Loading vulnerability…
CVE-2026-5147 is a SQL injection vulnerability in YunaiV yudao-cloud up to version 2026.01. The issue affects an unspecified code path associated with the /admin-api/system/tenant/get-by-website endpoint. According to the available information, improper handling of the Website argument allows attacker-controlled input to influence SQL query construction or execution, resulting in SQL injection. The vulnerability is remotely exploitable, and public exploit information is available. Specific details about the vulnerable function or underlying ORM/query logic are not currently available.
Mallory correlates every CVE against your assets, your vendors, and active adversary campaigns. Know which vulnerabilities matter for you, not just which ones are loud.
What it means. What to do now. Patch path, mitigations, and the assume-compromise checklist.
What an attacker gets, and what they’ve been doing with it.
If you can’t patch tonight, do this now.
Patch, then assume compromise.
1 valid exploit after Mallory filtered fakes, detection scripts, and README-only repos.
Repository is a small standalone Python exploit project consisting of one main script (cve-2026-5147.py), a README, requirements.txt, and .gitignore. It is not part of a larger exploitation framework. The Python script is the sole operational component and uses the requests library with optional retry handling to send crafted HTTP GET requests to a user-supplied target. Purpose: exploit a SQL injection vulnerability in a web API parameter to perform blind extraction against a MySQL backend. The script supports two inference modes: boolean-based blind SQL injection using EXP/INSTR or EXP/ASCII payloads, and time-based blind SQL injection using IF(...,SLEEP(...),0). According to the README and visible code, the workflow first attempts calibration of a boolean oracle by checking a JSON response field (default 'code'); if that fails, it automatically falls back to time-based calibration and extraction unless the user disables calibration. Main capabilities: build target URL from host or full URL, choose HTTP or HTTPS, set injectable parameter name, add custom headers and extra query parameters, tune timeout/sleep thresholds, and extract VERSION() and DATABASE() by default. Under the dump subcommand, it can enumerate schema information and dump data, including table names, column names, and row contents from the current database, with options such as schema-only dumping, row limits, custom scalar subqueries, and output to dump files. Code structure: helper functions create payloads (payload_instr, payload_ascii, payload_time_if_ascii), bind time-based payload builders, parse JSON fields, evaluate truth conditions, and perform iterative character-by-character extraction in extract_string(). The extraction loop tests candidate ASCII values for each character position, sends HTTP requests with the injected parameter, and interprets either JSON content or elapsed time as the oracle. The script prints progress to stderr and final results to stdout, indicating it is intended for interactive operator use. Notable observables: no hardcoded victim infrastructure is present; targets are fully operator-supplied. The only fixed network indicators are generic HTTP/HTTPS scheme handling and default request headers. The repository also references dump output files matching sqldump_*.txt. Overall, this is an operational blind-SQLi exploitation utility for authorized testing rather than a mere detector or README-only proof of concept.
2 sources tracked across advisories and community write-ups. News coverage will land here when it surfaces.
No news coverage yet. Advisories and community discussion only.
Query your assets running an affected version, and investigate the blast radius.
Every observed campaign linking this CVE to a named adversary.
Malware families riding this exploit, with evidence and IOCs.
YARA, Sigma, Snort, and vendor rules, auto-deployed to your SIEM.
Cross-references every affected SKU, including bundled OEM variants.
Community discussion across Reddit, Mastodon, and other social sources.