Although data exfiltration is a cornerstone of malicious cyber operations, its nature and impact have changed significantly over the past few years.
Historically, adversaries linked to state sponsors or similar activities would engage in data theft to further goals of industrial espionage or intellectual property theft. More recently, adversaries have engaged in exfiltration, based not on the value of the data to themselves or their sponsors, but rather on the desire of victims not to have their data exposed.
Gigamon expert Joe Slowik says: “The result is an ecosystem of exfiltration and disclosure for either monetary or ‘hacktivist’ purposes, where the malicious party seeks to damage, or threaten to damage, the victim.”
Recent occurrences include ransomware-related data loss at meat processor JBS, which prompted the company to pay $11 million to avoid disclosure of data, and an intrusion at gaming company Electronic Arts resulting in the loss and potential sale of nearly 800 gigabytes of sensitive information.
Given these trends, network owners, operators and defenders must work to ensure controls are in place to detect, mitigate, or outright block such activity given potential costs.
On its face, exfiltration appears less concerning than other cyber-nexus impacts such as network disruption. While data exfiltration typically has some immediate impact on an organisation, long-term hazards may make loss of sensitive data even more destructive than an especially virulent ransomware incident.
Three primary risks adhere to data loss or exfiltration events:
- Loss of sensitive data or intellectual property, which can place an organisation at a competitive disadvantage or even induce regulatory risk in the case of sensitive customer or client data
- Damage to an organisation’s reputation following a publicised incident, including but not limited to public relations work to reduce impacts and long-term association with a major, high-profile data loss event
- Opening up the organisation to extortion by actors who extract data, then threaten to publish it absent payment or some other action
The first case constitutes the typical focus of data loss concern, as this relates to incidents ranging from state-sponsored industrial espionage to criminal entities attempting to steal banking or personal information.
The second adheres to all (publicly revealed or acknowledged) exfiltration incidents, and forms a difficult-to-quantify cost that may take years to overcome.
The third is a relatively recent development that aligns with emerging trends in “double-extortion” ransomware and similar criminal activity, although such tactics possess a longer history among politically motivated hacktivist entities.
Overall, while risks are significant, their quantification and time to manifestation are typically unclear, making these items hard to assess. Theft of intellectual property may have impacts felt only years or decades after the action takes place, while potential reputational losses may be impossible to adequately quantify under any circumstance.
Yet given an increase in both hacktivist and especially ransomware operations incorporating data leak or data extortion activities, data exfiltration activity is rapidly becoming a more acute and immediate issue than in years past.
So network defenders and operators must align defences in such a fashion to identify and potentially even prevent such activity where appropriate.
Identifying exfiltration behaviours
Exfiltration activity would be expected to be noisy and relatively easy to spot, given the movement of large volumes of data to new or unfamiliar sources. Yet even making this determination requires a combination of network visibility and active network monitoring, such that most organisations either lack insight into such behaviours, or such activities blend in sufficiently with legitimate activity so as to produce significant ‘noise’ in detections.
What’s more, adversaries of various types employ multiple techniques to minimise visibility and evade detection, either through obfuscation or blending in with other actions. Examples include:
# Using legitimate third-party services, such as cloud backup systems or web-based storage, as destinations for leaked data. Examples range from common items such as Google Drive and Dropbox to more specialised products such as an ecosystem of items related to the Mega.io service.
# Tunnelling traffic over non-HTTP services or using alternative protocols for large data transfer that might not be monitored with the same level of scrutiny.
# Dividing data into smaller pieces for exfiltration to avoid abnormally large traffic flows leaving the network.
Although all of these are concerning, none are impossible to detect. Rather, defenders can employ a combination of monitoring for more general traffic pattern anomalies along with specific identification of certain techniques or behaviours to detect this type of activity.
Even in cases where data is encoded or encrypted, or where visibility into network activity is somewhat limited, access to data sets such as network flow provides a number of possibilities to flag suspicious operations.
At the most general level, identification of suspicious network flows represents a potentially powerful technique for identifying exfiltration activity. Aside from just looking for large data flows, with ‘large’ likely dependent on the normal operations and expectations within the monitored network, identification of directionality and upload/download ratios are necessary enrichments to accurately leverage flow monitoring.
For example, a ‘large’ flow may just be indicative of a long-running, data-rich connection such as streaming, remote access, or related activity. However, when also looking at directionality and data ratios, identifying large flows where the majority of data (80–90 percent) leaves the network can flag a large upload session.
While there may certainly be legitimate use cases for such activity, such as sharing large project files or datasets, these flows may also be indicative of data loss or other unauthorised activity.
All this becomes even more powerful when combined with an analytic approach to network connections. Identifying not just anomalous, outbound flows but connecting these to new, previously unobserved or suspicious network infrastructure enables powerful detection possibilities.
For example, linking a suspicious outbound flow to new network infrastructure or a virtual private server (VPS) instance such as Linode, DigitalOcean can identify activity of concern. Extending further, the same logic can also be applied to nonstandard connections like external FTP or other protocols to unknown or untrusted destinations to identify potential exfiltration activity.
While these strategies are effective in identifying exfiltration occurs, they are lacking since such approaches are fundamentally backward looking. In other words, these items will identify exfiltration as or shortly after it actually takes place.
This can be of value to alert organisations that something is wrong, enabling response and mitigation operations and reducing time to detection and time to response metrics. Yet ideally defenders and their respective employers would be able to prevent exfiltration from taking place.
Applying ‘whole of kill chain’ defence
Preventing exfiltration requires a ‘whole of cyber kill chain’ perspective on network monitoring and defence. In this scenario, defenders look for overall adversary intrusion pathways and dependencies to identify necessary precursors such as initial access, lateral movement, data collection, and data staging, prior to data leaving the network.
Through a critical examination of what an adversary requires to succeed, we can begin implementing more general controls across both host and network capabilities to defeat not just exfiltration operations but a variety of intrusions.
At minimum, identifying likely adversary ingress mechanisms and either monitoring them closely or implementing controls to reduce attack surface can significantly reduce the likelihood of a threat actor gaining access to the environment.
Relevant controls include patching external-facing systems, reducing available services from external access, limiting traffic types inbound to the network, and monitoring sensitive activity such as remote administration or access sessions.
Many organisations focus on these steps as a means to reduce risk, but true layered defence requires going beyond the network boundary to ensure coverage across subsequent adversary actions as well.
Given a near-infinite number of possible initial access mechanisms available to adversaries, as well as the possibility of subverting systems or users even when items are patched or otherwise monitored, defence must extend into internal network traffic flows and host items.
Defenders must identify lateral movement techniques and their artifacts across both network and host behaviours. While increasing adoption of endpoint defence and response (EDR) products covers host-centric observations, investment in East-West traffic visibility and monitoring is essential to capture adversary traversal of defended networks.
Implemented together, these items can ensure coverage of adversary operations ranging from opportunistic criminal actors to focused hacktivists to state-sponsored threats.
Overall, defenders need to combine an understanding of adversary behaviours and tendencies, such as through production or ingestion of cyber threat intelligence, to first determine how adversaries operate and what techniques are relevant in intrusion operations.
Once determined, a combination of EDR and network defence and response (NDR) must be employed to ensure layered detection and monitoring and cover any potential gaps in visibility that adversaries may attempt to use to their advantage.
Data exfiltration is increasingly incorporated into the actions of a variety of threats, from criminals to hacktivists to state-directed intrusions. Through the application and monitoring of robust network controls, defenders and asset owners can ensure awareness of such behaviours when they occur, allowing for quicker response and potentially event mitigation.
Defenders cannot focus solely on detecting anomalous outbound data flows. Instead, robust defence requires identifying adversary actions across all phases of intrusion operations.
Once understood, defenders can implement network and host controls or visibility to cover each step of an intrusion. Only through this robust, defence-in-depth approach can defenders ensure not just awareness of potential malicious activity (including but not limited to data exfiltration operations), but also enable possible intrusion interdiction or disruption when actions are caught early in the adversary’s lifecycle.
While neither easy nor inexpensive, defence against modern cyber threats, from ransomware operations to data exfiltration for a variety of purposes, demands sufficient investment to ensure defenders keep pace with a rapidly evolving threat landscape.