Insecure Deserialization in PyTorch Model Loading

CosyVoice thru commit 6e01309e01bc93bbeb83bdd996b1182a81aaf11e (2025-30-21) contains an insecure deserialization vulnerability (CWE-502) in its make_parquet_list.py data processing tool. The script loads PyTorch .pt files (utterance embeddings, speaker embeddings, speech tokens) using torch.load() without enabling the weights_only=True security parameter. This allows the deserialization of arbitrary Python objects via the pickle module. An attacker can exploit this by providing malicious .pt files within a data directory. When a victim processes this directory using the tool, arbitrary code is executed on the victim's system.

Casky was already ahead

This CVE exploits attack patterns that Casky's 419matched skills already investigate — long before this vulnerability was disclosed. Claude's reasoning model maps these techniques to MITRE ATT&CK, so practitioners who ran these skills have already seen the threat behaviour in their findings.

Analysis

CosyVoice, a voice synthesis platform, contains a critical insecure deserialization vulnerability in its make_parquet_list.py utility. The vulnerability stems from using torch.load() without the weights_only=True parameter when processing PyTorch .pt files containing utterance embeddings, speaker embeddings, and speech tokens. This allows attackers to execute arbitrary Python code by crafting malicious .pt files that deserialize untrusted objects through pickle. Any organization using CosyVoice for voice generation, voice cloning, or speech processing pipelines is at risk, particularly those accepting user-supplied audio models or embeddings.

Casky's 419 mapped skills leverage Claude AI with extended reasoning to detect attack patterns across three MITRE ATT&CK phases: Initial Access (TA0001), Discovery (TA0043), and Execution (TA0002). Practitioners using Casky would identify suspicious file handling behaviors—specifically unexpected deserialization operations on .pt files, anomalous process spawning following model loading, and lateral movement attempts post-exploitation. The platform's reasoning engine correlates insecure coding patterns (missing weights_only parameter) with runtime indicators like unauthorized file access to model directories and unexpected outbound connections, providing defenders with precise detection rules and remediation guidance tied to this specific attack chain.

Live Threat IntelligenceComing soon

Casky Risk Score

—

EPSS Probability

—

Shodan Exposure

—

GreyNoise Signal

—

Composite risk scoring from EPSS, CISA KEV, Shodan, and GreyNoise — 21 security APIs correlated into a single Casky Risk Score. Coming in Casky Pro. Join early access →

419 Casky skills cover this attack pattern

These skills use Claude AI's reasoning model to surface findings in the same attack categories as CVE-2026-31249.

analyzing-android-malware-with-apktool

malware analysis · medium

Run

analyzing-apt-group-with-mitre-navigator

threat intelligence · low

Run

analyzing-bootkit-and-rootkit-samples

malware analysis · medium

Run

MITRE ATT&CK Techniques

TA0001 TA0043 TA0002

Investigate the attack patterns behind CVE-2026-31249

Casky has 419 skills that investigate the attack patterns behind CVE-2026-31249. Run one and get CVSS-scored findings in 3 minutes.

Run the skill that detects this →

X Instagram LinkedIn