JNDI injection in Fastjson autoType processing
CVE-2025-70974 affects Alibaba Fastjson versions before 1.2.48. Improper handling of autoType allows an attacker-controlled JSON @type value to identify a Java class whose public methods may be invoked during processing. Depending on those methods, an attacker-supplied payload elsewhere in the same document can enable JNDI injection and potentially remote code execution. The vulnerability results from an incomplete fix for CVE-2017-18349 and was exploited in the wild from 2023 through 2025. A later bypass is tracked separately as CVE-2022-25845.
What it means and what to do now: patch path, interim mitigations, and the assume-compromise checklist.
What an attacker gets, and what they’ve been doing with it.
Processing attacker-controlled JSON can invoke public methods of attacker-selected Java classes and enable JNDI injection. Depending on the invoked methods and runtime environment, exploitation may lead to remote code execution.
If you can’t patch tonight, do this now.
Until upgrading, avoid processing untrusted JSON with Fastjson autoType enabled. Disable or strictly restrict autoType, block dangerous class-resolution paths, and prevent attacker-controlled @type values from resolving arbitrary classes. Where possible, prevent untrusted inputs from reaching Fastjson deserialization.
Patch, then assume compromise.
Upgrade com.alibaba:fastjson to version 1.2.48 or later to address this specific vulnerability. Version 1.2.48 should not be treated as resolving all related autoType vulnerabilities; account separately for the later bypass tracked as CVE-2022-25845.
No public exploits tracked yet. Mallory keeps watching.
12 sources tracked across advisories, community write-ups, and news. New activity surfaces here as Mallory finds it.