Blog
/
/
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

/

Network

/

November 13, 2025

Unmasking Vo1d: Inside Darktrace’s Botnet Detection

Default blog imageDefault blog image

What is vo1d APK malware?

Vo1d malware first appeared in the wild in September 2024 and has since evolved into one of the most widespread Android botnets ever observed. This large-scale Android malware primarily targets smart TVs and low-cost Android TV boxes. Initially, Vo1d was identified as a malicious backdoor capable of installing additional third-party software [1]. Its functionality soon expanded beyond the initial infection to include deploying further malicious payloads, running proxy services, and conducting ad fraud operations. By early 2025, it was estimated that Vo1d had infected 1.3 to 1.6 million devices worldwide [2].

From a technical perspective, Vo1d embeds components into system storage to enable itself to download and execute new modules at any time. External researchers further discovered that Vo1d uses Domain Generation Algorithms (DGAs) to create new command-and-control (C2) domains, ensuring that regardless of existing servers being taken down, the malware can quickly reconnect to new ones. Previous published analysis identified dozens of C2 domains and hundreds of DGA seeds, along with new downloader families. Over time, Vo1d has grown increasingly sophisticated with clear signs of stronger obfuscation and encryption methods designed to evade detection [2].

Darktrace’s coverage

Earlier this year, Darktrace observed a surge in Vo1d-related activity across customer environments, with the majority of affected customers based in South Africa. Devices that had been quietly operating as expected began exhibiting unusual network behavior, including excessive DNS lookups. Open-source intelligence (OSINT) has long highlighted South Africa as one of the countries most impacted by Vo1d infections [2].

What makes the recent activity particularly interesting is that the surge observed by Darktrace appears to be concentrated specifically in South African environments. This localized spike suggests that a significant number of devices may have been compromised, potentially due to vulnerable software, outdated firmware, or even preloaded malware. Regions with high prevalence of low-cost, often unpatched devices are especially susceptible, as these everyday consumer electronics can be quietly recruited into the botnet’s network. This specifically appears to be the case with South Africa, where public reporting has documented widespread use of low-cost boxes, such as non-Google-certified Android TV sticks, that frequently ship with outdated firmware [3].

The initial triage highlighted the core mechanism Vo1d uses to remain resilient: its use of DGA. A DGA deterministically creates a large list of pseudo-random domain names on a predictable schedule. This enables the malware to compute hundreds of candidate domains using the same algorithm, instead of using a hard-coded single C2 hostname that defenders could easily block or take down. To ensure reproducible from the infected device’s perspective, Vo1d utilizes DGA seeds. These seeds might be a static string, a numeric value, or a combination of underlying techniques that enable infected devices to generate the same list of candidate domains for a time window, provided the same DGA code, seed, and date are used.

Interestingly, Vo1d’s DGA seeds do not appear to be entirely unpredictable, and the generated domains lack fully random-looking endings. As observed in Figure 1, there is a clear pattern in the names generated. In this case, researchers identified that while the first five characters would change to create the desired list of domain names, the trailing portion remained consistent as part of the seed: 60b33d7929a, which OSINT sources have linked to the Vo1d botnet. [2]. Darktrace’s Threat Research team also identified a potential second DGA seed, with devices in some cases also engaging in activity involving hostnames matching the regular expression /[a-z]{5}fc975904fc9\.(com|top|net). This second seed has not been reported by any OSINT vendors at the time of writing.

Another recurring characteristic observed across multiple cases was the choice of top-level domains (TLDs), which included .com, .net, and .top.

Figure 1: Advanced Search results showing DNS lookups, providing a glimpse on the DGA seed utilized.

The activity was detected by multiple models in Darktrace / NETWORK, which triggered on devices making an unusually large volume of DNS requests for domains uncommon across the network.

During the network investigation, Darktrace analysts traced Vo1d’s infrastructure and uncovered an interesting pattern related to responder ASNs. A significant number of connections pointed to AS16509 (AMAZON-02). By hosting redirectors or C2 nodes inside major cloud environments, Vo1d is able to gain access to highly available and geographically diverse infrastructure. When one node is taken down or reported, operators can quickly enable a new node under a different IP within the same ASN. Another feature of cloud infrastructure that hardens Vo1d’s resilience is the fact that many organizations allow outbound connections to cloud IP ranges by default, assuming they are legitimate. Despite this, Darktrace was able to identify the rarity of these endpoints, identifying the unusualness of the activity.

Analysts further observed that once a generated domain successfully resolved, infected devices consistently began establishing outbound connections to ephemeral port ranges like TCP ports 55520 and 55521. These destination ports are atypical for standard web or DNS traffic. Even though the choice of high-numbered ports appears random, it is likely far from not accidental. Commonly used ports such as port 80 (HTTP) or 443 (HTTPS) are often subject to more scrutiny and deeper inspection or content filtering, making them riskier for attackers. On the other hand, unregistered ports like 55520 and 55521 are less likely to be blocked, providing a more covert channel that blends with outbound TCP traffic. This tactic helps evade firewall rules that focus on common service ports. Regardless, Darktrace was able to identify external connections on uncommon ports to locations that the network does not normally visit.

The continuation of the described activity was identified by Darktrace’s Cyber AI Analyst, which correlated individual events into a broader interconnected incident. It began with the multiple DNS requests for the algorithmically generated domains, followed by repeated connections to rare endpoints later confirmed as attacker-controlled infrastructure. Cyber AI Analyst’s investigation further enabled it to categorize the events as part of the “established foothold” phase of the attack.

Figure 2: Cyber AI Analyst incident illustrating the transition from DNS requests for DGA domains to connections with resolved attacker-controlled infrastructure.

Conclusion

The observations highlighted in this blog highlight the precision and scale of Vo1d’s operations, ranging from its DGA-generated domains to its covert use of high-numbered ports. The surge in affected South African environments illustrate how regions with many low-cost, often unpatched devices can become major hubs for botnet activity. This serves as a reminder that even everyday consumer electronics can play a role in cybercrime, emphasizing the need for vigilance and proactive security measures.

Credit to Christina Kreza (Cyber Analyst & Team Lead) and Eugene Chua (Principal Cyber Analyst & Team Lead)

Edited by Ryan Traill (Analyst Content Lead)

Appendices

Darktrace Model Detections

  • Anomalous Connection / Devices Beaconing to New Rare IP
  • Anomalous Connection / Multiple Connections to New External TCP Port
  • Anomalous Connection / Multiple Failed Connections to Rare Endpoint
  • Compromise / DGA Beacon
  • Compromise / Domain Fluxing
  • Compromise / Fast Beaconing to DGA
  • Unusual Activity / Unusual External Activity

List of Indicators of Compromise (IoCs)

  • 3.132.75[.]97 – IP address – Likely Vo1d C2 infrastructure
  • g[.]sxim[.]me – Hostname – Likely Vo1d C2 infrastructure
  • snakeers[.]com – Hostname – Likely Vo1d C2 infrastructure

Selected DGA IoCs

  • semhz60b33d7929a[.]com – Hostname – Possible Vo1d C2 DGA endpoint
  • ggqrb60b33d7929a[.]com – Hostname – Possible Vo1d C2 DGA endpoint
  • eusji60b33d7929a[.]com – Hostname – Possible Vo1d C2 DGA endpoint
  • uacfc60b33d7929a[.]com – Hostname – Possible Vo1d C2 DGA endpoint
  • qilqxfc975904fc9[.]top – Hostname – Possible Vo1d C2 DGA endpoint

MITRE ATT&CK Mapping

  • T1071.004 – Command and Control – DNS
  • T1568.002 – Command and Control – Domain Generation Algorithms
  • T1568.001 – Command and Control – Fast Flux DNS
  • T1571 – Command and Control – Non-Standard Port

[1] https://news.drweb.com/show/?lng=en&i=14900

[2] https://blog.xlab.qianxin.com/long-live-the-vo1d_botnet/

[3] https://mybroadband.co.za/news/broadcasting/596007-warning-for-south-africans-using-specific-types-of-tv-sticks.html

Continue reading
About the author
Christina Kreza
Cyber Analyst

Blog

/

Network

/

November 6, 2025

Darktrace Named the Only 2025 Gartner® Peer Insights™ Customers’ Choice for Network Detection and Response

Default blog imageDefault blog image

Darktrace: The only Customers’ Choice for NDR in 2025

In a year defined by rapid change across the threat landscape, recognition from those who use and rely on security technology every day means the most.

That’s why we’re proud to share that Darktrace has been named the only Customers’ Choice in the 2025 Gartner® Peer Insights™ Voice of the Customer for Network Detection and Response (NDR).

Out of 11 leading NDR vendors evaluated, Darktrace stood alone as the sole Customers’ Choice, a recognition that we feel reflects not just our innovation, but the trust and satisfaction of the customers who secure their networks with Darktrace every day.

What the Gartner® Peer Insights™ Voice of the Customer means

“Voice of the Customer” is a document that synthesizes Gartner Peer Insights reviews into insights for buyers of technology and services. This aggregated peer perspective, along with the individual detailed reviews, is complementary to Gartner expert research and can play a key role in your buying process. Peers are verified reviewers of a technology product or service, who not only rate the offering, but also provide valuable feedback to consider before making a purchase decision. Vendors placed in the upper-right “Customers’ Choice” quadrant of the “Voice of the Customer” have scores that meet or exceed the market average for both axes (User Interest and Adoption, and Overall Experience).It’s not just a rating. We feel it’s a reflection of genuine customer sentiment and success in the field.

In our view, Customers consistently highlight Darktrace’s ability to:

  • Detect and respond to unknown threats in real time
  • Deliver unmatched visibility across IT, OT, and cloud environments
  • Automate investigations and responses through AI-driven insights

We believe this recognition reinforces what our customers already know: that Darktrace helps them see, understand, and stop attacks others miss.

A rare double: recognized by customers and analysts alike

This distinction follows another major recogniton. Darktrace’s placement as a Leader in the Gartner® Magic Quadrant™ for Network Detection and Response earlier this year.

That makes Darktrace the only vendor to achieve both:

  • A Leader status in the Gartner Magic Quadrant for NDR, and
  • A Customers’ Choice in Gartner Peer Insights 2025

It’s a rare double that we feel reflects both industry leadership and customer trust, two perspectives that, together, define what great cybersecurity looks like.

A Customers’ Choice across the network and the inbox

To us, this recognition also builds on Darktrace’s momentum across multiple domains. Earlier this year, Darktrace was also named a Customers’ Choice for Email Security Platforms in the Gartner® Peer Insights™ report.

With more than 1,000 verified reviews across Network Detection and Response, Email Security Platforms, and Cyber Physical Systems (CPS), we at Darktrace are proud to be trusted across the full attack surface, from the inbox to the industrial network.

Thank you to our customers

We’re deeply grateful to every customer who shared their experience with Darktrace on Gartner Peer Insights. Your insights drive our innovation and continue to shape how we protect complex, dynamic environments across the world.

Discover why customers choose Darktrace for network and email security.

Gartner® Peer Insights™ content consists of the opinions of individual end users based on their own experiences, and should not be construed as statements of fact, nor do they represent the views of Gartner or its affiliates. Gartner does not endorse any vendor, product or service depicted in this content nor makes any warranties, expressed or implied, with respect to this content, about its accuracy or completeness, including any warranties of merchantability or fitness for a particular purpose.

GARTNER is a registered trademark and service mark of Gartner, Inc. and/or its affiliates in the U.S. and internationally and is used herein with permission. All rights reserved.

Magic Quadrant and Peer Insights are registered trademarks of Gartner, Inc. and/or its affiliates and is used herein with permission. All rights reserved.

Gartner, Voice of the Customer for Network Detection and Response, By Peer Community Contributor, 30 October 2025

Continue reading
About the author
Mikey Anderson
Product Marketing Manager, Network Detection & Response
Your data. Our AI.
Elevate your network security with Darktrace AI