Disclaimer: This content was generated using AI. While we strive for accuracy, please verify critical information with official sources.
Sources: National Vulnerability Database, CISA KEV, Exploit-DB
CVE-2026-108684 Overview
CVE-2026-108684 is a SQL injection vulnerability affecting erzhongxmu Jeewms up to version 3.7. The flaw resides in the getTreeData function within src/main/java/com/jeecg/demo/controller/JeecgFormDemoController.java, part of the Autocomplete Data Handler component. An attacker can manipulate the searchVal request parameter to inject arbitrary SQL into a backend Hibernate query. The attack can be launched remotely over the network and requires only low-privileged access. The maintainer has published a fix in commit 6e29bd57972a499e9c8a81a2dbe94d0d5cf23af0, and the issue is tracked under [CWE-74] (Improper Neutralization of Special Elements in Output Used by a Downstream Component).
Critical Impact
Remote authenticated attackers can inject arbitrary SQL through the searchVal parameter, exposing confidentiality, integrity, and availability of the backing database.
Affected Products
- erzhongxmu Jeewms versions up to and including 3.7
- Component: Autocomplete Data Handler (JeecgFormDemoController)
- Deployments using the vulnerable getAutocompleteData endpoint
Discovery Timeline
- 2026-10-11 - CVE-2026-108684 published to NVD
- 2026-10-15 - Last updated in NVD database
Technical Details for CVE-2026-108684
Vulnerability Analysis
The vulnerability exists in the getAutocompleteData handler of JeecgFormDemoController. The controller reads the searchVal HTTP parameter directly from the request and concatenates it into a Hibernate Query Language (HQL) string. Because the value is not escaped or bound as a parameter, an attacker-controlled payload becomes part of the executed query. The affected endpoint accepts both GET and POST requests, broadening the attack surface. Successful exploitation permits data extraction, record manipulation, or query-based denial of service against the backing database. The CWE classification [CWE-74] reflects the broader category of injection resulting from improper neutralization of user input in a downstream interpreter.
Root Cause
The root cause is unsafe string concatenation when constructing the HQL statement from TSUser where userName like '%" + searchVal + "%'. Hibernate executes the resulting query verbatim, so any SQL metacharacters supplied in searchVal alter the query structure rather than being treated as literal data.
Attack Vector
An authenticated remote attacker sends a crafted GET or POST request to the autocomplete endpoint with a malicious searchVal value. No user interaction is required. Attackers can use standard SQL injection techniques such as boolean-based blind, UNION-based, or time-based payloads to enumerate schema and exfiltrate data.
@RequestMapping(params = "getAutocompleteData", method = {RequestMethod.GET, RequestMethod.POST})
public void getAutocompleteData(HttpServletRequest request, HttpServletResponse response) {
String searchVal = request.getParameter("searchVal");
- String hql = "from TSUser where userName like '%" + searchVal + "%'";
- List autoList = systemService.findHql(hql);
+ String hql = "from TSUser where userName like ?";
+ List autoList = systemService.findHql(hql, new Object[]{"%" + searchVal + "%"});
try {
response.setContentType("application/json;charset=UTF-8");
response.setHeader("Pragma", "No-cache");
Source: GitHub Commit 6e29bd5. The patch replaces string concatenation with a positional bind parameter, forcing Hibernate to treat searchVal as data.
Detection Methods for CVE-2026-108684
Indicators of Compromise
- HTTP requests to getAutocompleteData containing SQL metacharacters such as ', --, ;, UNION, or SLEEP( in the searchVal parameter.
- Unexpected database errors or Hibernate QuerySyntaxException entries referencing the TSUser entity.
- Anomalous response-time spikes on the autocomplete endpoint indicating time-based blind SQL injection.
Detection Strategies
- Deploy web application firewall (WAF) rules that inspect searchVal for SQL injection patterns on both GET and POST requests.
- Enable database query logging and alert on queries against TSUser that contain nested SELECT, UNION, or boolean tautologies.
- Compare deployed JeecgFormDemoController.java against commit 6e29bd57972a499e9c8a81a2dbe94d0d5cf23af0 to identify unpatched instances.
Monitoring Recommendations
- Forward application and database logs to a centralized SIEM for correlation of authentication events with autocomplete endpoint usage.
- Track per-user request rates to getAutocompleteData to detect automated injection attempts.
- Monitor outbound database connections for abnormal result-set sizes consistent with mass data extraction.
How to Mitigate CVE-2026-108684
Immediate Actions Required
- Apply commit 6e29bd57972a499e9c8a81a2dbe94d0d5cf23af0 from the upstream repository to all Jeewms deployments.
- Restrict access to the getAutocompleteData endpoint to authenticated internal users where feasible.
- Review database audit logs for prior exploitation attempts against TSUser.
Patch Information
The maintainer merged the fix via Pull Request #4, applied in commit 6e29bd5. The patch rewrites the HQL query to use a bound parameter instead of string concatenation. See the GitHub Security Advisory GHSA-mc7c-w398-v7xp and the VulDB entry for additional detail.
Workarounds
- Block or rewrite requests to getAutocompleteData at a reverse proxy or WAF until patching is complete.
- Add server-side input validation that rejects non-alphanumeric characters in searchVal before the controller executes.
- Apply least-privilege permissions on the database account used by Jeewms to limit the impact of successful injection.
# Example WAF rule (ModSecurity) to block SQLi patterns in searchVal
SecRule ARGS:searchVal "@rx (?i)(union(\s)+select|sleep\(|--|;|')" \
"id:1026108684,phase:2,deny,status:403,msg:'CVE-2026-108684 SQLi attempt in searchVal'"
Disclaimer: This content was generated using AI. While we strive for accuracy, please verify critical information with official sources.
Sources: National Vulnerability Database, CISA KEV, Exploit-DB

