In psd-tools versions up to 1.17.0, the SmartObject.save() method writes embedded smart objects to file paths taken directly from the PSD file without sanitization, enabling arbitrary file write through path traversal (e.g., using '../'). The filename property is attacker-controlled and unsanitized, allowing writes outside the intended directory. Additionally, SmartObject.open() for external-kind smart objects uses an attacker-controlled fullPath descriptor without directory constraints, allowing arbitrary file reads. These vulnerabilities enable an attacker to write arbitrary bytes to arbitrary paths and read arbitrary files, potentially exfiltrating sensitive data. Both issues are fixed in version 1.17.1.

An attacker who can supply or manipulate PSD files processed by psd-tools up to version 1.17.0 can cause the software to write arbitrary data to arbitrary filesystem locations, potentially overwriting critical files. They can also read arbitrary files from the filesystem and write their contents to an output directory, enabling data exfiltration. This can lead to unauthorized file modification and disclosure.

Upgrade psd-tools to version 1.17.1 or later, where these vulnerabilities are fixed. The vendor advisory confirms that versions prior to 1.17.1 are affected and that the issues are resolved in 1.17.1. No other mitigations are specified.