NLTK: Thirteen Path Traversal, RCE, and DoS Vulnerabilities Disclosed Together
Thirteen vulnerabilities in the NLTK Python library, disclosed August 22, 2026, include path traversal, arbitrary file reads, DoS, and RCE risks.
Key findings
- Thirteen vulnerabilities disclosed for NLTK Python library between August 20-22, 2026.
- Multiple path traversal and arbitrary file read flaws affect NLTK versions before 3.10.0.
- Denial of Service via unbounded recursion in JSON parsing (CVE-2026-66393).
- Remote Code Execution possible through insecure unpickling (CVE-2026-71513).
- Affected versions range from pre-3.9.3 to pre-3.10.3, requiring urgent updates.
On August 22, 2026, a batch of thirteen vulnerabilities was disclosed for the Natural Language Toolkit (NLTK) Python library. These vulnerabilities, spanning from August 20 to August 22, 2026, primarily concern path traversal, arbitrary file reads, and denial of service, with several rated as High severity. The disclosures highlight significant security weaknesses in how NLTK handles file paths and data loading, potentially allowing local attackers to access sensitive files or crash the application.
A major theme among these vulnerabilities is path traversal and arbitrary file read flaws. CVE-2026-70626, CVE-2026-65915, CVE-2026-63312, CVE-2026-62388, CVE-2026-62384, and CVE-2026-62385 all involve attackers leveraging manipulated file paths or symlinks to access files outside the intended corpus directory. Specifically, CVE-2026-70626 and CVE-2026-62384 in CorpusReader.open() and FramenetCorpusReader respectively, allow symlink escape and arbitrary file reads. CVE-2026-65915 and CVE-2026-63312 exploit logic bugs in path validation, enabling file reads even when security checks are intended to be active. CVE-2026-62388 points out that security controls in pathsec.py were disabled by default in versions before 3.10.0, making these path traversal attacks easier to execute. CVE-2026-62385 affects FramenetCorpusReader and NKJPCorpusReader, allowing attackers to parse files outside the corpus root via crafted parameters.
Further exacerbating the file access issues, CVE-2026-63311 introduces a Server-Side Request Forgery (SSRF) vulnerability in nltk/pathsec.py's validate_network_url() function. This flaw allows attackers to potentially make requests to arbitrary internal or external resources by exploiting how network URL validation handles DNS resolution failures. Additionally, CVE-2026-63310 highlights a failure to verify file integrity after package downloads, opening the door for man-in-the-middle attacks or DNS poisoning to inject malicious content during package extraction.
Denial of Service (DoS) is also a concern, as demonstrated by CVE-2026-66393. This vulnerability in JSONTaggedDecoder.decode_obj() allows attackers to cause a RecursionError and crash the application by providing deeply nested JSON structures.
A critical vulnerability, CVE-2026-71513, presents a remote code execution (RCE) risk. It affects AllowlistUnpickler and allows attackers to bypass allowlisting by exploiting attribute traversal, enabling the execution of arbitrary callables outside the intended namespace. This could be triggered by crafting untrusted transition-parser data.
The NLTK library has seen multiple patches addressing these issues. Versions before 3.9.4 are affected by CVE-2026-70626. Versions before 3.10.0 are impacted by CVE-2026-65915, CVE-2026-63312, CVE-2026-63311, CVE-2026-62388, and CVE-2026-62385. CVE-2026-63310 affects versions before 3.9.3. CVE-2026-66393 is also fixed in versions 3.9.4 and later. CVE-2026-62384 and CVE-2026-62383 are addressed in versions before 3.10.2. CVE-2026-71514 affects versions 3.9.4 through 3.10.2. Finally, CVE-2026-71513 is fixed in versions before 3.10.3, and CVE-2026-72818 is addressed in versions before 3.10.2. Users are strongly advised to update to the latest available versions to mitigate these risks.
This extensive batch of vulnerabilities underscores the importance of robust security practices in Python libraries, particularly those handling external data and file system interactions. The prevalence of path traversal and arbitrary file read flaws suggests a need for more rigorous input validation and secure file handling mechanisms within NLTK. Users should prioritize updating their NLTK installations to the patched versions to protect against potential exploitation.
CVE-2026-70626, CVE-2026-66393, CVE-2026-65915, CVE-2026-63312, CVE-2026-63311, CVE-2026-63310, CVE-2026-62388, CVE-2026-62385, CVE-2026-62384, CVE-2026-62383, CVE-2026-71514, CVE-2026-71513, CVE-2026-72818