Blog
/
AI
/
October 30, 2023

Exploring AI Threats: Package Hallucination Attacks

Learn how malicious actors exploit errors in generative AI tools to launch packet attacks. Read how Darktrace products detect and prevent these threats!
Inside the SOC
Darktrace cyber analysts are world-class experts in threat intelligence, threat hunting and incident response, and provide 24/7 SOC support to thousands of Darktrace customers around the globe. Inside the SOC is exclusively authored by these experts, providing analysis of cyber incidents and threat trends, based on real-world experience in the field.
Written by
Charlotte Thompson
Cyber Analyst
Written by
Tiana Kelly
Deputy Team Lead, London & Cyber Analyst
Default blog imageDefault blog imageDefault blog imageDefault blog imageDefault blog imageDefault blog image
30
Oct 2023

AI tools open doors for threat actors

On November 30, 2022, the free conversational language generation model ChatGPT was launched by OpenAI, an artificial intelligence (AI) research and development company. The launch of ChatGPT was the culmination of development ongoing since 2018 and represented the latest innovation in the ongoing generative AI boom and made the use of generative AI tools accessible to the general population for the first time.

ChatGPT is estimated to currently have at least 100 million users, and in August 2023 the site reached 1.43 billion visits [1]. Darktrace data indicated that, as of March 2023, 74% of active customer environments have employees using generative AI tools in the workplace [2].

However, with new tools come new opportunities for threat actors to exploit and use them maliciously, expanding their arsenal.

Much consideration has been given to mitigating the impacts of the increased linguistic complexity in social engineering and phishing attacks resulting from generative AI tool use, with Darktrace observing a 135% increase in ‘novel social engineering attacks’ across thousands of active Darktrace/Email™ customers from January to February 2023, corresponding with the widespread adoption of ChatGPT and its peers [3].

Less overall consideration, however, has been given to impacts stemming from errors intrinsic to generative AI tools. One of these errors is AI hallucinations.

What is an AI hallucination?

AI “hallucination” is a term which refers to the predictive elements of generative AI and LLMs’ AI model gives an unexpected or factually incorrect response which does not align with its machine learning training data [4]. This differs from regular and intended behavior for an AI model, which should provide a response based on the data it was trained upon.  

Why are AI hallucinations a problem?

Despite the term indicating it might be a rare phenomenon, hallucinations are far more likely than accurate or factual results as the AI models used in LLMs are merely predictive and focus on the most probable text or outcome, rather than factual accuracy.

Given the widespread use of generative AI tools in the workplace employees are becoming significantly more likely to encounter an AI hallucination. Furthermore, if these fabricated hallucination responses are taken at face value, they could cause significant issues for an organization.

Use of generative AI in software development

Software developers may use generative AI for recommendations on how to optimize their scripts or code, or to find packages to import into their code for various uses. Software developers may ask LLMs for recommendations on specific pieces of code or how to solve a specific problem, which will likely lead to a third-party package. It is possible that packages recommended by generative AI tools could represent AI hallucinations and the packages may not have been published, or, more accurately, the packages may not have been published prior to the date at which the training data for the model halts. If these hallucinations result in common suggestions of a non-existent package, and the developer copies the code snippet wholesale, this may leave the exchanges vulnerable to attack.

Research conducted by Vulcan revealed the prevalence of AI hallucinations when ChatGPT is asked questions related to coding. After sourcing a sample of commonly asked coding questions from Stack Overflow, a question-and-answer website for programmers, researchers queried ChatGPT (in the context of Node.js and Python) and reviewed its responses. In 20% of the responses provided by ChatGPT pertaining to Node.js at least one un-published package was included, whilst the figure sat at around 35% for Python [4].

Hallucinations can be unpredictable, but would-be attackers are able to find packages to create by asking generative AI tools generic questions and checking whether the suggested packages exist already. As such, attacks using this vector are unlikely to target specific organizations, instead posing more of a widespread threat to users of generative AI tools.

Malicious packages as attack vectors

Although AI hallucinations can be unpredictable, and responses given by generative AI tools may not always be consistent, malicious actors are able to discover AI hallucinations by adopting the approach used by Vulcan. This allows hallucinated packages to be used as attack vectors. Once a malicious actor has discovered a hallucination of an un-published package, they are able to create a package with the same name and include a malicious payload, before publishing it. This is known as a malicious package.

Malicious packages could also be recommended by generative AI tools in the form of pre-existing packages. A user may be recommended a package that had previously been confirmed to contain malicious content, or a package that is no longer maintained and, therefore, is more vulnerable to hijack by malicious actors.

In such scenarios it is not necessary to manipulate the training data (data poisoning) to achieve the desired outcome for the malicious actor, thus a complex and time-consuming attack phase can easily be bypassed.

An unsuspecting software developer may incorporate a malicious package into their code, rendering it harmful. Deployment of this code could then result in compromise and escalation into a full-blown cyber-attack.

Figure 1: Flow diagram depicting the initial stages of an AI Package Hallucination Attack.

For providers of Software-as-a-Service (SaaS) products, this attack vector may represent an even greater risk. Such organizations may have a higher proportion of employed software developers than other organizations of comparable size. A threat actor, therefore, could utilize this attack vector as part of a supply chain attack, whereby a malicious payload becomes incorporated into trusted software and is then distributed to multiple customers. This type of attack could have severe consequences including data loss, the downtime of critical systems, and reputational damage.

How could Darktrace detect an AI Package Hallucination Attack?

In June 2023, Darktrace introduced a range of DETECT™ and RESPOND™ models designed to identify the use of generative AI tools within customer environments, and to autonomously perform inhibitive actions in response to such detections. These models will trigger based on connections to endpoints associated with generative AI tools, as such, Darktrace’s detection of an AI Package Hallucination Attack would likely begin with the breaching of one of the following DETECT models:

  • Compliance / Anomalous Upload to Generative AI
  • Compliance / Beaconing to Rare Generative AI and Generative AI
  • Compliance / Generative AI

Should generative AI tool use not be permitted by an organization, the Darktrace RESPOND model ‘Antigena / Network / Compliance / Antigena Generative AI Block’ can be activated to autonomously block connections to endpoints associated with generative AI, thus preventing an AI Package Hallucination attack before it can take hold.

Once a malicious package has been recommended, it may be downloaded from GitHub, a platform and cloud-based service used to store and manage code. Darktrace DETECT is able to identify when a device has performed a download from an open-source repository such as GitHub using the following models:

  • Device / Anomalous GitHub Download
  • Device / Anomalous Script Download Followed By Additional Packages

Whatever goal the malicious package has been designed to fulfil will determine the next stages of the attack. Due to their highly flexible nature, AI package hallucinations could be used as an attack vector to deliver a large variety of different malware types.

As GitHub is a commonly used service by software developers and IT professionals alike, traditional security tools may not alert customer security teams to such GitHub downloads, meaning malicious downloads may go undetected. Darktrace’s anomaly-based approach to threat detection, however, enables it to recognize subtle deviations in a device’s pre-established pattern of life which may be indicative of an emerging attack.

Subsequent anomalous activity representing the possible progression of the kill chain as part of an AI Package Hallucination Attack could then trigger an Enhanced Monitoring model. Enhanced Monitoring models are high-fidelity indicators of potential malicious activity that are investigated by the Darktrace analyst team as part of the Proactive Threat Notification (PTN) service offered by the Darktrace Security Operation Center (SOC).

Conclusion

Employees are often considered the first line of defense in cyber security; this is particularly true in the face of an AI Package Hallucination Attack.

As the use of generative AI becomes more accessible and an increasingly prevalent tool in an attacker’s toolbox, organizations will benefit from implementing company-wide policies to define expectations surrounding the use of such tools. It is simple, yet critical, for example, for employees to fact check responses provided to them by generative AI tools. All packages recommended by generative AI should also be checked by reviewing non-generated data from either external third-party or internal sources. It is also good practice to adopt caution when downloading packages with very few downloads as it could indicate the package is untrustworthy or malicious.

As of September 2023, ChatGPT Plus and Enterprise users were able to use the tool to browse the internet, expanding the data ChatGPT can access beyond the previous training data cut-off of September 2021 [5]. This feature will be expanded to all users soon [6]. ChatGPT providing up-to-date responses could prompt the evolution of this attack vector, allowing attackers to publish malicious packages which could subsequently be recommended by ChatGPT.

It is inevitable that a greater embrace of AI tools in the workplace will be seen in the coming years as the AI technology advances and existing tools become less novel and more familiar. By fighting fire with fire, using AI technology to identify AI usage, Darktrace is uniquely placed to detect and take preventative action against malicious actors capitalizing on the AI boom.

Credit to Charlotte Thompson, Cyber Analyst, Tiana Kelly, Analyst Team Lead, London, Cyber Analyst

References

[1] https://seo.ai/blog/chatgpt-user-statistics-facts

[2] https://darktrace.com/news/darktrace-addresses-generative-ai-concerns

[3] https://darktrace.com/news/darktrace-email-defends-organizations-against-evolving-cyber-threat-landscape

[4] https://vulcan.io/blog/ai-hallucinations-package-risk?nab=1&utm_referrer=https%3A%2F%2Fwww.google.com%2F

[5] https://twitter.com/OpenAI/status/1707077710047216095

[6] https://www.reuters.com/technology/openai-says-chatgpt-can-now-browse-internet-2023-09-27/

Inside the SOC
Darktrace cyber analysts are world-class experts in threat intelligence, threat hunting and incident response, and provide 24/7 SOC support to thousands of Darktrace customers around the globe. Inside the SOC is exclusively authored by these experts, providing analysis of cyber incidents and threat trends, based on real-world experience in the field.
Written by
Charlotte Thompson
Cyber Analyst
Written by
Tiana Kelly
Deputy Team Lead, London & Cyber Analyst

More in this series

No items found.

Blog

/

/

April 24, 2025

The Importance of NDR in Resilient XDR

picture of hands typing on laptop Default blog imageDefault blog image

As threat actors become more adept at targeting and disabling EDR agents, relying solely on endpoint detection leaves critical blind spots.

Network detection and response (NDR) offers the visibility and resilience needed to catch what EDR can’t especially in environments with unmanaged devices or advanced threats that evade local controls.

This blog explores how threat actors can disable or bypass EDR-based XDR solutions and demonstrates how Darktrace’s approach to NDR closes the resulting security gaps with Self-Learning AI that enables autonomous, real-time detection and response.

Threat actors see local security agents as targets

Recent research by security firms has highlighted ‘EDR killers’: tools that deliberately target EDR agents to disable or damage them. These include the known malicious tool EDRKillShifter, the open source EDRSilencer, EDRSandblast and variants of Terminator, and even the legitimate business application HRSword.

The attack surface of any endpoint agent is inevitably large, whether the software is challenged directly, by contesting its local visibility and access mechanisms, or by targeting the Operating System it relies upon. Additionally, threat actors can readily access and analyze EDR tools, and due to their uniformity across environments an exploit proven in a lab setting will likely succeed elsewhere.

Sophos have performed deep research into the EDRShiftKiller tool, which ESET have separately shown became accessible to multiple threat actor groups. Cisco Talos have reported via TheRegister observing significant success rates when an EDR kill was attempted by ransomware actors.

With the local EDR agent silently disabled or evaded, how will the threat be discovered?

What are the limitations of relying solely on EDR?

Cyber attackers will inevitably break through boundary defences, through innovation or trickery or exploiting zero-days. Preventive measures can reduce but not completely stop this. The attackers will always then want to expand beyond their initial access point to achieve persistence and discover and reach high value targets within the business. This is the primary domain of network activity monitoring and NDR, which includes responsibility for securing the many devices that cannot run endpoint agents.

In the insights from a CISA Red Team assessment of a US CNI organization, the Red Team was able to maintain access over the course of months and achieve their target outcomes. The top lesson learned in the report was:

“The assessed organization had insufficient technical controls to prevent and detect malicious activity. The organization relied too heavily on host-based endpoint detection and response (EDR) solutions and did not implement sufficient network layer protections.”

This proves that partial, isolated viewpoints are not sufficient to track and analyze what is fundamentally a connected problem – and without the added visibility and detection capabilities of NDR, any downstream SIEM or MDR services also still have nothing to work with.

Why is network detection & response (NDR) critical?

An effective NDR finds threats that disable or can’t be seen by local security agents and generally operates out-of-band, acquiring data from infrastructure such as traffic mirroring from physical or virtual switches. This means that the security system is extremely inaccessible to a threat actor at any stage.

An advanced NDR such as Darktrace / NETWORK is fully capable of detecting even high-end novel and unknown threats.

Detecting exploitation of Ivanti CS/PS with Darktrace / NETWORK

On January 9th 2025, two new vulnerabilities were disclosed in Ivanti Connect Secure and Policy Secure appliances that were under malicious exploitation. Perimeter devices, like Ivanti VPNs, are designed to keep threat actors out of a network, so it's quite serious when these devices are vulnerable.

An NDR solution is critical because it provides network-wide visibility for detecting lateral movement and threats that an EDR might miss, such as identifying command and control sessions (C2) and data exfiltration, even when hidden within encrypted traffic and which an EDR alone may not detect.

Darktrace initially detected suspicious activity connected with the exploitation of CVE-2025-0282 on December 29, 2024 – 11 days before the public disclosure of the vulnerability, this early detection highlights the benefits of an anomaly-based network detection method.

Throughout the campaign and based on the network telemetry available to Darktrace, a wide range of malicious activities were identified, including the malicious use of administrative credentials, the download of suspicious files, and network scanning in the cases investigated.

Darktrace / NETWORK’s autonomous response capabilities played a critical role in containment by autonomously blocking suspicious connections and enforcing normal behavior patterns. At the same time, Darktrace Cyber AI Analyst™ automatically investigated and correlated the anomalous activity into cohesive incidents, revealing the full scope of the compromise.

This case highlights the importance of real-time, AI-driven network monitoring to detect and disrupt stealthy post-exploitation techniques targeting unmanaged or unprotected systems.

Unlocking adaptive protection for evolving cyber risks

Darktrace / NETWORK uses unique AI engines that learn what is normal behavior for an organization’s entire network, continuously analyzing, mapping and modeling every connection to create a full picture of your devices, identities, connections, and potential attack paths.

With its ability to uncover previously unknown threats as well as detect known threats using signatures and threat intelligence, Darktrace is an essential layer of the security stack. Darktrace has helped secure customers against attacks including 2024 threat actor campaigns against Fortinet’s FortiManager , Palo Alto firewall devices, and more.  

Stay tuned for part II of this series which dives deeper into the differences between NDR types.

Credit to Nathaniel Jones VP, Security & AI Strategy, FCISO & Ashanka Iddya, Senior Director of Product Marketing for their contribution to this blog.

Continue reading
About the author
Nathaniel Jones
VP, Security & AI Strategy, Field CISO

Blog

/

/

April 22, 2025

Obfuscation Overdrive: Next-Gen Cryptojacking with Layers

man looking at multiple computer screensDefault blog imageDefault blog image

Out of all the services honeypotted by Darktrace, Docker is the most commonly attacked, with new strains of malware emerging daily. This blog will analyze a novel malware campaign with a unique obfuscation technique and a new cryptojacking technique.

What is obfuscation?

Obfuscation is a common technique employed by threat actors to prevent signature-based detection of their code, and to make analysis more difficult. This novel campaign uses an interesting technique of obfuscating its payload.

Docker image analysis

The attack begins with a request to launch a container from Docker Hub, specifically the kazutod/tene:ten image. Using Docker Hub’s layer viewer, an analyst can quickly identify what the container is designed to do. In this case, the container is designed to run the ten.py script which is built into itself.

 Docker Hub Image Layers, referencing the script ten.py.
Figure 1: Docker Hub Image Layers, referencing the script ten.py.

To gain more information on the Python file, Docker’s built in tooling can be used to download the image (docker pull kazutod/tene:ten) and then save it into a format that is easier to work with (docker image save kazutod/tene:ten -o tene.tar). It can then be extracted as a regular tar file for further investigation.

Extraction of the resulting tar file.
Figure 2: Extraction of the resulting tar file.

The Docker image uses the OCI format, which is a little different to a regular file system. Instead of having a static folder of files, the image consists of layers. Indeed, when running the file command over the sha256 directory, each layer is shown as a tar file, along with a JSON metadata file.

Output of the file command over the sha256 directory.
Figure 3: Output of the file command over the sha256 directory.

As the detailed layers are not necessary for analysis, a single command can be used to extract all of them into a single directory, recreating what the container file system would look like:

find blobs/sha256 -type f -exec sh -c 'file "{}" | grep -q "tar archive" && tar -xf "{}" -C root_dir' \;

Result of running the command above.
Figure 4: Result of running the command above.

The find command can then be used to quickly locate where the ten.py script is.

find root_dir -name ten.py

root_dir/app/ten.py

Details of the above ten.py script.
Figure 5: Details of the above ten.py script.

This may look complicated at first glance, however after breaking it down, it is fairly simple. The script defines a lambda function (effectively a variable that contains executable code) and runs zlib decompress on the output of base64 decode, which is run on the reversed input. The script then runs the lambda function with an input of the base64 string, and then passes it to exec, which runs the decoded string as Python code.

To help illustrate this, the code can be cleaned up to this simplified function:

def decode(input):
   reversed = input[::-1]

   decoded = base64.decode(reversed)
   decompressed = zlib.decompress(decoded)
   return decompressed

decoded_string = decode(the_big_text_blob)
exec(decoded_string) # run the decoded string

This can then be set up as a recipe in Cyberchef, an online tool for data manipulation, to decode it.

Use of Cyberchef to decode the ten.py script.
Figure 6: Use of Cyberchef to decode the ten.py script.

The decoded payload calls the decode function again and puts the output into exec. Copy and pasting the new payload into the input shows that it does this another time. Instead of copy-pasting the output into the input all day, a quick script can be used to decode this.

The script below uses the decode function from earlier in order to decode the base64 data and then uses some simple string manipulation to get to the next payload. The script will run this over and over until something interesting happens.

# Decode the initial base64

decoded = decode(initial)
# Remove the first 11 characters and last 3

# so we just have the next base64 string

clamped = decoded[11:-3]

for i in range(1, 100):
   # Decode the new payload

   decoded = decode(clamped)
   # Print it with the current step so we

   # can see what’s going on

   print(f"Step {i}")

   print(decoded)
   # Fetch the next base64 string from the

   # output, so the next loop iteration will

   # decode it

   clamped = decoded[11:-3]

Result of the 63rd iteration of this script.
Figure 7: Result of the 63rd iteration of this script.

After 63 iterations, the script returns actual code, accompanied by an error from the decode function as a stopping condition was never defined. It not clear what the attacker’s motive to perform so many layers of obfuscation was, as one round of obfuscation versus several likely would not make any meaningful difference to bypassing signature analysis. It’s possible this is an attempt to stop analysts or other hackers from reverse engineering the code. However,  it took a matter of minutes to thwart their efforts.

Cryptojacking 2.0?

Cleaned up version of the de-obfuscated code.
Figure 8: Cleaned up version of the de-obfuscated code.

The cleaned up code indicates that the malware attempts to set up a connection to teneo[.]pro, which appears to belong to a Web3 startup company.

Teneo appears to be a legitimate company, with Crunchbase reporting that they have raised USD 3 million as part of their seed round [1]. Their service allows users to join a decentralized network, to “make sure their data benefits you” [2]. Practically, their node functions as a distributed social media scraper. In exchange for doing so, users are rewarded with “Teneo Points”, which are a private crypto token.

The malware script simply connects to the websocket and sends keep-alive pings in order to gain more points from Teneo and does not do any actual scraping. Based on the website, most of the rewards are gated behind the number of heartbeats performed, which is likely why this works [2].

Checking out the attacker’s dockerhub profile, this sort of attack seems to be their modus operandi. The most recent container runs an instance of the nexus network client, which is a project to perform distributed zero-knowledge compute tasks in exchange for cryptocurrency.

Typically, traditional cryptojacking attacks rely on using XMRig to directly mine cryptocurrency, however as XMRig is highly detected, attackers are shifting to alternative methods of generating crypto. Whether this is more profitable remains to be seen. There is not currently an easy way to determine the earnings of the attackers due to the more “closed” nature of the private tokens. Translating a user ID to a wallet address does not appear to be possible, and there is limited public information about the tokens themselves. For example, the Teneo token is listed as “preview only” on CoinGecko, with no price information available.

Conclusion

This blog explores an example of Python obfuscation and how to unravel it. Obfuscation remains a ubiquitous technique employed by the majority of malware to aid in detection/defense evasion and being able to de-obfuscate code is an important skill for analysts to possess.

We have also seen this new avenue of cryptominers being deployed, demonstrating that attackers’ techniques are still evolving - even tried and tested fields. The illegitimate use of legitimate tools to obtain rewards is an increasingly common vector. For example,  as has been previously documented, 9hits has been used maliciously to earn rewards for the attack in a similar fashion.

Docker remains a highly targeted service, and system administrators need to take steps to ensure it is secure. In general, Docker should never be exposed to the wider internet unless absolutely necessary, and if it is necessary both authentication and firewalling should be employed to ensure only authorized users are able to access the service. Attacks happen every minute, and even leaving the service open for a short period of time may result in a serious compromise.

References

1. https://www.crunchbase.com/funding_round/teneo-protocol-seed--a8ff2ad4

2. https://teneo.pro/

Continue reading
About the author
Nate Bill
Threat Researcher
Your data. Our AI.
Elevate your network security with Darktrace AI