Python stack vulnerabilities · Page 81

BoxLite: Timeout Bypass Vulnerability

Boxlite is a sandbox service that allows users to create lightweight virtual machines (Boxes) and launch OCI containers within them to run untrusted code. In versions 0.8.2 and prior, Boxlite allows users to configure a timeout for services running inside the virtual machine. When the timeout is triggered, Boxlite sends a signal to kill the process. However, instead of using the uncatchable SIGKILL signal, Boxlite uses the catchable SIGALRM signal. Malicious code running inside the sandbox can exploit this vulnerability to continue running after the timeout is triggered, leading to resource exhaustion within the virtual machine and affecting

Dulwich's submodule path traversal in porcelain.submodule_update / porcelain.clone(recurse_submodules=True) yields RCE via attacker-dropped .git/hooks payload

Dulwich is a pure-Python implementation of the Git file formats and protocols. Starting in version 0.23.2 and prior to version 1.2.5, `dulwich.porcelain.submodule_update`, and by extension `porcelain.clone(..., recurse_submodules=True)`, materializes attacker-controlled submodule paths from a crafted upstream repository without path validation. A malicious `.gitmodules` plus a matching tree gitlink whose `path` is `.git/hooks` (or any other directory inside the parent repository's `.git` directory) causes the attacker's submodule tree contents to be written directly into the victim's `.git/hooks/` directory, preserving executable mode bits. T

• Update to 1.2.5
• Update to 1.2.5-1

Dulwich has unbounded memory allocation in receive-pack from crafted thin packs

Dulwich is a pure-Python implementation of the Git file formats and protocols. Starting in version 0.1.0 and prior to version 1.2.5, a client with push access could push a tiny crafted thin pack (~174 bytes) whose delta header declares a huge dest_size. When dulwich ingested it via add_thin_pack / apply_delta, it would allocate hundreds of MB of memory based on that attacker-controlled size, with no relationship to the actual bytes received. Operators running a Dulwich-based Git server that exposes git-receive-pack (i.e. accepts pushes) - for example via dulwich.server functionality, the HTTP smart server, or anything built on ReceivePa

• Update to 1.2.5
• Update to 1.2.5-1

Dulwich doesn't sanitize commit subjects in `porcelain.format_patch`

Dulwich is a pure-Python implementation of the Git file formats and protocols. Starting in version 0.24.0 and prior to version 1.2.5, dulwich.porcelain.format_patch(outdir=...) derives each patch filename from the commit's subject line. Prior to this fix, get_summary only replaced spaces with dashes - path separators (/, \), parent-directory components (..), and other filename-hostile characters (e.g. :) were preserved verbatim and passed straight into os.path.join(outdir, f"{i:04d}-{summary}.patch"). A malicious commit subject could therefore direct the generated patch file outside the requested outdir. This is fixed in Dulwich 1.2.5. Users

• Update to 1.2.5
• Update to 1.2.5-1

Dulwich has an arbitrary file write via NTFS-hostile tree entries on Windows

Dulwich is a pure-Python implementation of the Git file formats and protocols. Versions starting with 0.10.0 and prior to 1.2.5 have an arbitrary file write leading to remote code execution when cloning or checking out a malicious Git repository on Windows. Dulwich's path-element validator accepted tree entries whose filenames contained bytes that Windows interprets as structural path syntax. Contributing configuration bugs made matters worse. The core.protectNTFS and core.protectHFS settings were looked up under a wrong option name and so user-set values were silently ignored, and core.protectNTFS only defaulted to true on Windows (Git upstr

• Update to 1.2.5

Dulwich Vulnerable to Command Injection via Merge Driver Path

Dulwich is a pure-Python implementation of the Git file formats and protocols. Starting in version 0.24.0 and prior to version 1.2.5, Dulwich's `ProcessMergeDriver` substitutes the file path (from the git tree, controllable by an attacker via a malicious branch) into the merge driver command via the `%P` placeholder and executes it with `subprocess.run(..., shell=True)`. An attacker who can cause a victim to merge an untrusted branch can achieve arbitrary command execution by crafting malicious file paths. Version 1.2.5 fixes the issue.

• Update to 1.2.5
• Update to 1.2.5-1

kafka-python prior to 2.3.2 DoS via SCRAM Iteration Count in scram.py

kafka-python prior to 2.3.2 contains a denial-of-service vulnerability in SCRAM authentication handling that allows a malicious or machine-in-the-middle broker to freeze the client event loop by supplying an excessively large iteration count. In scram.py, ScramClient.process_server_first_message() passes the broker-controlled SCRAM iteration count directly to hashlib.pbkdf2_hmac() without validation, blocking producer sends, consumer polls, admin operations, and heartbeats, which can cause consumer group eviction and repeated reconnect failures.

• Update to 2.0.2-12
• Update to 2.3.2

kafka-python prior to 2.3.2 Denial of Service via Protocol Parser Frame Length

kafka-python prior to 2.3.2 contains a denial-of-service vulnerability in the protocol parser that allows a malicious broker or machine-in-the-middle attacker to exhaust memory or hang connections by sending a crafted 4-byte frame length value without bounds validation. Attackers can send a specially crafted frame length through the receive_bytes() function to trigger either a multi-gigabyte memory allocation or an uncaught ValueError that leaves the connection in a broken state, causing requests to hang and consumers to stop heartbeating until restart.

• Update to 2.0.2-13
• Update to 2.3.2

Weblate: Stored HTML injection in editor search preview

Weblate is a web based localization tool. Prior to version 2026.5, Weblate's live search preview renders unit source and context as HTML without escaping. Any contributor whose content reaches those fields stores HTML and CSS that runs inside the authenticated editor of every user who runs a matching search. This issue has been patched in version 2026.5.

• Update to 2026.5

Weblate SSRF: outbound URL guard misses the NAT64 well-known prefix (64:ff9b::/96)

Weblate is a web based localization tool. From version 5.15 to before version 2026.6, Weblate's VCS_RESTRICT_PRIVATE did not properly account for some transitional IPv6 ranges, multicast addresses, or some semi-private IPv4 ranges, which allowed some addresses to bypass private range restrictions. This issue has been patched in version 2026.6.

• Update to 2026.6

SSRF via sitemap-derived URLs in Crawlee for Python

Crawlee is a web scraping and browser automation library. From version 1.0.0 to before version 1.7.0, Crawlee is vulnerable to SSRF via sitemap-derived URLs. This issue has been patched in version 1.7.0.

• Update to 1.7.0

Pipecat: Path Traversal in Pipecat Runner `/files` Endpoint — Arbitrary File Read via `%2F`-Encoded Separator

Pipecat is an open-source Python framework for building real-time voice and multimodal conversational agents. From version 0.0.90 to before version 1.2.0, a path traversal vulnerability exists in Pipecat's development runner (src/pipecat/runner/run.py). When the runner is started with the --folder flag, it exposes a GET /files/{filename:path} download endpoint. The filename path parameter is concatenated directly onto args.folder with no containment check. Starlette normalises literal ../ sequences in URLs, but %2F-encoded slashes bypass this normalisation: the path parameter is URL-decoded after routing, so ..%2F..%2Fetc%2Fpasswd resolves to

• Update to 1.2.0

LMDeploy: Hardcoded trust_remote_code=True is an implicit unsafe remote-code load path with no user opt-out

LMDeploy is a toolkit for compressing, deploying, and serving large language models. In versions 0.12.3 and prior, hardcoded "trust_remote_code=True" enables HF supply-chain RCE without user opt-in. Version 0.13.0 patches the issue.

• Update to 0.13.0

LMDeploy: Arbitrary code execution via hardcoded trust_remote_code=True in lmdeploy model initialization

LMDeploy is a toolkit for compressing, deploying, and serving large language models. In versions 0.12.3 and prior, LMDeploy is vulnerable to arbitrary code execution through hardcoded "trust_remote_code=True" in multiple HuggingFace model-loading call sites. At time of publication, there are no publicly available patches.

• Update to 0.13.0

SQLFluff: Uncontrolled Resource Consumption in Parser

SQLFluff is a modular SQL linter and auto-formatter with support for multiple dialects and templated code. Prior to version 4.2.0, in deployments where untrusted users can provide SQL queries to be linted, an untrusted user can submit a malicious long query to any application using the parser to trigger a Denial of Service through resource exhaustion. This issue has been patched in version 4.2.0.

• Update to 4.2.0
• Update to 4.3.0-1

SQLFluff: Recursive Stack Overflow in Parser

SQLFluff is a modular SQL linter and auto-formatter with support for multiple dialects and templated code. Prior to version 4.1.0, in deployments where untrusted users can provide SQL queries to be linted, an untrusted user can submit a malicious query with deliberate excessive nesting to any application using the parser to trigger a Denial of Service through resource exhaustion. This issue has been patched in version 4.1.0.

• Update to 4.1.0
• Update to 4.3.0-1

durabletask 1.4.1, 1.4.2, and 1.4.3 contain malicious code distributed via a compromised maintainer account

`durabletask` versions 1.4.1, 1.4.2, and 1.4.3 were published on 2026-05-19 within a 35-minute window through a compromised PyPI maintainer account and contained malicious code. On import, the package fetched a remote payload (`rope.pyz`) from an attacker-controlled host and executed it. The payload was a credential-theft framework that interrogated cloud instance metadata (AWS/Azure/GCP) and secret stores, harvested Kubernetes service-account tokens, HashiCorp Vault tokens, and credentials from known filesystem paths, attempted to brute-force password manager vaults. Anything obtained was exfiltrated to command-and-control infrastructure

Awxkit: path traversal via yaml !include directive

A path traversal vulnerability was found in awxkit, the CLI tool for AWX. The YAML !include directive does not sanitize file paths, allowing an attacker to craft a malicious YAML file that reads arbitrary YAML-formatted files from the local filesystem when a user imports it using "awx --conf.format yaml import". This is a client-side vulnerability requiring user interaction.

Apache Airflow Samba provider: Path traversal in GCSToSambaOperator via GCS object names

The Apache Airflow Samba provider's `GCSToSambaOperator` joined GCS object names to the SMB destination path without a containment check, so an object named with `../` segments resolved a write path outside the configured `destination_path`. An attacker able to write objects into the source GCS bucket — typically an external data producer distinct from the trusted DAG author — could write files to arbitrary locations on the Samba target when the operator ran. Upgrade apache-airflow-providers-samba to 4.12.6 or later, which validates the resolved destination stays within `destination_path`.

• Update to 4.12.6

bz2.BZ2Decompressor reuse after error can cause a stack buffer overflow

bz2.BZ2Decompressor objects could be reused after a decompression error. If an application caught the resulting OSError and retried with the same decompressor, crafted input could cause the decompressor to resume from an invalid internal state and perform out-of-bounds writes to a stack buffer. This could crash the process when processing untrusted data.

• Update to 3.10.21
• Update to 3.11.16
• Update to 3.11.2-6+deb12u9
• Update to 3.12.14
• Update to 3.13.14
• Update to 3.13.14-1
• Update to 3.13.5-2+deb13u3
• Update to 3.14.6
• Update to 3.14.6-1
• Update to 3.15.0b3