Unsafe Pickle Deserialization in NLTK AllowlistUnpickler

NLTK before 3.10.3 contains a remote code execution vulnerability in AllowlistUnpickler that validates only the pickle module string and not the global name, allowing attackers to resolve dotted names by attribute traversal to callables outside the allowlisted namespace. Attackers can craft untrusted transition-parser models that execute arbitrary commands when TransitionParser.parse loads the model through allowlisted_pickle_load.

Casky was already ahead

This CVE exploits attack patterns that Casky's 446matched 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

NLTK versions before 3.10.3 contain a critical remote code execution vulnerability in the AllowlistUnpickler class, which fails to properly validate pickle objects during deserialization. While the implementation attempts to restrict unpickling to allowlisted modules, it only validates the pickle module string itself and overlooks the global name resolution. This allows attackers to bypass the allowlist by using attribute traversal to access callables outside the intended namespace. The vulnerability is particularly dangerous because it affects the TransitionParser.parse() function, which automatically loads transition-parser models—a common operation in natural language processing pipelines. Any organization using NLTK for NLP tasks that processes untrusted models or accepts external parser models is at risk of arbitrary code execution with the privileges of the application.

Casky's 446 matching skills leverage Claude's extended reasoning to detect the attack patterns underlying this vulnerability across the MITRE ATT&CK framework. Practitioners using Casky would identify initial access vectors (TA0001) through malicious model delivery mechanisms, execution techniques (TA0002) where arbitrary commands run during TransitionParser.parse invocation, and persistence methods (TA0043) that attackers could establish post-compromise. The platform would flag suspicious pickle deserialization patterns, attribute traversal abuse in object instantiation, and unexpected subprocess or system calls spawned from NLTK processing workflows. By correlating these behavioral indicators with the specific CWE-502 insecure deserialization pattern, security teams gain actionable intelligence to detect compromised models before they execute malicious payloads in production environments.

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 →

446 Casky skills cover this attack pattern

These skills use Claude AI's reasoning model to surface findings in the same attack categories as CVE-2026-71513.

Abuse of Public-Facing API: Mobile API Abuse

mobile security · low

Run

abusing-dpapi-for-credential-access

red teaming · high

Run

abusing-shadow-credentials-for-privesc

red teaming · high

Run

MITRE ATT&CK Techniques

TA0001 TA0043 TA0002

Investigate the attack patterns behind CVE-2026-71513

Casky has 446 skills that investigate the attack patterns behind CVE-2026-71513. Run one and get CVSS-scored findings in 3 minutes.

Run the skill that detects this →

X Instagram LinkedIn