description) Adversaries can modify key-value pairs in plist files to influence system behaviors, such as hiding the execution of an application (i.e. Hidden Window) or running additional commands for persistence (ex: Launch Agent/Launch Daemon or Re-opened Applications). For example, adversaries can add a malicious application path to the `~/Library/Preferences/com.apple.dock.plist` file, which controls apps that appear in the Dock. Adversaries can also modify the <code>LSUIElement</code> key in an application’s <code>info.plist</code> file to run the app in the background. Adversaries can also insert key-value pairs to insert environment variables, such as <code>LSEnvironment</code>, to enable persistence via Dynamic Linker Hijacking.(Citation: wardle chp2 persistence)(Citation: eset_osx_flashback)

Adversaries may impersonate a trusted person or organization in order to persuade and trick a target into performing some action on their behalf. For example, adversaries may communicate with victims (via Phishing for Information, Phishing, or Internal Spearphishing) while impersonating a known sender such as an executive, colleague, or third-party vendor. Established trust can then be leveraged to accomplish an adversary’s ultimate goals, possibly against multiple victims. In many cases of business email compromise or email fraud campaigns, adversaries use impersonation to defraud victims -- deceiving them into sending money or divulging information that ultimately enables Financial Theft. Adversaries will often also use social engineering techniques such as manipulative and persuasive language in email subject lines and body text such as `payment`, `request`, or `urgent` to push the victim to act quickly before malicious activity is detected. These campaigns are often specifically targeted against people who, due to job roles and/or accesses, can carry out the adversary’s goal. Impersonation is typically preceded by reconnaissance techniques such as Gather Victim Identity Information and Gather Victim Org Information as well as acquiring infrastructure such as email domains (i.e. Domains) to substantiate their false identity.(Citation: CrowdStrike-BEC) There is the potential for multiple victims in campaigns involving impersonation. For example, an adversary may Compromise Accounts targeting one organization which can then be used to support impersonation against other entities.(Citation: VEC)

Adversaries may attempt to modify hierarchical structures in infrastructure-as-a-service (IaaS) environments in order to evade defenses. IaaS environments often group resources into a hierarchy, enabling improved resource management and application of policies to relevant groups. Hierarchical structures differ among cloud providers. For example, in AWS environments, multiple accounts can be grouped under a single organization, while in Azure environments, multiple subscriptions can be grouped under a single management group.(Citation: AWS Organizations)(Citation: Microsoft Azure Resources) Adversaries may add, delete, or otherwise modify resource groups within an IaaS hierarchy. For example, in Azure environments, an adversary who has gained access to a Global Administrator account may create new subscriptions in which to deploy resources. They may also engage in subscription hijacking by transferring an existing pay-as-you-go subscription from a victim tenant to an adversary-controlled tenant. This will allow the adversary to use the victim’s compute resources without generating logs on the victim tenant.(Citation: Microsoft Peach Sandstorm 2023)(Citation: Microsoft Subscription Hijacking 2022) In AWS environments, adversaries with appropriate permissions in a given account may call the `LeaveOrganization` API, causing the account to be severed from the AWS Organization to which it was tied and removing any Service Control Policies, guardrails, or restrictions imposed upon it by its former Organization. Alternatively, adversaries may call the `CreateAccount` API in order to create a new account within an AWS Organization. This account will use the same payment methods registered to the payment account but may not be subject to existing detections or Service Control Policies.(Citation: AWS RE:Inforce Threat Detection 2024)

Adversaries may fake, or spoof, a sender’s identity by modifying the value of relevant email headers in order to establish contact with victims under false pretenses.(Citation: Proofpoint TA427 April 2024) In addition to actual email content, email headers (such as the FROM header, which contains the email address of the sender) may also be modified. Email clients display these headers when emails appear in a victim's inbox, which may cause modified emails to appear as if they were from the spoofed entity. This behavior may succeed when the spoofed entity either does not enable or enforce identity authentication tools such as Sender Policy Framework (SPF), DomainKeys Identified Mail (DKIM), and/or Domain-based Message Authentication, Reporting and Conformance (DMARC).(Citation: Cloudflare DMARC, DKIM, and SPF)(Citation: DMARC-overview)(Citation: Proofpoint-DMARC) Even if SPF and DKIM are configured properly, spoofing may still succeed when a domain sets a weak DMARC policy such as `v=DMARC1; p=none; fo=1;`. This means that while DMARC is technically present, email servers are not instructed to take any filtering action when emails fail authentication checks.(Citation: Proofpoint TA427 April 2024)(Citation: ic3-dprk) Adversaries may abuse Microsoft 365’s Direct Send functionality to spoof internal users by using internal devices like printers to send emails without authentication.(Citation: Barnea DirectSend) Adversaries may also abuse absent or weakly configured SPF, SKIM, and/or DMARC policies to conceal social engineering attempts(Citation: ic3-dprk) such as Phishing. They may also leverage email spoofing for Impersonation of legitimate external individuals and organizations, such as journalists and academics.(Citation: ic3-dprk)

Adversaries may employ various time-based methods to evade detection and analysis. These techniques often exploit system clocks, delays, or timing mechanisms to obscure malicious activity, blend in with benign activity, and avoid scrutiny. Adversaries can perform this behavior within virtualization/sandbox environments or natively on host systems. Adversaries may utilize programmatic `sleep` commands or native system scheduling functionality, for example Scheduled Task/Job. Benign commands or other operations may also be used to delay malware execution or ensure prior commands have had time to execute properly. Loops or otherwise needless repetitions of commands, such as `ping`, may be used to delay malware execution and potentially exceed time thresholds of automated analysis environments.(Citation: Revil Independence Day)(Citation: Netskope Nitol) Another variation, commonly referred to as API hammering, involves making various calls to Native API functions in order to delay execution (while also potentially overloading analysis environments with junk data).(Citation: Joe Sec Nymaim)(Citation: Joe Sec Trickbot)

Adversaries may intentionally exclude certain files, folders, directories, file types, or system components from encryption or tampering during a ransomware or malicious payload execution. Some file extensions that adversaries may avoid encrypting include `.dll`, `.exe`, and `.lnk`.(Citation: Palo Alto Unit 42 Medusa Group Medusa Ransomware January 2024) Adversaries may perform this behavior to avoid alerting users, to evade detection by security tools and analysts, or, in the case of ransomware, to ensure that the system remains operational enough to deliver the ransom notice. Exclusions may target files and components whose corruption would cause instability, break core services, or immediately expose the attack. By carefully avoiding these areas, adversaries maintain system responsiveness while minimizing indicators that could trigger alarms or otherwise inhibit achieving their goals.