NextGen Mirth Connect Flaws Expose Downstream System Logins

Health records storage vendor NextGen Connect has a batch of high-severity flaws that could let attackers walk away with administrator passwords that open access to databases and downstream systems the logins connect to, an independent researcher found.

See Also: Reduce Cloud Risk in Healthcare with Security by Default

The U.S. Cybersecurity and Infrastructure Security Agency on Thursday added the authenticated SQL injection flaw, CVE-2026-82583 , and two unsafe XML parser vulnerabilities, CVE-2026-78224 and CVE-2026-82578 , to its catalog, all with CVSS scores between 7 and 8.

The cloud-based Mirth Connect platform by NextGen Healthcare is an integration engine used by hundreds of organizations in more than 40 countries to exchange medical data between clinical and business systems.

"This is the box that shuttles lab results and discharge summaries between hospital systems, and it typically holds credentials for the systems it talks to. That is a lot of trust in one server," said Abhinav Agarwal, who found the flaws with autonomous agents running with his published methodology on artificial intelligence-assisted vulnerability discovery.

The bugs are found in the last open-source release published by NextGen, version 4.5.2, after which the company shifted to a commercial, proprietary model. Agarwal said he reported the vulnerabilities to CISA and NextGen, which issued fixes in version 4.7.2 on Aug. 5.

"The tested vulnerable release, 4.5.2, is the last tagged public one, while the fixes sit in 4.7.x commercial builds," Agarwal said. "Customers can get patches, but outsiders cannot inspect them."

The SQL injection flaw can expose administrator password hashes, all channel definitions and full server configurations, Agarwal said. Because Mirth 4.5.2 stores connector passwords unencrypted, attackers could easily access plain-text credentials for connected databases, file servers, mail relays and clinical endpoints.

The sensitive information is written into a directory the product serves over HTTP with no login required, so the data can be collected without the original account, Agarwal said. "Export that and you have working logins for whatever is downstream."

The same malicious request can also freeze the bundled database, preventing maintainers from reversing the action through the same interface and forcing them to restart the server, Agarwal said.

The flaw lies in the Database Connector endpoint, which accepts a caller-controlled selectLimit value and inserts it directly into a database query without separating user input from the SQL command. It allows authenticated users who can reach the endpoint to manipulate queries against Mirth's bundled Apache Derby database.

The two other vulnerabilities can be exploited without a Mirth account, Agarwal said. NextGen Connect routes incoming data through independent processing pipelines called channels. The bugs affect deployed channels that receive messages from connected systems, allowing an attacker to exploit them by sending a malicious input, he said.

"The attack is just a message - the same kind of message the channel was built to accept," Agarwal said.

CVE-2026-78224 stems from an unsafe parser used when XML messages are transformed into a different format as they move through the pipeline, Agarwal said. An unauthenticated client can exploit the flaw by sending a malicious message through an affected channel to read files stored on the server.

CVE-2026-82578 affects a separate XML parser in the XML Batch Adaptor, but it can also let "an unauthenticated channel client read a server-local file when XML batch processing is enabled in an affected split mode," Agarwal said.

Either XML external entity flaw can allow attackers to send the contents of files readable by the Mirth service to an address they control.

Agarwal said his findings quickly received attention from the vendor because NextGen Connect had previously been exploited. A high-severity deserialization flaw, tracked as CVE-2023-43208, was exploited in 2024 after an incomplete patch for CVE-2023-37679, disclosed in 2023, left the software vulnerable to unauthenticated remote code execution.

Separately, NextGen Healthcare has faced a series of class action lawsuits since 2023 after a system intrusion exposed data belonging to 1 million individuals. The cases were ultimately consolidated and finalized in March with a $19.4 million settlement.

Electronic Healthcare Records

Governance & Risk Management

Cybersecurity reporter

Wang previously reported on cybersecurity for IANS. She received a master's in data journalism from Columbia University in August 2025. She attended Boston University.