Analytical Model - Tags as Analysis
Analysis consists of collecting and evaluating data and drawing conclusions based on the data available to you. Assuming data is collected and modeled accurately within Synapse, the data itself - nodes and their properties - should not change. But as you collect more data or re-evaluate existing data, your assessment of that data - often encoded in tags - may change over time. Nodes and properties are largely stable; tags are meant to be flexible and readily modified if needed.
Every knowledge domain has its own focus and set of analytical questions it attempts to answer. The “answers” to some of these questions can be recorded in Synapse as tags applied to relevant nodes. These tags provide context to the data in Synapse.
The Synapse data model for tags is simple - it consists of the single
syn:tag form. The appropriate use of
tags to annotate data is more nuanced. You can think of tags - their structure and application - as an
analytical model that complements and extends the power of the data model.
This analytical model:
Is largely independent from the data model. You do not need to write code to implement new tags or design a tag structure; you simply need to create the appropriate
Is specific to an analytical discipline. Tags used for cyber threat analysis may be very different from tags used to track financial fraud.
Is tightly coupled with the specific questions you want to answer. The tags you create and apply should be driven by your particular analysis goals.
The tags should annotate assessments and conclusions that are important to your analysis.
The tags should allow you to ask meaningful questions of your data.
The effective use of Synapse to conduct analysis depends on:
The data model: how you represent the data you care about using Synapse’s forms, properties, and types.
The analytical model: how you design and use a set of tags to annotate data, provide context, and answer the questions that matter to you.
The sections below provide a few examples of the kinds of context - observations and assessments - that you can represent with tags. Recording these assessments in Synapse alongside the relevant data provides immediate context to that data, and allows you to query both data (nodes) and assessments (tags).
These examples are simply meant to illustrate a few possible “real-world” applications for tags. There is no “right” or “wrong” set of tags (although there are “better” and “worse” design decisions that may impact your ability to answer questions efficiently).
See Design Concepts - Analytical Model for considerations in designing tags and tag trees.
See Design Concepts - General for considerations on whether to model something as a form, a property, a tag, or a tag associated with a node.
See Tags Associated with Forms for Synapse’s ability to link tags to nodes to more easily cross-reference tags with data model elements that those tags represent.
The purpose of analysis is to draw relevant conclusions from the data at hand. The conclusions will vary based on the knowledge domain, but could include big-picture assessments such as “The increase in widget manufacturing due to lower production costs has had a negative effect on the demand for gizmos” or “The threat group Vicious Wombat is working on behalf of the Derpistan government”.
Those large assessments can be made based on numerous smaller assessments (tags) which are themselves based on the observables (nodes) in Synapse. To build up to those larger assessments, you must start by recording those smaller assessments as tags on nodes.
The folowing examples from cyber threat intelligence illustrate some of the assessments that can be recorded using tags.
The specific tags referenced below are based on The Vertex Project’s tag trees and use our conventions. Use what works for you!
A common practice in threat intelligence involves deciding not only whether an indicator (such as a file, domain, or IP address) is malicious, but whether it should be associated with a threat cluster. That is, can an indicator be linked to other indicators (e.g., from the same indcident or intrusion) to create a known set of related indicators and activity. “Threat clusters” may be built up and expanded over time to represent a broader set of activity presumed to be carried out by some (generally unknown) set of malicious actors (a “threat group”).
You can tag nodes to indicate that the node is associated with a particular threat cluster. For example:
cno is a top-level tag for assessments related to Computer Network Operations (CNO),
threat is a
sub-tag used for threat clusters / threat groups, and
<cluster> is the “name” of the particular threat cluster
based on your organization’s conventions (names, numbers, etc.)
Tactics, Techniques, and Procedures (TTPs)
The methodologies (sometimes known as tactics, techniques, and procedures or TTPs) that a threat group uses to conduct activity provide insight into the group and its operations. Knowledge of past TTPs may help predict future actions or operations. Sets of TTPs observed together may provide a “fingerprint” of a group’s activity. General knowledge of TTPs in current use can help organizations more effectively protect and defend their assets.
“TTP” can cover a broad range of observed activity, from whether a group uses zero-day exploits to the specific packer used to obfuscate a piece of malware. When a node represents an instance of the use of a TTP, it may be useful to tag the node with the TTP in question.
For example, you have an email message (RFC822 file) that you assess is a phishing attack. You can tag the relevant
node or nodes (such as the
file:bytes of the message and / or the
inet:email:message node representing the
message metadata) with that TTP:
cno is our top-level tag,
ttp represents the TTP sub-tree,
phish represents assessments related to
message indicates the node(s) represent the phishing email (e.g., as opposed to an attachment or
URL representing the phishing
payload, or the sending email address or IP representing the
Some third-party data sources provide both data and tags or labels associated with that data. For example, Shodan may
provide data on an IPv4 address (such as which ports were open as of the last Shodan scan) as well as tags such as
vpn. Similarly, VirusTotal may provide metadata and multiscanner data for files along with
tags such as
In addition, many commercial organizations conduct their own threat tracking and analysis and publish their research. This type of research commonly includes “indicators of compromise” or IOCs - hashes, domains, IP addresses, and so on indicative of the reported activity. These reports do not necessarily include tags provided by the reporting organization. But the report may make it clear that the reporter associates the IOCs with particular malware families, “campaigns”, or threat groups.
Shodan’s label indicating that an IPv4 address hosted a VPN and ESET’s reporting that a SHA1 hash is associated with the X-Agent malware family are both assertions. These assertions are valuable data and can be useful to your analysis.
That said, you may not have the means to verify these assertions yourself. To accept the assertion at face value means you need to trust the third-party in question. “Trust” may include things like understanding the source of the data; knowing their general reputation (i.e., within your analysis community); or building trust over time as you determine the reliabilty and accuracy of their reporting.
Your own assertions are presumably “more trustworthy” based on direct access to your internal data and processes. Assertions made by others may be open to question or validation, so it can be useful to record these third-party assessments separately. This allows you to retain the context of what “other people” say while keeping those (potentially lower-confidence) assertions separate from your own.
You can use tags to annotate “other people’s analysis” by tagging relevant nodes with what “other people” say about them:
rep.eset.sednit: ESET says this SHA1 hash is associated with Sednit
rep.shodan.vpn: Shodan says this IPv4 hosts a VPN
rep.vt.peexe: VirusTotal says this file is a PE executable
rep is a top-level tag for third-party reporting, the second tag element (e.g.,
eset) is the name
of the reporting organization, and the third tag element is the information the third party is reporting.
Within a particular knowledge domain, it may be useful to record observations that support your analysis process in some way. In other words, the observations are relevant to your analysis, but do not represent the specific output or objective of your analysis.
In cyber threat intelligence, a primary goal is to track malicious activity and maintain awareness of the current threat landscape, often in terms of malware, threat groups, and techniques / TTPs. Part of this tracking includes noting infrastructure (such as IP addresses, netblocks, or domains) used in malicious activity.
Identifying network infrastructure as TOR nodes, anonymous VPN endpoints, or sinkhole IPs is not a primary goal of threat intelligence, but knowing this information can be useful and help prevent analysts from mis-identifying threat actor infrastructure.
You can use tags to annotate identified infrastructure (such as
inet:ipv4 nodes) of interest:
cno.infra.anon.tor: The IPv4 is a TOR exit node
cno.infra.anon.vpn: The IPv4 is an anonymous VPN exit point
cno.infra.dns.sink.hole: The IPv4 is used to resolve sinkholed FQDNs
cno is our top-level tag for Computer Network Operations,
infra indicates the “infrastructure”
sub-tree, the third element indicates the kind of infrastructure (
anon for anonymous,
dns for DNS, etc.),
and so on.
Tags as Hypotheses
Another way to look at tags is as hypotheses. If a tag represents the outcome of an assessment, then every tag can be seen as having an underlying question - a hypothesis - it is attempting to answer. Deciding to apply the tag is equivalent to deciding that the underlying hypothesis is true.
Making these assessments typically involves the judgment of a human analyst; so evaluating and tagging data within Synapse is one of an analyst’s primary tasks.
Hypotheses may be simple or complex; tags typically represent relatively simple concepts that are used collectively to support (or refute) more complex theories. Because the concept of encoding analytical conclusions within a system like Synapse may be unfamiliar, a few examples may be helpful.
The question “can this newly identified FQDN be associated with any known threat cluster?” can be thought of as n number of individual hypotheses based on the number of known threat clusters:
Hypothesis 1: This domain is associated with Threat Cluster 1.
Hypothesis 2: This domain is associated with Threat Cluster 2.
Hypothesis n: This domain is associated with Threat Cluster n.
If an analyst determines that the domain is associated with Threat Cluster 46, placing a Threat Cluster 46 tag (e.g.,
cno.threat.t46) on that FQDN effectively means that the hypothesis “This domain is associated with
Threat Cluster 46” has been assessed to be true (and by implication, that all competing hypotheses are false).
Deciding whether a domain is meant to imitate (masquerade as) a legitimate domain for malicious purposes can also be thought of as a set of hypotheses.
“Masquerading” is a threat actor technique (TTP) designed to influence a targeted user to trust something enough to perform an action. A domain that “looks like” a valid FQDN or an email address that “looks like” a trusted sender may encourage the victim to click a link or open an attachment. In threat intelligence, the focus is on threat actor TTPs, so the TTPs we’re interested in are (by definition) malicious.
Let’s say an analyst comes across the suspicious domain
akcdndata.com. To decide whether this is an example of a
masquerade, the analyst needs to decide:
Is the FQDN
akcdndata.comassociated with known malicious activity?
Does the FQDN
akcdndata.comimitate a legitimate company, site, or service?
A number of possibilities (hypotheses) exist, such as:
Hypothesis 1: The domain is NOT malicious.
Hypothesis 2: The domain IS malicious, but is not meant to imitate anything.
Hypothesis 3: The domain IS malicious, and is meant to imitate a legitimate resource.
The tag (or tags) the analyst decides to apply depend on which hypotheses they can prove or disprove (assert are true, or not).
Deciding on Hypothesis 1 vs. Hypothesis 2 may involve things like reviewing domain registration data, associated DNS infrastructure, or seeing if the FQDN shows up in public reporting of malicious activity.
If Hypothesis 1 is true, we would not tag the FQDN. If Hypothesis 2 is true, we can simply assert that the FQDN is
malicious (with a tag such as
If Hypothesis 2 is true, deciding on Hypothesis 3 may be trickier. Does the FQDN “look like” anything familiar? It may “look like” Akamai CDN (content delivery network) but that’s a bit of a stretch…maybe it is just a coincidence? Do we have any context around how the FQDN was used maliciously that might indicate that the threat actors wanted to mislead victims into thinking the FQDN was associated with Akamai?
If we have enough evidence to support Hypothesis 3, we can apply a TTP tag such as
our top-level tag,
ttp for our TTP sub-tree,
se for social engineering TTPs, and
masq for masquerade).
Individual Hypotheses to Broader Reasoning
The hypotheses represented by the tags in the examples above are fairly narrow in scope - an indicator is
associated with a threat cluster (
cno.threat.t42), a domain was designed to mislead users by imitating a
legitimate web site or service (
cno.ttp.se.masq). With Synapse, you can leverage these more focused
hypotheses to answer broader, more complex questions.
A newly identified zero-day exploit has been circulating in the wild and is in use by multiple threat groups. The associated vulnerability has been assigned CVE-2021-9999 (a number we made up). The exploit is delivered via a malicious XLSX file sent as an email (phishing) attachment.
You believe that “Threat Group 12 was the first group to use the zero day associated with CVE-2021-9999”. To prove
or disprove this hypothesis, you could query Synapse for all files (
file:bytes nodes) that:
exploit CVE-2021-9999 (i.e., have a tag such as
are associated with a known threat cluster or threat group (i.e., are tagged
If you have data for any associated phishing messages, you can pivot from the malicious XLSX files to their
associated emails (
inet:email:message:attachment -> inet:email:message) and look for the phishing message
with the oldest date. By identifying the threat group associated with the earliest known email, you can determine
whether the zero-day was first used by Threat Group 12 or some other group.
You are able to take tags associated with simple assessments (“this file exploits CVE-2021-9999” or “this file is
associated with Threat Cluster 12”) and combine nodes (files /
file:bytes), properties (
and tags to answer a more complex question. That’s the power of Synapse!
This example is simplified; you would of course perform additional research besides what is described above (such as searching for additional samples that exploit the vulnerability and any associated phishing attempts, attributing identified samples that are not yet associated with a Threat Cluster, etc.)
Assuming you have completed your research and the data is in Synapse and tagged appropriately, you can easily answer the above question using the Storm query language using a query such as the following:
file:bytes#rep.vt.cve_2021_9999 +#cno.threat -> inet:email:message:attachment -> inet:email:message | min :date | -> # +syn:tag^=cno.threat