← All articles

Exploited vulnerabilities hit four systems in active attacks

Exploited vulnerabilities hit four systems in active attacks

Exploited vulnerabilities affecting four systems are being used in attacks linked to cryptocurrency mining, ransomware and LLMjacking, according to the U.S. Cybersecurity and Infrastructure Security Agency (CISA). The agency added flaws in Synology DiskStation Manager (DSM), BerriAI LiteLLM, Apache Axis and Vanderbilt University Research Electronic Data Capture (REDCap) to its Known Exploited Vulnerabilities (KEV) catalog, which tracks security flaws with evidence of active exploitation.

The four vulnerabilities are CVE-2026-90436 in Apache Axis, CVE-2026-90817 in REDCap, CVE-2026-40217 in LiteLLM and CVE-2024-10441 in Synology DSM. Each has a high or critical Common Vulnerability Scoring System (CVSS) rating, a standard measure of how serious a security flaw can be. The Apache Axis, REDCap and Synology flaws each have a CVSS score of 9.8, while the LiteLLM flaw is rated 8.8.

CVE-2026-90436 is an unsafe reflection vulnerability in Apache Axis that could let an unauthenticated remote attacker run arbitrary code. In other words, an attacker would not need to log in before attempting to make the affected server execute commands. The flaw affects Axis 1.x, which has reached end of life and remains unpatched. Recommended options are migrating to another SOAP engine, such as Apache Axis 2/Java, or using an Axis build with the patch from a July 12, 2026, commit applied.

The KEV listing followed reports that attackers were using CVE-2026-90436 together with the 4EverAble2Shell vulnerabilities affecting N-able N-central to deploy cryptocurrency miners on vulnerable servers. Rapid7 separately detected exploitation against Atlassian Confluence Data Center in a customer environment because Confluence uses the Axis library for its XML-RPC and SOAP application interfaces. Atlassian said Confluence is vulnerable when the Remote API (XML-RPC & SOAP) is enabled. That setting was disabled by default for new installations from Confluence 3.0, and the feature was fully removed in Confluence 10.0.

Rapid7 said the activity was flagged on November 13, 2026. The attackers used shell commands and scripts fetched from temporary Pinggy and Cloudflare Tunnel domains, then examined the compromised environment. Just over three minutes after the first observed execution, they authenticated to the database server using Confluence's database credentials and queried page bodies, application links and attachments with names associated with credentials, tokens, VPN configurations and private keys.

The scripts, which Rapid7 said were likely generated by artificial intelligence, searched Confluence's default attachment location, /var/atlassian/application-data/confluence/attachments/v4/. They parsed Office documents, searched for credentials and looked for payment card numbers, using Luhn checks to reduce false positives. The attackers also mapped internal services, used a Confluence application credential to query the internal Atlassian Crowd API for users and group memberships, and attempted AS-REP Roasting, a Kerberos-based method for harvesting credentials.

Those efforts were unsuccessful, but the attackers obtained a service account credential and queried a production cloud data warehouse. About 12 hours after the first exploit request, they signed in to Microsoft 365 as an integration service account through legacy authentication from 192.238.253[.]176.

CVE-2026-90817 is a code injection flaw in REDCap that could allow an unauthenticated attacker to run arbitrary code on the server. CISA said an attacker could send crafted HTTP requests that reach an unintended controller route from a public survey context and provide a crafted file-path or stream parameter during data import. Exploitation requires knowledge of a valid public survey hash.

Safe Security said it observed an unknown threat abusing this REDCap flaw to stage a temporary web shell, meaning a hidden tool for remotely controlling a web server. The attacker then used an in-memory loader to retrieve XMRig, a cryptocurrency miner, disguised as image files. The miner was made to resemble syslog-ng and persistence was established through cron and an added SSH key, allowing the malicious activity to survive future sessions or restarts.

CVE-2026-40217 affects LiteLLM and is an unprotected alternate channel vulnerability. It could allow a remote attacker to run arbitrary code through bytecode rewriting at the /guardrails/test_custom_code URI. The Google Threat Intelligence Group reported that attackers used the flaw to turn limited-access inference keys into full administrative credentials and maintain access during LLMjacking attacks. LLMjacking is the unauthorised use of artificial intelligence services or computing resources for an attacker's benefit. The affected targets included financial services, retail, transportation and technology organisations in North America, South America and the Asia Pacific region.

The fourth flaw, CVE-2024-10441, is an improper encoding or escaping of output vulnerability in Synology DSM. It could allow a remote attacker to run arbitrary code through unspecified methods. Synology disclosed earlier this month that the flaw was being exploited against Network Attached Storage (NAS) devices, which are systems used to store files on a network. Those attacks resulted in the deployment of C3RB3R ransomware, malicious software that encrypts or otherwise disrupts access to data.

Because the four vulnerabilities are being actively exploited, CISA recommended that federal agencies apply available fixes or necessary mitigations for the DSM, REDCap and Axis flaws by November 21, 2026. The recommended deadline for the LiteLLM flaw is December 2, 2026. Organisations should identify whether any of these products are exposed, especially where public surveys, remote APIs, AI inference services or network storage are involved.

For teams that need broader help securing systems and infrastructure, AEU-I provides security-first IT, infrastructure and consulting. This does not replace vendor remediation, but it is relevant to the wider task of reviewing exposed services and affected environments.

Also available in: DeutschItalianoShqip