Thursday, November 26, 2020

, , , , , , ,

Using similarity to expand context and map out threat campaigns

TL;DR: VirusTotal allows you to search for similar files according to different orthogonal notions (structure, visual layout, icons, execution behaviour, etc.). File similarity can be combined with the “have:” search modifier in order to gain more context about threats, e.g. what are the emails or URLs that distribute them.

This is the second blog post in our similarity series, the first article focused on how to trigger file similarity searches and the different similarity vectors at your disposal. In the context of this series we have also done a webinar that can be viewed on-demand, it focuses on using similarity to automatically produce optimal YARA rules to detect a given malware framework/family/campaign via VTDIFF.

This situation might sound familiar. As a SOC analyst or Incident Responder you are often confronted with files you know nothing about. Your SIEM describes their internal sightings and actions but fails to transmit the bigger picture. You are constrained by the narrow visibility of your corporate logs. Context is king and the problem is that you are fighting threat actors that operate globally with just a piece of the puzzle, your local data.

What is this file? Who is behind it? What is their modus operandi? How did it get there? Are there other related components? What does it do? Are there other variants that could have impacted my organization in the past? Any that could impact us in the future? How do I contain it? Your SIEM, case management system, EDR, firewall, IDS etc. don’t answer these questions. You are missing a necessary layer in your defense-in-depth security strategy.

VirusTotal is your saving grace. You jump into VT ENTERPRISE and look up the hash: threat reputation is useful, but you need further context. Your task is to identify IoCs that can be used for remediation, e.g. by blocking a command-and-control domain in the network perimeter, as well as artefacts that can be used for proactive threat hunting purposes, to determine whether there has been a breach and what is its scope. The issue is that sometimes VirusTotal does not have full context for a specific individual file in terms of sandbox reports, in-the-wild sightings, relationships, etc. and so your investigation might end here.

How to do it better

Isolated hashes are of limited value. Many times they are unique per victim or campaign, so a better idea would be finding the cluster/family/campaign they belong to in order to unearth remediation IoCs and threat hunting patterns. Most importantly, you need to leverage those groupings in order to surface command-and-control domains, dropzones, distribution URLs, phishing emails, etc. that can be used for mitigation and containment, and, to build proper understanding and situational awareness.

Similarity and the “have” search modifier to the rescue. Let’s imagine the initial hash that popped up as an alert in our environment was a first stage EMOTET dropper, i.e. a document that delivers a malicious payload through macros.

Threat reputation allows you to perform an immediate first assessment (alert triage), but other than that there is little context in terms of remediation IoCs and hunting artifacts. We still know nothing about how this file gets distributed, i.e. its delivery vector. Similarly, we fully ignore whether this is something spear phished exclusively against our organization or part of a larger campaign. What about the threat network infrastructure? Does it download additional payloads? Does it communicate with a command-and-control?

The next step in an incident response engagement - and this is what most analysts fail to do - is to jump into the file’s cluster (its family/framework/campaign) in order to expand context and surface IoCs. This is just one click away:

For documents there is a limited number of approaches to find similar files (other file formats will expose more), this said, they are very rich because they are fully orthogonal: structural features, visual layout, local sensitive fuzzy hashing, execution behaviour similarity. Let’s jump to other similar files based on the document’s visual layout by clicking on “Similar by icon/thumbnail” or on the thumbnail itself, located in the top right: main_icon_dhash:23232b2b00010000.

There are too many matches, we would have to iterate over every single one in order to surface particular patterns that may allow us to understand the campaign.

Finding phishing emails that distribute the threat

We can narrow down the search above to match exclusively those files that have been seen as an attachment in some email uploaded to VirusTotal:

main_icon_dhash:23232b2b00010000 AND have:email_parents
(Note that you can also use tag:attachment instead of have:email_parents)

We can now run through the matching files, open up their Relations tab and jump into the pertinent email parent, so as to understand the deception techniques being used in the campaign:

This particular instance poses as some kind of World Health Organization report on COVID. It is important to inspect all the other emails because not only will they tell us more about the lures, it will also allow us to identify targeted industries, geographical spread, activity time spans, etc. For instance, there could be other localized variants that could be targeting some other corporate branches. Access to these emails will not only give us greater insight into the attacker, it is also something we can leverage tactically in order to improve filtering in our email gateways.

Discovering URLs that distribute this threat

We want to see if this campaign is also being distributed via download URLs. If that´s the case we can block them in our network perimeter or use them to search across web proxy logs. Let’s ask VirusTotal whether any of the files in the cluster have associated in-the-wild URLs:
main_icon_dhash:23232b2b00010000 AND have:itw

We can now jump into the Relations tab in order to export these additional IoCs:

There are over 3K files with in-the-wild URLs, note that we can automate all of this via the API.

Identifying command-and-control/exfiltration infrastructure

The next step is to understand whether any of the machines in our corporate fleet are beaconing out to infrastructure tied to this campaign. At the same time, we will probably want to block the CnC and exfiltration points in order to mitigate the impact of historical undetected breaches. Let’s filter down the search to focus exclusively on those files that exhibited network communications when executed in a dynamic analysis sandbox:

main_icon_dhash:23232b2b00010000 AND have:behaviour_network

Most of the matching files have been analysed by several sandboxes participating in our multi-sandbox effort. This gives us unparalleled visibility into the campaign. For an attacker it is easy to evade a single sandbox, it is far more complex to do so for 17+ of them at the same time. Each one of them set up in a different geographical region, going out to the internet through a different IP address, running different OS versions, with different software and language packages installed, etc. As a result, we now have very interesting sightings in terms of infrastructure:

These communication points can be very easily triaged. Remember that VirusTotal also characterizes domains, IP addresses and URLs. Threat reputation for these domains further confirms that they are accurate IoCs:

The domain relationships (in-the-wild sightings) tell the same story:

We now have additional IoCs that we can feed into our stack in order to proactively defend our organization from other variants. As a bonus point, pivoting to other campaign files that have sandbox behaviour reports allows us to shed more light into other TTPs that we might be tracking via MITRE ATT&CK (e.g. installation, actions on objectives, etc.).

Gaining context through the community

Furthering on the use of the “have” search modifier, we can also leverage it to find files on which some VT Community user has placed a comment providing more context:

main_icon_dhash:23232b2b00010000 AND have:comments

Community comments often give us interesting details in terms of in-the-wild observations, malware capabilities, reverse engineering reports, attribution, etc. For example, in this particular case we learn about additional distribution URLs:

This other case helps us understand that this first stage is EMOTET and allows us to jump into a pastebin dump with further context about the campaign in terms of related hashes and network infrastructure:

Additional context

The “have” modifier accepts many other values, some of the more representative ones are:

  • compressed_parents: the files were seen inside a compressed file uploaded to VirusTotal.
  • pcap_parents: the files were seen in a network traffic recording uploaded to VirusTotal.
  • embedded_(urls/domains/ips): a URL/domain/IP address pattern was extracted from the binary bodies of the files.
  • behaviour: the files managed to execute in at least one sandbox and produced the pertinent dynamic analysis report.
  • behaviour_registry: the files executed in a sandbox and interacted with the Windows Registry.
  • crowdsource_yara_rule: the files match some YARA rule coming from open source community repositories, these rules often provide additional references and descriptions about a threat.

Summing up

VirusTotal aggregates orthogonal means to cluster together groups of related files. Files which may belong to the same malware family/framework/campaign/actor. These file similarity vectors range from structural features to dynamic analysis observations.

We started off with a single IoC for which we had little context, neither did VirusTotal, beyond basic threat reputation. By leveraging file similarity we managed to find thousands of other files related to the campaign/malware framework. Through the “have” search modifier we then narrowed down our searches to identify phishing emails used by the attackers, distribution URLs, additional network infrastructure such as CnCs and context shared by other threat researchers.

All of this is tactical intelligence that can be fed into network perimeter defenses, but also context that can be operationalized and digested into TTPs in order to characterize threat actors. Finally, this blog post presented an incident response scenario but the very same logic can be applied to threat actor tracking or campaign monitoring use cases.

This post was authored by Emiliano Martinez.

Thursday, November 19, 2020

Why is similarity so relevant when investigating attacks

The concept of similarity is pretty straightforward: are two files similar? There are many ways to figure it out. That's why different similarity algorithms exist. Now, why is this useful? 

Attackers need tools for their attacks, basically malware. Malware in the end is a piece of software, built from frameworks, code and libraries, and takes some time and expertise to create. The result is that two different malware files built from the same developer using the same pieces will look alike.

Imagine you are investigating some attack and you find some suspicious file. After taking a look in VirusTotal, you find nothing really meaningful about the file itself. One idea at this point would be finding similar files: maybe the attacker used similar malware in other campaigns than the one under investigation, and maybe these files will tell more about the infection chain and infrastructure. Here is where similarity comes handy!

Additionally, the same approach can be applied to attribution. We find some malware that looks new, there are no references about it. Can we find similar malware? Maybe the new artefacts will tell more about the author, maybe they are well-known by the security industry. This is how attribution is built in many cases.

There are many situations where similarity becomes useful. We can always reduce the problem to the following: IOCs can easily be replaced, malware frameworks not. 

If you want to know more about how to use similarity in real cases, join us next November 25th for our “Similarity brings your threat hunting to the next level” webinar with TrendMicro and Trinity Cyber. Register here.

In this blogpost we will discuss some interesting ideas of what can be done with similarity in VirusTotal.

File similarity in VT

You came across the following sample c9b96d5d694e4e25e03d97c7b95eff637525e539b9c47c8eda498f72ecd51b22 within your network and you want to find some context. Crowdsourced sigma rules already warn that something fishy might be going on. 

At this point we want to get a better understanding of the whole picture, which means getting more artifacts. When we run out of indicators, similarity to the rescue!

How to find similar samples? Right from the Details panel in the sample report there are several hashes that correspond to the output of different similarity algorithms: vhash, authentihash, imphash, rich PE header hash, ssdeep and TLSH:

It is important to understand that different similarity algorithms provide different results. Choosing the right similarity many times depends on the samples we are working with, that's why sometimes it is just easier to check them all at the same time and take a look at the results.

Clicking on any of the hashes shown in the report will return all similar samples. In this case, vhash returns 57 additional files, imphash finds no other hits and rich PE header hash returns around 1.16 million other files in VT (we can spot potential non-malicious files adding the search operator positives:0).

All the above might sound too technical, that's why sometimes we can approach this similarity problem with a different angle. For instance, we implement visual similarity. This is specially useful for suspicious documents distributed by attackers, but it also works for executables sharing similar icons. In this case, visual similarity returns 3,390 new files by clicking on the icon above.

We do our best to detonate in a sandbox every file we receive in VirusTotal. Would it be possible to find files with a similar behavior? It is! Even better, we integrate multiple sandboxes, offering us different options. We can do this similarity search either by selecting it in the multiple similarity button, or in the Behavior tab. Following the example, JujuBox behaviour similarity returns 11 additional files. This is an interesting feature when we want to make TTPs actionable, but we will get back to all these topics in a future post.

We have used clustering hashes (both static/structural and behavioural), but are there concrete features that we could pivot on? We can look at the Capabilities and Indicators. Specifically, let's try to find some pivotable features (or clues) among the million files caught by rich PE header hash using the VT Enterprise query [rich_pe_header_hash:640b9fb49577f39427b39125155c2425 have:clue_rule]. One of the results, 15e5353c8d5d1b1dba8d9c99e77075d737771335eac9597eba95d1f3efc3b6cd, shows interesting dropped files, registry keys set and DNS resolutions in the Details panel. We can click in any of these indicators to find their respective clusters.

We can even drop all of this into a VT-Graph to see the whole picture and the different clusters in a single panel, including the rest of the attack like dropped files, contacted URLs, etc.


To sum up, once we understand the value of using similarity for our threat hunting, it is very important to have all the options available depending on our needs. Different investigations, or different malware families, need different approaches. Behavioural similarity for instance can be very interesting when the samples are different but the TTPs are common.

But we cannot apply similarity without any data to compare. In VirusTotal we have 2.5 Billion files to make sure you get the most from your Threat Intel investigations.

Happy hunting!

This post was co-authored by Marta Gomez and Jose Martin.

Thursday, November 05, 2020

, , , ,

Keep your friends close; keep ransomware closer

“How to avoid being a ransomware victim?” is one of the main questions every single company and organization asks themselves every day. Unfortunately there is no silver bullet against that, but there are several good practices we can follow to minimize our exposure.

We can start by enumerating what are the main vectors that attackers use to get into victims: phishing, brute forcing and the use of exploits. Let's use this information to understand what exactly are attackers doing from a technical point of view, but more importantly, to monitor how their campaigns evolve. And here we want to highlight the importance of continuously tracking malicious activity in order to feed our systems accordingly: attackers evolve their methods and the IOCs used constantly change. We need the whole movie, not just a static picture.

This post describes different examples of techniques we can use to monitor ransomware campaigns, with a special focus on the infection vectors previously mentioned in order to minimize the risk of becoming a victim.

For more details, you can check our recorded anti-ransomware webinar in English and Spanish.

Ransomware in phishing attacks

Phishing is the most common technique used to distribute ransomware. We want to be able to discover how it is being used in new ransomware campaigns and to obtain the infrastructure behind the attack, gathering valuable IOCs and TTPs to feed our defenses.

We can start looking for emails involved in phishing campaigns uploaded this year to VirusTotal:

engines:ransom type:email fs:2020-01-01+

We get a list of generic ransomware email files. We can specify a certain malware family we are interested in. For instance, the following query returns emails related to some of the most common campaigns:

(engines:bitpaymer OR engines:maze OR engines:Ryuk OR engines:gandcrab OR engines:clop OR engines:revil OR engines:sodibiniki OR engines:matrix) type:email

Trickbot is a malware family frequently used to distribute ransomware. By searching for recent samples delivered by email (engines:trickbot fs:2020-09-01+ type:email) we can quickly find an interesting sample implementing an exploit and pretending to be a well known financial institution. We can quickly expand all the domains, URLs and IP addresses embedded into this file into our investigation graph, getting a broader overview of the campaign:

Expanding different nodes uncovers new IOCs to feed our defenses and unfolds this campaign, showing domain names that were used to bait victims into opening the malicious word document attached to the phishing email. 

This kind of phishing attacks where legitimate logos, domains and brand images are used to bait victims into executing malware can hurt a company's reputation, not to speak of being used against the company itself. The sooner we detect a campaign the faster we can perform actions to shut it down. VirusTotal’s Livehunt checks any submitted file against a search criteria written in Yara.

For example, to check for embedded domains in emails detected as phishing, we could use:

import "vt"
rule brandmon_google {
        $domain1 = ""
        $domain2 = ""
        $domain6 = ""
        for any engine, signature in vt.metadata.signatures : (
            signature contains "phishing" and vt.metadata.file_type == vt.FileType.EMAIL and (any of them)

Exploits used in ransomware attacks

Exploits are commonly used for installing malware or for escalating privileges into your system.

According to this report, the four CVEs that are most frequently used for performing ransomware attacks this year are:
  • CVE-2019-19781 → Revil/Sodinokibi, Ragnarok, DopplePaymer, Maze, CLOP y Nephilim.
  • CVE-2019-11510 → Revil/Sodinokibi y Black Kingdom
  • CVE-2012-0158 → EDA2 y RASOM
  • CVE-2018-8453 → Revil/Sodinokibi
We can add to the list a couple of recent remarkable exploits: zerologon (CVE-2020-1472) and SMBGhost (CVE-2020-0796). We observed several ransomware lookups in VirusTotal tagged with this last vulnerability during the last months:

We could use the following query to get more detailed information about what CVEs were used in ransomware attacks during 2020:

engines:ransom tag:exploit fs:2020-01-01+ tag:CVE-2020*

We can once again filter by malware families. For instance, the previous query is mostly GandCrab malware, which can be easily checked using the query: (engines:ransom and not engines:gandcrab) tag:exploit fs:2020-01-01+ tag:CVE-2020*).

Now, we are ready to create a Livehunt rule to find new files tagged with one of the exploits frequently used by ransomware. 

import "vt"
rule ransomware_exploits {
        for any tag in vt.metadata.tags : (
            tag == "cve-2019-19781" or
            tag == "cve-2019-11510" or
            tag == "cve-2012-0158" or
            tag == "cve-2018-8453" or
            tag == "cve-2020-1472" or
            tag == "cve-2020-0796"
        ) and not vt.metadata.file_type == vt.FileType.CAP

This will result in an immediate notification, allowing us tracking any new IOCs we can use to protect our system.

More importantly, this is very valuable information we can use on a regular basis to manage our patching policy, prioritizing patches based on fresh data of how different exploits are being used in real attacks.

Tracking fresh campaigns

Now, we want to make sure that we monitor any new ransomware campaign in order to understand how it evolves and what new artefacts and techniques they use.

As an example, we can start with a recent DFIR Report investigation revealing Ryuk exploiting zerologon. There are many ways to track campaigns, however VT Graph is a great choice to get together all the discovered observables and extend our knowledge in a visual way. Here are some tips that could help you during this process:
  • Start by adding all the known observables to a new VT Graph.
  • Expand domains, URLs and IPs to unfold relations and obtain new observables.
  • In order to keep our list of observables up to date, we can translate common Yara rules into Livehunt rules to catch new files, injecting Livehunts results directly into the graph.
  • Additionally, we can use Retrohunt rules to look for similar samples in our collection.
We start the investigation dropping one of the files included in the publication in a new graph, showing domains, urls and ip addresses embedded in the file, ITW URLs hosting the file and network observables contacted by this sample when executed. This file is detected as “bazar” malware, used to install Ryuk. We dropped all this information in our graph:

Additionally to keep pivoting using our graph and indicators, we can also translate the Yara rules from the DFIR report into Livehunt rules.

We can integrate Livehunt results into our graph in just two clicks. Just click on the target icon at the right in the VTGrap interface, select the rule desired and choose "Load results". This will add all the new observables that match our rules to the current graph. We can expand these new nodes to unveil new observables and create relationships.

All these new IOCs are fresh observables that are clearly related to this campaign. All this continuous flow of fresh indicators will help us improve our security mechanisms to stop Ryuk from passing through our defenses.

Summarizing, the knowledge of what attackers are using is the first necessary step for us to minimize our exposure to different campaigns. It wouldn't be right to put all the different ransomware attacks under the same umbrella, as they became highly specialized and protecting from different actors is not exactly the same. The techniques described in this post are a good starting point for automatically minimizing our exposure to more spread ransomware campaigns, however they can be applied both for generic and targeted attacks.

Stay safe and happy hunting!

This post was co-authored by Vicente Diaz.