Blog
/
Network
/
May 25, 2022

Uncovering the Sysrv-Hello Crypto-Jacking Bonet

Discover the cyber kill chain of a Sysrv-hello botnet infection in France and gain insights into the latest TTPs of the botnet in March and April 2022.
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
Shuh Chin Goh
Default blog imageDefault blog imageDefault blog imageDefault blog imageDefault blog imageDefault blog image
25
May 2022

In recent years, the prevalence of crypto-jacking botnets has risen in tandem with the popularity and value of cryptocurrencies. Increasingly crypto-mining malware programs are distributed by botnets as they allow threat actors to harness the cumulative processing power of a large number of machines (discussed in our other Darktrace blogs.1 2 One of these botnets is Sysrv-hello, which in addition to crypto-mining, propagates aggressively across the Internet in a worm-like manner by trolling for Remote Code Execution (RCE) vulnerabilities and SSH worming from the compromised victim devices. This all has the purpose of expanding the botnet.

First identified in December 2020, Sysrv-hello’s operators constantly update and change the bots’ behavior to evolve and stay ahead of security researchers and law enforcement. As such, infected systems can easily go unnoticed by both users and organizations. This blog examines the cyber kill chain sequence of a Sysrv-hello botnet infection detected at the network level by Darktrace DETECT/Network, as well as the botnet’s tactics, techniques, and procedures (TTPs) in March and April 2022.

Figure 1: Timeline of the attack

Delivery and exploitation

The organization, which was trialing Darktrace, had deployed the technology on March 2, 2022. On the very same day, the initial host infection was seen through the download of a first-stage PowerShell loader script from a rare external endpoint by a device in the internal network. Although initial exploitation of the device happened prior to the installation and was not observed, this botnet is known to target RCE vulnerabilities in various applications such as MySQL, Tomcat, PHPUnit, Apache Solar, Confluence, Laravel, JBoss, Jira, Sonatype, Oracle WebLogic and Apache Struts to gain initial access to internal systems.3 Recent iterations have also been reported to have been deployed via drive-by-downloads from an empty HTML iframe pointing to a malicious executable that downloads to the device from a user visiting a compromised website.4

Initial intrusion

The Sysrv-hello botnet is distributed for both Linux and Windows environments, with the corresponding compatible script pulled based on the architecture of the system. In this incident, the Windows version was observed.

On March 2, 2022 at 15:15:28 UTC, the device made a successful HTTP GET request to a malicious IP address5 that had a rarity score of 100% in the network. It subsequently downloaded a malicious PowerShell script named ‘ldr.ps1'6 onto the system. The associated IP address ‘194.145.227[.]21’ belongs to ‘ASN AS48693 Rices Privately owned enterprise’ and had been identified as a Sysrv-hello botnet command and control (C2) server in April the previous year. 3

Looking at the URI ‘/ldr.ps1?b0f895_admin:admin_81.255.222.82:8443_https’, it appears some form of query was being executed onto the object. The question mark ‘?’ in this URI is used to delimit the boundary between the URI of the queryable object and the set of strings used to express a query onto that object. Conventionally, we see the set of strings contains a list of key/value pairs with equal signs ‘=’, which are separated by the ampersand symbol ‘&’ between each of those parameters (e.g. www.youtube[.]com/watch?v=RdcCjDS0s6s&ab_channel=SANSCyberDefense), though the exact structure of the query string is not standardized and different servers may parse it differently. Instead, this case saw a set of strings with the hexadecimal color code #b0f895 (a light shade of green), admin username and password login credentials, and the IP address ‘81.255.222[.]82’ being applied during the object query (via HTTPS protocol on port 8443). In recent months this French IP has also had reports of abuse from the OSINT community.7

On March 2, 2022 at 15:15:33 UTC, the PowerShell loader script further downloaded second-stage executables named ‘sys.exe’ and ‘xmrig.2 sver.8 9 These have been identified as the worm and cryptocurrency miner payloads respectively.

Establish foothold

On March 2, 2022 at 17:46:55 UTC, after the downloads of the worm and cryptocurrency miner payloads, the device initiated multiple SSL connections in a regular, automated manner to Pastebin – a text storage website. This technique was used as a vector to download/upload data and drop further malicious scripts onto the host. OSINT sources suggest the JA3 client SSL fingerprint (05af1f5ca1b87cc9cc9b25185115607d) is associated with PowerShell usage, corroborating with the observation that further tooling was initiated by the PowerShell script ‘ldr.ps1’.

Continual Pastebin C2 connections were still being made by the device almost two months since the initiation of such connections. These Pastebin C2 connections point to new tactics and techniques employed by Sysrv-hello — reports earlier than May do not appear to mention any usage of the file storage site. These new TTPs serve two purposes: defense evasion using a web service/protocol and persistence. Persistence was likely achieved through scheduling daemons downloaded from this web service and shellcode executions at set intervals to kill off other malware processes, as similarly seen in other botnets.10 Recent reports have seen other malware programs also switch to Pastebin C2 tunnels to deliver subsequent payloads, scrapping the need for traditional C2 servers and evading detection.11

Figure 2: A section of the constant SSL connections that the device was still making to ‘pastebin[.]com’ even in the month of April, which resembles beaconing scheduled activity

Throughout the months of March and April, suspicious SSL connections were made from a second potentially compromised device in the internal network to the infected breach device. The suspicious French IP address ‘81.255.222[.]82’ previously seen in the URI object query was revealed as the value of the Server Name Indicator (SNI) in these SSL connections where, typically, a hostname or domain name is indicated.

After an initial compromise, attackers usually aim to gain long-term remote shell access to continue the attack. As the breach device does not have a public IP address and is most certainly behind a firewall, for it to be directly accessible from the Internet a reverse shell would need to be established. Outgoing connections often succeed because firewalls generally filter only incoming traffic. Darktrace observed the device making continuous outgoing connections to an external host listening on an unusual port, 8443, indicating the presence of a reverse shell for pivoting and remote administration.

Figure 3: SSL connections to server name ‘81.255.222[.]8’ at end of March and start of April

Accomplish mission

On March 4, 2022 at 15:07:04 UTC, the device made a total of 16,029 failed connections to a large volume of external endpoints on the same port (8080). This behavior is consistent with address scanning. From the country codes, it appears that public IP addresses for various countries around the world were contacted (at least 99 unique addresses), with the US being the most targeted.

From 19:44:36 UTC onwards, the device performed cryptocurrency mining using the Minergate mining pool protocol to generate profits for the attacker. A login credential called ‘x’ was observed in the Minergate connections to ‘194.145.227[.]21’ via port 5443. JSON-RPC methods of ‘login’ and ‘submit’ were seen from the connection originator (the infected breach device) and ‘job’ was seen from the connection responder (the C2 server). A high volume of connections using the JSON-RPC application protocol to ‘pool-fr.supportxmr[.]com’ were also made on port 80.

When the botnet was first discovered in December 2020, mining pools MineXMR and F2Pool were used. In February 2021, MineXMR was removed and in March 2021, Nanopool mining pool was added,12 before switching to the present SupportXMR and Minergate mining pools. Threat actors utilize such proxy pools to help hide the actual crypto wallet address where the contributions are made by the crypto-mining activity. From April onwards, the device appears to download the ‘xmrig.exe’ executable from a rare IP address ‘61.103.177[.]229’ in Korea every few days – likely in an attempt to establish persistency and ensure the cryptocurrency mining payload continues to exist on the compromised system for continued mining.

On March 9, 2022 from 18:16:20 UTC onwards, trolling for various RCE vulnerabilities (including but not limited to these four) was observed over HTTP connections to public IP addresses:

  1. Through March, the device made around 5,417 HTTP POSTs with the URI ‘/vendor/phpunit/phpunit/src/Util/PHP/eval-stdin.php’ to at least 99 unique public IPs. This appears to be related to CVE-2017-9841, which in PHPUnit allows remote attackers to execute arbitrary PHP code via HTTP POST data beginning with a ‘13 PHPUnit is a common testing framework for PHP, used for performing unit tests during application development. It is used by a variety of popular Content Management Systems (CMS) such as WordPress, Drupal and Prestashop. This CVE has been called “one of the most exploitable CVEs of 2019,” with around seven million attack attempts being observed that year.14 This framework is not designed to be exposed on the critical paths serving web pages and should not be reachable by external HTTP requests. Looking at the status messages of the HTTP POSTs in this incident, some ‘Found’ and ‘OK’ messages were seen, suggesting the vulnerable path could be accessible on some of those endpoints.

Figure 4: PCAP of CVE-2017-9841 vulnerability trolling

Figure 5: The CVE-2017-9841 vulnerable path appears to be reachable on some endpoints

  1. Through March, the device also made around 5,500 HTTP POSTs with the URI ‘/_ignition/execute-solution’ to at least 99 unique public IPs. This appears related to CVE-2021-3129, which allows unauthenticated remote attackers to execute arbitrary code using debug mode with Laravel, a PHP web application framework in versions prior to 8.4.2.15 The POST request below makes the variable ‘username’ optional, and the ‘viewFile’ parameter is empty, as a test to see if the endpoint is vulnerable.16

Figure 6: PCAP of CVE-2021-3129 vulnerability trolling

  1. The device made approximately a further 252 HTTP GETs with URIs containing ‘invokefunction&function’ to another minimum of 99 unique public IPs. This appears related to a RCE vulnerability in ThinkPHP, an open-source web framework.17

Figure 7: Some of the URIs associated with ThinkPHP RCE vulnerability

  1. A HTTP header related to a RCE vulnerability for the Jakarta Multipart parser used by Apache struts2 in CVE-2017-563818 was also seen during the connection attempts. In this case the payload used a custom Content-Type header.

Figure 8: PCAP of CVE-2017-5638 vulnerability trolling

Two widely used methods of SSH authentication are public key authentication and password authentication. After gaining a foothold in the network, previous reports3 19 have mentioned that Sysrv-hello harvests private SSH keys from the compromised device, along with identifying known devices. Being a known device means the system can communicate with the other system without any further authentication checks after the initial key exchange. This technique was likely performed in conjunction with password brute-force attacks against the known devices. Starting from March 9, 2022 at 20:31:25 UTC, Darktrace observed the device making a large number of SSH connections and login failures to public IP ranges. For example, between 00:05:41 UTC on March 26 and 05:00:02 UTC on April 14, around 83,389 SSH connection attempts were made to 31 unique public IPs.

Figure 9: Darktrace’s Threat Visualizer shows large spikes in SSH connections by the breach device

Figure 10: Beaconing SSH connections to a single external endpoint, indicating a potential brute-force attack

Darktrace coverage

Cyber AI Analyst was able to connect the events and present them in a digestible, chronological order for the organization. In the aftermath of any security incidents, this is a convenient way for security users to conduct assisted investigations and reduce the workload on human analysts. However, it is good to note that this activity was also easily observed in real time from the model section on the Threat Visualizer due to the large number of escalating model breaches.

Figure 11: Cyber AI Analyst consolidating the events in the month of March into a summary

Figure 12: Cyber AI Analyst shows the progression of the attack through the month of March

As this incident occurred during a trial, Darktrace RESPOND was enabled in passive mode – with a valid license to display the actions that it would have taken, but with no active control performed. In this instance, no Antigena models breached for the initial compromised device as it was not tagged to be eligible for Antigena actions. Nonetheless, Darktrace was able to provide visibility into these anomalous connections.

Had Antigena been deployed in active mode, and the breach device appropriately tagged with Antigena All or Antigena External Threat, Darktrace would have been able to respond and neutralize different stages of the attack through network inhibitors Block Matching Connections and Enforce Group Pattern of Life, and relevant Antigena models such as Antigena Suspicious File Block, Antigena Suspicious File Pattern of Life Block, Antigena Pastebin Block and Antigena Crypto Currency Mining Block. The first of these inhibitors, Block Matching Connections, will block the specific connection and all future connections that matches the same criteria (e.g. all future outbound HTTP connections from the breach device to destination port 80) for a set period of time. Enforce Group Pattern of Life allows a device to only make connections and data transfers that it or any of its peer group typically make.

Conclusion

Resource hijacking results in unauthorized consumption of system resources and monetary loss for affected organizations. Compromised devices can potentially be rented out to other threat actors and botnet operators could switch from conducting crypto-mining to other more destructive illicit activities (e.g. DDoS or dropping of ransomware) whilst changing their TTPs in the future. Defenders are constantly playing catch-up to this continual evolution, and retrospective rules and signatures solutions or threat intelligence that relies on humans to spot future threats will not be able to keep up.

In this case, it appears the botnet operator has added an object query in the URL of the initial PowerShell loader script download, added Pastebin C2 for evasion and persistence, and utilized new cryptocurrency mining pools. Despite this, Darktrace’s Self-Learning AI was able to identify the threats the moment attackers changed their approach, detecting every step of the attack in the network without relying on known indicators of threat.

Appendix

Darktrace model detections

  • Anomalous File / Script from Rare Location
  • Anomalous File / EXE from Rare External Location
  • Compromise / Agent Beacon (Medium Period)
  • Compromise / Slow Beaconing Activity To External Rare
  • Compromise / Beaconing Activity To External Rare
  • Device / External Address Scan
  • Compromise / Crypto Currency Mining Activity
  • Compromise / High Priority Crypto Currency Mining
  • Compromise / High Volume of Connections with Beacon Score
  • Compromise / SSL Beaconing to Rare Destination
  • Anomalous Connection / Multiple HTTP POSTs to Rare Hostname
  • Device / Large Number of Model Breaches
  • Anomalous Connection / Multiple Failed Connections to Rare Endpoint
  • Anomalous Connection / SSH Brute Force
  • Compromise / SSH Beacon
  • Compliance / SSH to Rare External AWS
  • Compromise / High Frequency SSH Beacon
  • Compliance / SSH to Rare External Destination
  • Device / Multiple C2 Model Breaches
  • Anomalous Connection / POST to PHP on New External Host

MITRE ATT&CK techniques observed:

IoCs

Thanks to Victoria Baldie and Yung Ju Chua for their contributions.

Footnotes

1. https://www.darktrace.com/en/blog/crypto-botnets-moving-laterally

2. https://www.darktrace.com/en/blog/how-ai-uncovered-outlaws-secret-crypto-mining-operation

3. https://www.lacework.com/blog/sysrv-hello-expands-infrastructure

4. https://www.riskiq.com/blog/external-threat-management/sysrv-hello-cryptojacking-botnet

5. https://www.virustotal.com/gui/ip-address/194.145.227.21

6. https://www.virustotal.com/gui/url/c586845daa2aec275453659f287dcb302921321e04cb476b0d98d731d57c4e83?nocache=1

7. https://www.abuseipdb.com/check/81.255.222.82

8. https://www.virustotal.com/gui/file/586e271b5095068484446ee222a4bb0f885987a0b77e59eb24511f6d4a774c30

9. https://www.virustotal.com/gui/file/f5bef6ace91110289a2977cfc9f4dbec1e32fecdbe77326e8efe7b353c58e639

10. https://www.ironnet.com/blog/continued-exploitation-of-cve-2021-26084

11. https://www.zdnet.com/article/njrat-trojan-operators-are-now-using-pastebin-as-alternative-to-central-command-server

12. https://blogs.juniper.net/en-us/threat-research/sysrv-botnet-expands-and-gains-persistence

13. https://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2017-9841

14. https://www.imperva.com/blog/the-resurrection-of-phpunit-rce-vulnerability

15. https://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2021-3129

16. https://isc.sans.edu/forums/diary/Laravel+v842+exploit+attempts+for+CVE20213129+debug+mode+Remote+code+execution/27758

17. https://securitynews.sonicwall.com/xmlpost/thinkphp-remote-code-execution-rce-bug-is-actively-being-exploited

18. https://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2017-5638

19. https://sysdig.com/blog/crypto-sysrv-hello-wordpress

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
Shuh Chin Goh

More in this series

No items found.

Blog

/

/

May 20, 2026

Prompt Security in Enterprise AI: Strengths, Weaknesses, and Common Approaches

prompt securityDefault blog imageDefault blog image

How enterprise AI Agents are changing the risk landscape  

Generative AI Agents are changing the way work gets done inside enterprises, and subsequently how security risks may emerge. Organizations have quickly realized that providing these agents with wider access to tooling, internal information, and granting permissions for the agent to perform autonomous actions can greatly increase the efficiency of employee workflows.

Early deployments of Generative AI systems led many organizations to scope individual components as self-contained applications: a chat interface, a model, and a prompt, with guardrails placed at the boundary. Research from Gartner has shown that while the volume and scope of Agentic AI deployments in enterprise environments is rapidly accelerating, many of the mechanisms required to manage risk, trust, and cost are still maturing.

The issue now resides on whether an agent can be influenced, misdirected, or manipulated in ways that leads to unsafe behavior across a broader system.

Why prompt security matters in enterprise AI

Prompt security matters in enterprise AI because prompts are the primary way users and systems interact with Agentic AI models, making them one of the earliest and most visible indicators of how these systems are being used and where risk may emerge.

For security teams, prompt monitoring is a logical starting point for understanding enterprise AI usage, providing insight into what types of questions are being asked and tasks are being given to AI Agents, how these systems are being guided, and whether interactions align with expected behavior. Complete prompt security takes this one step further, filtering out or blocking sensitive or dangerous content to prevent risks like prompt injection and data leakage.

However, visibility only at the prompt layer can create a false sense of security. Prompts show what was asked, but not always why it was asked, or what downstream actions were triggered by the agent across connected systems, data sources, or applications.

What prompt security reveals  

The primary function of prompt security is to minimize risks associated with generative and agentic AI use, but monitoring and analysis of prompts can also grant insight into use cases for particular agents and model. With comprehensive prompt security, security teams should be able to answer the following questions for each prompt:

  • What task was the user attempting to complete?
  • What data was included in the request, and was any of the data high-risk or confidential?
  • Was the interaction high-risk, potentially malicious, or in violation of company policy?
  • Was the prompt anomalous (in comparison to previous prompts sent to the agent / model)?

Improving visibility at this layer is a necessary first step, allowing organizations to establish a baseline for how AI systems are being used and where potential risks may exist.  

Prompt security alone does not provide a complete view of risk. Further data is needed to understand how the prompt is interpreted, how context is applied, what autonomous actions the agent takes (if any), or what downstream systems are affected. Understanding the outcome of a query is just as important for complete prompt security as understanding the input prompt itself – for example, a perfectly normal, low-risk prompt may inadvertently result in an agent taking a high-risk action.

Comprehensive AI security systems like Darktrace / SECURE AI can monitor and analyze both the prompt submitted to a Generative AI system, as well as the responses and chain-of-thought of the system, providing greater insight into the behavior of the system. Darktrace / SECURE AI builds on the core Darktrace methodology, learning the expected behaviors of your organization and identifying deviations from the expected pattern of life.

How organizations address prompt security today

As prompt-level visibility has become a focus, a range of approaches have emerged to make this activity more observable and controllable. Various monitoring and logging tools aim to capture prompt inputs to be analyzed after the fact.  

Input validation and filtering systems attempt to intervene earlier, inspecting prompts before they reach the model. These controls look for known jailbreak patterns, language indicative of adversarial attacks, or ambiguous instructions which could push the system off course.

Importantly, for a prompt security solution to be accurate and effective, prompts must be continually observed and governed, rather than treated as a point-in-time snapshot.  

Where prompt security breaks down in real environments

In more complex environments, especially those involving multiple agents or extensive tool use, AI security becomes harder to define and control.

Agent-to-Agent communications can be harder to monitor and trace as these happen without direct user interaction. Communication between agents can create routes for potential context leakage between agents, unintentional privilege escalation, or even data leakage from a higher privileged agent to a lower privileged one.

Risk is shaped not just by what is asked, but by the conditions in which that prompt operates and the actions an agent takes. Controls at the orchestration layer are starting to reflect this reality. Techniques such as context isolation, scoped memory, and role-based boundaries aim to limit how far a prompt’s influence can extend.  

Furthermore, Shadow AI usage can be difficult to monitor. AI systems that are deployed outside of formal governance structures and Generative AI systems hosted on unknown endpoints can fly under the radar and can go unseen by monitoring tools, leaving a critical opening where adversarial prompts may go undetected. Darktrace / SECURE AI features comprehensive detection of Shadow AI usage, helping organizations identify potential risk areas.

How prompt security fits in a broader AI risk model

Prompt security is an important starting point, but it is not a complete security strategy. As AI systems become more integrated into enterprise environments, the risks extend to what resources the system can access, how it interprets context, and what actions it is allowed to take across connected tools and workflows.

This creates a gap between visibility and control. Prompt security alone allows security teams to observe prompt activity but falls short of creating a clear understanding of how that activity translates into real-world impact across the organization.

Closing that gap requires a broader approach, one that connects signals across human and AI agent identities, SaaS, cloud, and endpoint environments. It means understanding not just how an AI system is being used, but how that usage interacts with the rest of the digital estate.

Prompt security, in that sense, is less of a standalone solution and more of an entry point into a larger problem: securing AI across the enterprise as a whole.

Explore how Darktrace / SECURE AI brings prompt security to enterprises

Darktrace brings more than a decade of AI expertise, built on an enterprise‑wide platform designed to operate in and understand the behaviors of the complex, ambiguous environments where today’s AI now lives. With Darktrace / SECURE AI, enterprises can safely adopt, manage, monitor, and build AI within their business.  

Learn about Darktrace / SECURE AI here.

Sign up today to stay informed about innovations across securing AI.

[related-resource]

Continue reading
About the author
Jamie Bali
Technical Author (AI) Developer

Blog

/

/

May 20, 2026

State of AI Cybersecurity 2026: 77% of security stacks include AI, but trust is lagging

Default blog imageDefault blog image

Findings in this blog are taken from Darktrace’s annual State of AI Cybersecurity Report 2026.

AI is a contributing member of nearly every modern cybersecurity team. As we discussed earlier in this blog series, rapid AI adoption is expanding the attack surface in ways that security professionals have never before experienced while also empowering attackers to operate at unprecedented speed and scale. It’s only logical that defenders are harnessing the power of AI to fight back.

After all, AI can help cybersecurity teams spot the subtle signs of novel threats before humans can, investigate events more quickly and thoroughly, and automate response. But although AI has been widely adopted, this technology is also frequently misunderstood, and occasionally viewed with suspicion.

For CISOs, the cybersecurity marketplace can be noisy. Making sense of competing vendors’ claims to distinguish the solutions that truly deliver on AI’s full potential from those that do not isn’t always easy. Without a nuanced understanding of the different types of AI used across the cybersecurity stack, it is difficult to make informed decisions about which vendors to work with or how to gain the most value from their solutions. Many security leaders are turning to Managed Security Service Providers (MSSPs) for guidance and support.

The right kinds of AI in the right places?

Back in 2024, when we first conducted this annual survey, more than a quarter of respondents were only vaguely familiar with generative AI or hadn’t heard of it at all. Today, GenAI plays a role in 77% of security stacks. This percentage marks a rapid increase in both awareness and adoption over a relatively short period of time.

According to security professionals, different types of AI are widely integrated into cybersecurity tooling:

  • 67% report that their organization’s security stack uses supervised machine learning
  • 67% report that theirs uses agentic AI
  • 58% report that theirs uses natural language processing (NLP)
  • 35% report that theirs uses unsupervised machine learning

But their responses suggest that organizations aren’t always using the most valuable types of AI for the most relevant use cases.

Despite all the recent attention AI has gotten, supervised machine learning isn’t new. Cybersecurity vendors have been experimenting with models trained on hand-labeled datasets for over a decade. These systems are fed large numbers of examples of malicious activity – for instance, strains of ransomware – and use these examples to generalize common indicators of maliciousness – such as the TTPs of multiple known ransomware strains – so that the models can identify similar attacks in the future. This approach is more effective than signature-based detection, since it isn’t tied to an individual byte sequence or file hash. However, supervised machine learning models can miss patterns or features outside the training data set. When adversarial behavior shifts, these systems can’t easily pivot.

Unsupervised machine learning, by contrast, can identify key patterns and trends in unlabeled data without human input. This enables it to classify information independently and detect anomalies without needing to be taught about past threats. Unsupervised learning can continuously learn about an environment and adapt in real time.

One key distinction between supervised and unsupervised machine learning is that supervised learning algorithms require periodic updating and re-training, whereas unsupervised machine learning trains itself while it works.

The question of trust

Even as AI moves into the mainstream, security professionals are eyeing it with a mix of enthusiasm and caution. Although 89% say they have good visibility into the reasoning behind AI-generated outputs, 74% are limiting AI’s ability to take autonomous action in their SOC until explainability improves. 86% do not allow AI to take even small remediation actions without human oversight.

This model, commonly known as “human in the loop,” is currently the norm across the industry. It seems like a best-of-both-worlds approach that allows teams to experience the benefits of AI-accelerated response without relinquishing control – or needing to trust an AI system.

Keeping humans somewhat in the loop is essential for getting the best out of AI. Analysts will always need to review alerts, make judgement calls, and set guardrails for AI's behavior. Their input helps AI models better understand what “normal” looks like, improving their accuracy over time.

However, relying on human confirmation has real costs – it delays response, increases the cognitive burden analysts must bear, and creates potential coverage gaps when security teams are overwhelmed or unavailable. The traditional model, in which humans monitor and act on every alert, is no longer workable at scale.

If organizations depend too heavily on in-the-loop humans, they risk recreating the very problem AI is meant to solve: backlogs of alerts waiting for analyst review. Removing the human from the loop can buy back valuable time, which analysts can then invest in building a proactive security posture. They can also focus more closely on the most critical incidents, where human attention is truly needed.

Allowing AI to operate autonomously requires trust in its decision-making. This trust can be built gradually over time, with autonomous operations expanding as trust grows. But it also requires knowledge and understanding of AI — what it is, how it works, and how best to deploy it at enterprise scale.

Looking for help in all the right places

To gain access to these capabilities in a way that’s efficient and scalable, growing numbers of security leaders are looking for outsourced support. In fact, 85% of security professionals prefer to obtain new SOC capabilities in the form of a managed service.

This makes sense: Managed Security Service Providers (MSSPs) can deliver deep, continuously available expertise without the cost and complexity of building an in-house team. Outsourcing also allows organizations to scale security coverage up or down as needs change, stay current with evolving threats and regulatory requirements, and leverage AI-native detection and response without needing to manage the AI tools themselves.

Preferences for MSSP-delivered security operations are particularly strong in the education, energy (87%), and healthcare sectors. This makes sense: all are high-value targets for threat actors, and all tend to have limited cybersecurity budgets, so the need for a partner who can deliver affordable access to expertise at scale is strong. Retailers also voiced a strong preference for MSSP-delivered services. These companies are tasked with managing large volumes of consumer personal and financial data, and with transforming an industry traditionally thought of as a late adopter to a vanguard of cyber defense. Technology companies, too, have a marked preference for SOC capabilities delivered by MSSPs. This may simply be because they understand the complexity of the threat landscape – and the advantages of specialized expertise — so well.

In order to help as many organizations as possible – from major enterprises to small and midmarket companies – benefit from enterprise-grade, AI-native security, Darktrace is making it easier for MSSPs to deliver its technology. The ActiveAI Security Portal introduces an alert dashboard designed to increase the speed and efficiency of alert triage, while a new AI-powered managed email security solution is giving MSSPs an edge in the never-ending fight against advanced phishing attacks – helping partners as well as organizations succeed on the frontlines of cyber defense.

Explore the full State of AI Cybersecurity 2026 report for deeper insights into how security leaders are responding to AI-driven risks.

Learn more about securing AI in your enterprise.

[related-resource]

Continue reading
About the author
The Darktrace Community
Your data. Our AI.
Elevate your network security with Darktrace AI