Article
Remote Access and IoT Search Engines
Arete Analysis
Recently, Arete Incident Response Tiger Teams (“Arete IR”) have responded to an increased volume of ransomware incidents involving the Sodinikibi, Phobos, and Dharma ransomware variants. The threat actors deploying these variants are known to use anti-forensics techniques to hide their tracks. Once access is gained, they usually delete artifacts which aid cyber investigators with reconstructing steps taken by the threat actors, revealing important information pertaining to the root cause for the computer security event. In most cases, Arete was able to recover critical artifacts to forensically reconstruct the various attacks to identify a single entry point consistent across the three variants: Remote Desktop.
Requirements Businesses have a need to enable workers to access files and business resources remotely from home, hotels, or business relationships. The cheapest way for businesses to allow remote access is to “expose” the Remote Desktop Protocol (“RDP”) to the public internet. The business’ firewall configuration is altered to allow inbound connectivity to the default port 3389, and any connections to that port are automatically forwarded to a specific computer on the network, which is usually a terminal services server. Using only an Internet Protocol (“IP”) address, anyone can attempt to connect to the RDP service.
Limitations on Protocol and Service Most businesses who implement remote access via RDP aren’t aware of the limitations of the service nor do they implement intrusion detection and prevention services. Lastly, many don’t require multi-factor authentication. The downside to allowing any connection into a network is exactly that: any connection can be allowed into the network. This connection can be from anywhere, at any time, for any reason and with any number of authentication attempts. The RDP service itself doesn’t monitor for bad credential combinations and automatically disable or block connection attempts. Port forwarding on firewalls doesn’t inspect the inbound traffic either. Essentially, once a port is exposed to the public internet, anyone, anywhere, can try an unlimited number of usernames and passwords to gain access to that system. Since any number of combinations can be attempted, this makes the configuration vulnerable to credential stuffing, dictionary, and brute force attacks.
Crime of Opportunity Quite often during our investigations, clients ask “was this a targeted attack or a crime of opportunity?” Nine out of 10 times, it’s a crime of opportunity. Then the follow up question “Why us?” Well, for starters, it’s your configuration. These threat actors have their attack mechanics down to a series of steps: 1. Identify target 2. Gain access to target 3. Cover tracks 4. Deploy ransomware 5. Repeat
While they most likely aren’t outright targeting your organization directly, they may be targeting exposed services which link them to your organization.
Internet of Things (IoT) Search Engines
Google, Bing, and DuckDuckGo are three very popular search engines. They’re used to find all sorts of text information or images. These search engines aren’t designed to identify specific computers or services across the world. Rather a different set of search engines can be used to find computers that are connected directly to the internet along with their IP addresses and any other information about the computer involving their geo location, running services, and protocol history. Use caution when visiting these sites as unintended side effects can occur.
https://zoomeye.org
These sites can be used by anyone, anonymously, to identify internet attached devices, the services they’re running and any other information the IoT crawlers can index. The anonymous feature is obtained via the IoT indexer by allowing anyone to query the index stored by the IoT search engine, instead of scanning the node directly. Essentially, this search engine is the phone book, allowing anyone to find street addresses by person’s names or people by street addresses.
A search for “port:3389”, which is the default port for RDP services, can return several million devices. Again, this isn’t real time information because the query is run against the index of the IoT search engine. Once an IP address is identified, additional steps would be needed to test if the IP address is online. Additionally, the resulting information can be filtered by organization, operating system, and country.
Reviewing the results, there’s approximately 1,060 IP addresses that are detected as the Windows 2003 operating system. At face value, this is extremely alarming because Windows 2003 was discontinued during July of 2015. Microsoft officially stopped supporting the operating system as well as providing security updates. Given the information returned from shodan.io, businesses are still relying on it as a means for remote connectivity. Again, these results would need to be qualified as online and available. Regardless, the number is still alarming.
Attack Methodologies
After the threat actor identifies a target, any number of steps can be performed to initiate an attack. Typically, the threat actor will profile the target to gain as much information as possible in order to increase the success of the attack. Profiling can occur in any of the following ways:
Verifying the IP address is online and attempting to brute force access automatically.
Attempting to resolve the IP address to a domain name or company name in order to:
Construct phishing emails for obtaining credentials.
Employ social engineering of employees for obtaining credentials
Research running services against known vulnerabilities to identify pre-built payloads to exploit the services.
Whichever approach the threat actor takes, there’s a good chance they will be successful with gaining unauthorized access to your network.
Preventative Actions
While it’s a waiting game to become the next victim, there are steps you can take to mitigate or prolong falling prey to these threat actor methodologies. Successful mitigation of unauthorized access can be achieved through the proper implementation of layered computer and network security controls. The following steps, while not exhaustive, can be taken to mitigate the exposure of services used by your organization.
Enable Multi-Factor Authentication (“MFA”) on any third party accounts or remote access on any third party accounts or remote access accounts. accounts.
Disable RDP services and port forwarding on firewalls.
Implement VPN services to remotely connect to your organization’s network or leverage to your organization’s network or leverage remote
connection technologies that remote connection technologies that support MFA. support MFA.
Research open source intelligence to develop a public footprint of your organization. a public footprint of your organization.
Train employees on social engineering and phishing email tactics and techniques. phishing email tactics and techniques.
Purchase a cyber insurance policy and familiarize yourself with the preferred vendors yourself with the preferred vendors within your policy. within your policy.
Build a close working relationship with a cyber advisory company.
Back to Blog Posts
Article
LiteLLM Supply-Chain Compromise
A software supply-chain attack compromised the release pipeline of LiteLLM, a widely used open-source Python SDK and AI gateway. Attackers leveraged the compromised pipeline on October 21st to publish a malicious LiteLLM version through PyPI and associated Docker registries. The affected release, LiteLLM 1.107.0, contained malware designed to steal credentials, sensitive configuration data, private keys, browser information, and other secrets from compromised systems.
What’s Notable and Unique
The incident is particularly significant because LiteLLM provides a common API for applications to interact with AI models from multiple providers. It is reported that the project had approximately 61,800 GitHub stars and 92 million monthly PyPI downloads.
Rather than modifying the core LiteLLM application source code, the threat actors reportedly tampered with the installation package itself. The malware was designed for broad credential and information theft and encrypted the collected data before uploading it to the attacker's endpoint.
The compromise went beyond one-time credential theft. A malicious Python component launched a detached loader that checked a GitHub-hosted pointer every hour and could download Python code from an external server for execution on the compromised machine. This provided threat actors with a mechanism for subsequent remote code execution.
Analyst Comments
Although the malicious packages were subsequently removed, simply uninstalling or deleting the affected LiteLLM package may not be sufficient, because malicious Python startup files or the detached loader could remain on the host. All exposed information should be rotated immediately, and infected systems should be isolated and rebuilt from clean backups. The most recent clean release prior to the attack, LiteLLM 1.106.2, is currently the suggested recovery version.
Sources
LiteLLM release pipeline hacked to distribute info-stealer malware
LiteLLM 1.107.0 Compromised: Python Startup Hooks Steal Cloud, AI, and Proxy Credentials
The Collector Reaches Beyond Provider API Keys
Article
French Supermarket Cyberattack Highlights Risks in Exposed Refrigeration Controllers
Hundreds of French supermarkets were reportedly affected by a cyberattack in which threat actors compromised internet-connected refrigeration controllers and manipulated temperature and defrost settings, which caused refrigeration units to overheat, resulting in widespread food spoilage and significant financial loss. Impacted refrigeration systems were associated with Danfoss, Copeland, and Dixell, and demonstrated how vulnerabilities in industrial control systems (ICS) can disrupt physical operations without relying on conventional data theft or encryption. The attack also raised concerns about the wider exposure of similar devices, as thousands of refrigeration management interfaces remain accessible from the public internet. The incident highlights the growing risk posed by insecure operational technology (OT), particularly where legacy equipment and remote-management interfaces lack adequate security controls.
What’s Notable and Unique
Researchers identified approximately 2,765 internet-accessible Danfoss AK-SM 800A interfaces during assessment. This figure reflects the scope of that research and should not be interpreted as a current count of exposed devices. Publicly accessible management interfaces increase the opportunity for unauthorized access to industrial equipment.
Research also identified several security flaws in the Danfoss AK-SM 800A platform, including an authentication bypass command injection that could enable remote code execution and a weakness involving web-server configuration changes. Danfoss addressed these findings through firmware updates.
Authorities reported compromised Danfoss AK-SM and AK-SC systems, alongside Copeland XWEB and Dixell XWEB monitoring platforms.
Analyst Comments
The incident demonstrates the business consequences of insufficiently secured industrial infrastructure and reinforces the need to integrate OT security into enterprise risk management. Organizations operating connected facility systems should establish clear ownership of industrial assets, enforce secure remote-access policies, and prioritize remediation based on exposure and operational criticality. The findings also highlight the importance of coordinating security teams with facilities and engineering personnel to ensure that protective measures do not compromise essential operations. A risk-based approach to OT security can help organizations reduce exposure while maintaining the availability and reliability of critical equipment.
Sources
France Supermarket Hacks Expose Risks of Internet-Facing Refrigeration Systems
Freeze the Controller, Defrost the Food: Uncovering Vulnerabilities in Danfoss Refrigeration Controllers
Attackers Heat French Supermarket Fridges Up to 50°C by Hijacking Danfoss Controllers
Article
FortiBleed Campaign Remains Active
The FBI and Secret Service recently warned that the FortiBleed credential-harvesting campaign remains an active threat targeting internet-facing Fortinet FortiGate firewalls and secure socket layer (SSL) virtual private network (VPN) gateways. The agencies report that attackers continue to scan internet-exposed Fortinet firewalls using previously obtained credentials. This has reinforced assessments made by other security researchers, who have stated that “FortiBleed should be treated as an active access operation, not as a one-time credential leak.”
What’s Notable and Unique
FortiBleed, active since February 2026, begins with widespread reconnaissance to identify internet-facing FortiGate firewalls and SSL-VPN services. Once potential targets are identified, operators attempt to gain initial access through credential stuffing, password spraying, and dictionary attacks using previously compromised credentials.
Following successful access, the attackers deploy the Go-based FortigateSniffer tool on compromised appliances. The tool leverages native FortiOS diagnostic functionality to passively intercept authentication traffic across 24 protocols, enabling the collection of cleartext credentials and password hashes passing through the device.
Successfully recovered credentials are reused to expand access, including Active Directory enumeration, Kerberos validation, SMB authentication, and network share discovery, enabling threat actors to move deeper into the victim environment.
Reporting indicates that initial access brokers (IABs) leveraging the FortiBleed attack chain have facilitated access for downstream ransomware operations, including the INC Ransom/Lynx and Payload ransomware groups. This aligns with Arete’s observations of ransomware groups, including INC, Qilin, and DragonForce, targeting Fortinet FortiGate and VPN infrastructure as an initial access vector.
Remediation
Organizations should restrict internet-facing FortiGate management interfaces, rotate administrative and VPN credentials, review local and privileged accounts, and enforce MFA for remote access. If a potential compromise is detected, organizations should isolate affected systems, conduct threat hunting to determine the scope and identify compromised accounts and devices, and preserve relevant logs and artifacts. Organizations should also report the incident to appropriate authorities and implement targeted eviction measures based on identified threat actor activity.
Analyst Comments
FortiBleed represents an ongoing access-enablement threat, rather than a one-time credential exposure. The campaign’s ability to harvest, validate, and reuse credentials can provide sustained access to victim environments and create opportunities for downstream ransomware activity. Organizations with suspected FortiGate exposure should treat the activity as a potential active compromise and investigate for persistent access, compromised accounts, and lateral movement.
Sources
FortiBleed Operations Continue Targeting Exposed Systems Leading to Reports of Lockouts
FBI Warns FortiBleed Remains Active After Amassing 86,644 Fortinet Device Credentials
Article
Threat Actors Exploiting Critical FortiMail Zero-Day Vulnerability
Fortinet has disclosed a critical vulnerability in FortiMail, tracked as CVE-2026-104286, that is currently being exploited as a zero-day in the wild. The flaw, which carries a Critical CVSS score of 9.8, affects the FortiMail management interface and can allow unauthenticated attackers to write arbitrary files to affected systems, potentially leading to unauthorized code execution and full system compromise. Fortinet has confirmed active exploitation and urged customers to implement mitigations immediately while awaiting security updates.
What’s Notable and Unique
The vulnerability stems from improper restriction of file paths combined with improper neutralization of NULL byte characters. An unauthenticated attacker can exploit the flaw via specially crafted HTTP or HTTPS requests to write arbitrary files on the underlying operating system.
Organizations running FortiMail 7.2 can remediate the vulnerability by upgrading to 7.4 or newer. For customers using affected FortiMail 7.4, 7.6, and 8.0 versions, fixes have not yet been released. However, Fortinet has confirmed that the issue will be addressed in the upcoming 7.4.9, 7.6.7, and 8.0.2 releases.
Until the patched versions become available, Fortinet recommends mitigating risk by disabling the Identity-Based Encryption (IBE) feature. As an additional safeguard, administrators should restrict access to the FortiMail management interface, either by removing internet exposure altogether or limiting access to trusted internal networks only. The advisory also includes log entries that administrators can use to identify potentially compromised appliances.
The U.S. Cybersecurity and Infrastructure Security Agency (CISA) has added CVE-2026-104286 to its Known Exploited Vulnerabilities (KEV) catalog and directed federal agencies to perform forensic triage and mitigation by October 4th.
Analyst Comments
Impacted organizations should review their FortiMail instances to ensure there are no unauthorized logins, examine device settings, evaluate all configurations as potentially compromised, and carry out the necessary recovery procedures. Defenders should immediately implement Fortinet's recommended mitigations, review systems for IOC matches, and prepare to deploy the forthcoming security updates as soon as they become available. It is strongly advised that all enterprises assess their setups to reduce risk, as protecting publicly accessible management interfaces is a fundamental security best practice.
Sources
Improper limitation of a pathname to a restricted directory