The newest Python resources are on the Python page. For a reading order instead of a date order, see the Python learning path.
| Date Added | Link | Excerpt |
|---|---|---|
| 2026-05-20 2026 | Microsoft's durabletask package on PyPi Compromised. Mini Shai Hulud attacks again... again! news 8 min read | Library versions 1.4.1, 1.4.2, and 1.4.3 of Microsoft's `durabletask` Python package on PyPI were compromised with a dropper that executes a sophisticated infostealer and worm. This payload targets credentials from cloud providers, password managers, and developer tools, propagating via SSM or kubectl in cloud environments. It also includes a destructive component triggered by specific system locales. The malware exfiltrates data encrypted with an RSA key and utilizes a GitHub-based dead-drop for command and control. → aikido.dev |
| 2026-05-12 2026 | What AI 'fingerprints' helped expose the 1st AI-made zero-day exploit? | The exploit was a Python script beginner | Researchers discovered the first zero-day exploit generated by AI. The exploit was written as a Python script. The article's title suggests that unique "AI fingerprints" were crucial in identifying this novel threat, distinguishing it from human-crafted exploits. This marks a significant development in cybersecurity, highlighting AI's potential for both creating and detecting sophisticated attacks. The specific details of these "fingerprints" and how they led to the exposure of the exploit are likely discussed within the linked content. |
| 2026-05-10 2026 | JDownloader Website Supply Chain Attack: Installers Replaced with Python RAT Malware (May 2026) news 5 min read | Writeup of the JDownloader website supply chain attack (May 2026), detailing how an unpatched CMS vulnerability allowed attackers to replace Windows and Linux installers with a Python RAT and ELF binaries respectively. The attack, active for approximately 24 hours, utilized obfuscation and persistence techniques, including SUID-root binaries for Linux. This incident highlights the risks of unauthorized changes to web content and the importance of verifying digital signatures. → rescana.com |
| 2026-05-08 2026 | Linux Kernel Elevation of Privilege Vulnerability news 1 min read | Writeup on CVE-2026-31431, a "Copy Fail" logic bug in the Linux kernel's authencesn cryptographic template. This vulnerability allows an unprivileged local user to perform a deterministic, controlled 4-byte write into the page cache of any readable file, enabling elevation of privilege to root. The exploit is a 732-byte Python script that can modify setuid binaries, impacting all Linux distributions shipped since 2017. Vendor-specific fixes are available for Ubuntu, Debian, Red Hat, SUSE, Amazon, Arch, AlmaLinux, Cloudlinux, and Gentoo. → hkcert.org |
| 2026-05-05 2026 | Bootstrap script exposes PyPI to domain takeover attacks news 8 min read Supply Chain | Library detailing a domain takeover vulnerability in legacy Python package bootstrap scripts. The vulnerability, discovered by ReversingLabs, affects numerous packages including tornado and slapos.core, by exploiting the now-available python-distribute[.]org domain. This could allow attackers to execute arbitrary code when developers run affected bootstrap scripts, potentially impacting software supply chain security. → reversinglabs.com |
| 2026-05-02 2026 | Script Injection and Data Theft: Python Data Analysis Tool Compromised intermediate 2 min read RCE | Tool update compromise elementary-data (version 0.23.3) allowed script injection via GitHub Actions workflows, leading to theft of SSH keys, AWS credentials, API tokens, and cryptocurrency wallet files. The malicious package was available on PyPI and as a Docker image. Countermeasures include uninstalling the compromised version, installing 0.23.4, renewing credentials, and checking for malware marker files. This incident is cataloged as MAL-2026-3083 in OSV. |
| 2026-05-01 2026 | "Copy Fail": Linux root in all major distributions with 732 bytes of Python intermediate 2 min read RCE | Writeup of CVE-2026-31431, "Copy Fail," a Linux kernel vulnerability allowing local privilege escalation. This logic error enables a deterministic 4-byte write to the page cache of any readable filesystem, exploitable with a 732-byte Python script. The vulnerability, discovered with AI assistance and affecting major distributions since 2017, can be mitigated by blocking AF_ALG socket creation or blacklisting the algif_aead module. |
| 2026-04-30 2026 | New Linux 'Copy Fail' Vulnerability Enables Root Access on Major Distributions news 2 min read RCE | Library for detecting and mitigating the "Copy Fail" vulnerability (CVE-2026-31431), a Linux kernel flaw in the algif_aead module. This high-severity issue allows unprivileged local users to gain root access by writing controlled bytes into the page cache of any readable file, targeting setuid binaries like `/usr/bin/su`. The exploit, a small Python script, corrupts the page cache of files it doesn't own, bypassing sandboxing and affecting distributions shipped since 2017, including RHEL and Ubuntu. → thehackernews.com |
| 2026-04-23 2026 | wapiti-scanner/wapiti: Web vulnerability scanner written in Python3 beginner 5 min read API Sec | Library for "black-box" web vulnerability scanning in Python3, acting as a fuzzer to detect issues like SQL Injections, XSS, File Disclosure, Command Execution, XXE, CRLF Injection, Shellshock, SSRF, and Log4Shell (CVE-2021-44228) by attacking scripts and forms. It supports various reporting formats, session management for resuming scans, proxy integration, authenticated scans, URL scope limitation, cookie import from browsers, and includes modules for CMS enumeration, subdomain takeover detection, and security header checks. |
| 2026-04-22 2026 | CVE-2025-68664: Critical LangChain Flaw Enables Secret Extraction news 4 min read | Writeup of CVE-2025-68664, a critical serialization injection vulnerability in LangChain Core, enabling secret extraction and unintended object instantiation. The flaw, stemming from improper handling of the "lc" key during data serialization and deserialization, affects Python versions >= 1.0.0 and < 1.2.5 and < 0.3.81, and a similar issue, CVE-2025-68665, impacts LangChain.js. Exploitation involves crafting attacker-controlled LLM outputs that masquerade as trusted objects, leading to risks like secret leakage and network operations. Patched versions implement deserialization allowlists and disable environment-based secret loading by default. → socradar.io |
| 2026-04-22 2026 | Bandit Python: Free SAST in 10 Seconds (2026 Review) beginner 3 min read | Library for static analysis of Python code, Bandit identifies common security issues through Abstract Syntax Tree analysis. It ships with 47 built-in checks targeting vulnerabilities like hardcoded credentials, weak cryptography, and injection flaws, with specialized plugins for issues such as insecure Hugging Face model downloads (B615). Bandit offers flexible configuration, multiple output formats including SARIF, baseline comparisons for incremental scans, and integrates with pre-commit hooks and Docker. It’s recommended for Python projects needing a free, focused security linter to complement broader SAST solutions. → appsecsanta.com |
| 2026-04-22 2026 | CVE-2026-22607: Fickling Python RCE Vulnerability news 4 min read | Writeup of CVE-2026-22607 details an Insecure Deserialization vulnerability in Fickling, a Python pickling decompiler. Versions up to 0.1.6 incorrectly classify pickle files using `cProfile.run()` as "SUSPICIOUS" instead of "OVERTLY_MALICIOUS". This misclassification allows attackers to craft malicious pickle files, bypass Fickling's analysis, and achieve arbitrary code execution on systems relying on its security assessment for deserialization. → sentinelone.com |
| 2026-04-22 2026 | CVE-2026-21226: Azure Core Python Library RCE Vulnerability news 4 min read | Library for Python applications using Azure SDKs, addressing CVE-2026-21226, an insecure deserialization vulnerability (CWE-502). Attackers with low-level authorization can execute arbitrary code over a network by crafting malicious serialized payloads processed by the vulnerable Azure Core library. Mitigation involves updating the `azure-core` package via `pip install --upgrade azure-core` and implementing input validation or network segmentation. → sentinelone.com |
| 2026-04-22 2026 | SGLang CVE-2026-5760 (CVSS 9.8) Enables RCE via Malicious GGUF Model Files news 2 min read RCE | Writeup on CVE-2026-5760, a CVSS 9.8 remote code execution vulnerability in SGLang. Attackers exploit this by crafting malicious GGUF model files with Jinja2 server-side template injection payloads in the `tokenizer.chat_template` parameter. Loading these models and hitting the `/v1/rerank` endpoint allows arbitrary Python code execution on the SGLang server, similar to CVE-2024-34359 (Llama Drama) and CVE-2025-61620 in vLLM. Mitigation involves using `ImmutableSandboxedEnvironment` for rendering templates. → thehackernews.com |
| 2026-04-22 2026 | Marimo RCE Flaw CVE-2026-39987 Exploited Within 10 Hours of Disclosure news 3 min read RCE | Writeup on CVE-2026-39987, a pre-authenticated RCE vulnerability in Marimo exploited within 10 hours of disclosure. The flaw, unpatched until version 0.23.0, allowed unauthenticated attackers to gain a full PTY shell by connecting to the `/terminal/ws` WebSocket endpoint without proper authentication. Attackers leveraged the exploit for credential theft, environment variable extraction, and deployment of the NKAbuse variant via Hugging Face Spaces, with CISA adding it to the KEV catalog. → thehackernews.com |
| 2026-04-22 2026 | Critical SQL Injection Vulnerability in Django (CVE-2025-64459) news 4 min read | Library detailing CVE-2025-64459, a critical SQL injection vulnerability in Django that allows attackers to manipulate query logic via internal parameters like `_connector` and `_negated`. The analysis covers exploitation scenarios such as authentication bypass and data exfiltration, outlines the fix implemented in patched Django versions (5.2.8, 5.1.14, 4.2.26), and provides mitigation strategies including code review, parameter whitelisting, and testing for vulnerable patterns. → endorlabs.com |
| 2026-04-22 2026 | Malicious PyPI Packages Deliver SilentSync RAT news 7 min read | Library for Python package installers that delivers the SilentSync RAT. Malicious PyPI packages named `sisaws` and `secmeasure`, uploaded by the same author, are used to deploy SilentSync. This RAT enables remote command execution, file exfiltration, screen capturing, and web browser data theft from Chrome, Brave, Edge, and Firefox on Windows systems. The malicious packages leverage typosquatting and mimic legitimate library functionalities to evade detection. |
| 2026-04-22 2026 | Bearer: SAST Tool to Discover, Filter, and Prioritize Security and Privacy Risks beginner 8 min read | Tool for static application security testing (SAST), Bearer scans source code to identify, filter, and prioritize security and privacy risks. It supports multiple languages including Go, Java, JavaScript, TypeScript, PHP, Python, and Ruby, with advanced cross-file analysis and additional languages available in its commercial version. Bearer CLI detects vulnerabilities based on OWASP Top 10 and CWE Top 25, such as Path Traversal (CWE-22), SQL Injection (CWE-89), and Cross-Site Scripting (CWE-79), and also identifies PII and PHI data flows for privacy compliance reporting. |
| 2026-04-19 2026 | PyPI Supply Chain Attack: Colorama and Colorizr Name Confusion news 5 min read | Library of malicious Python packages exploiting typo-squatting and name-confusion attacks against the Colorama library on PyPI. These packages, designed to mimic legitimate libraries, deliver payloads for persistent remote access, data exfiltration, and attempts to evade antivirus controls on both Windows and Linux systems. The campaign exhibits cross-ecosystem tactics, using NPM package names to target PyPI users, and features sophisticated persistence mechanisms and stealth techniques. → checkmarx.com |
| 2026-04-19 2026 | Compromised LiteLLM PyPI Package Delivers Credential Stealer news 6 min read Supply Chain | Library versions 1.82.7 and 1.82.8 of the popular Python package litellm, an abstraction for interacting with LLMs from providers like OpenAI and Google, were compromised on PyPI. This malicious code acted as a multi-stage credential stealer, exfiltrating sensitive data including API keys, cloud provider credentials, and Kubernetes secrets. The payload employed AES-256-CBC encryption for data and RSA for key protection, ultimately attempting to establish persistence via a system service and download further payloads from attacker-controlled infrastructure. → sonatype.com |
| 2026-04-19 2026 | LiteLLM PyPI Package Compromised in TeamPCP Supply Chain Attack news 3 min read | Library compromised in a supply-chain attack, where malicious versions of the LiteLLM Python package (1.82.7 and 1.82.8) were uploaded to PyPI by the TeamPCP hacking group. These versions deployed an infostealer that harvested sensitive data including SSH keys, cloud credentials, Kubernetes secrets, and cryptocurrency wallet data. The payload also attempted lateral movement and installed a persistent systemd backdoor, exfiltrating data to attacker-controlled infrastructure. Organizations are advised to rotate credentials and inspect systems for persistence artifacts. → bleepingcomputer.com |
| 2026-04-19 2026 | Malicious PyPI Package — LiteLLM Supply Chain Compromise news 1 min read | Writeup detailing the LiteLLM supply chain compromise, where malicious Python `.pth` files in `site-packages/` automatically execute embedded, double base64-encoded payloads. These payloads exfiltrate environment variables, SSH keys, and cloud credentials to attacker-controlled servers like `models[.]litellm[.]cloud`. The attack, attributed to TeamPCP, exploits versions 1.82.8 and potentially 1.82.7 of LiteLLM, necessitating immediate credential rotation and checks for suspicious `.pth` files. |
| 2026-04-19 2026 | The PyPI Supply Chain Attacks of 2025 news Supply Chain | The PyPI Supply Chain Attacks of 2025 |
| 2026-04-16 2026 | PYPI Security: How to Prevent Supply Chain Attacks in Python Projects beginner | PYPI Security: How to Prevent Supply Chain Attacks in Python Projects |
| 2026-04-16 2026 | Python Tools for Penetration Testers beginner 6 min read | Library offering Python tools for penetration testers, vulnerability researchers, and reverse engineers. It lists libraries and programs for packet manipulation (Scapy, Impacket, dpkt), network reconnaissance (AutoRecon, Mitm6, SMBMap), web application security (XSStrike, Powerfuzzer, waymap), fuzzing (afl-python, Peach Fuzzing Platform), disassembly and emulation (Capstone, Unicorn Engine, Frida, Angr), memory analysis (Volatility, Rekall), and reverse engineering of applications (Androguard, Ghidatron, pefile). |
| 2026-04-16 2026 | Escalating Deserialization Attacks in Python intermediate 7 min read | Library for escalating Python deserialization attacks, demonstrating how insecure deserialization with `pickle` can lead to Remote Code Execution (RCE). The entry details how to exploit Python 2 and Python 3 vulnerabilities using techniques like `__reduce__` methods, `eval`, `compile`, and `exec` to achieve code injection and access sensitive files like `/etc/passwd` without leaving obvious artifacts of direct shell access. |
| 2026-04-16 2026 | Exploiting Python Pickles - David Hamann intermediate 5 min read | Writeup detailing the exploitation of Python's `pickle` module for remote code execution. It explains how the `__reduce__` method can be abused during deserialization to execute arbitrary commands, demonstrating this with a Flask application and a reverse shell payload. The writeup emphasizes the security risks of unpickling untrusted data and suggests alternatives like JSON or data signing. |
| 2026-04-16 2026 | Attack on Software Supply Chains Using Fake Python Infrastructure intermediate 9 min read | Library detailing a sophisticated software supply chain attack where an attacker distributed malware by creating a fake Python infrastructure with a typosquatted domain and a malicious "colorama" package. This campaign affected over 170,000 users, leveraging compromised GitHub accounts and multi-stage execution with obfuscation techniques to steal sensitive data like credentials and session tokens from various applications. → checkmarx.com |
| 2026-04-16 2026 | Defense in Depth: A Practical Guide to Python Supply Chain Security beginner 32 min read | Library for Python supply chain security, detailing defenses against attacks like the Ultralytics compromise. It advocates for layered security, starting with static analysis using Ruff, dependency pinning with cryptographic hashes via uv, and vulnerability scanning with pip-audit. The library also covers generating SBOMs with CycloneDX for rapid impact assessment, and adopting Trusted Publishing with OIDC for secure package distribution, emphasizing that no single control is perfect but multiple layers mitigate risk. |
| 2026-04-16 2026 | How Python Pickle Deserialization Security Exploit Works intermediate 16 min read | Library for understanding Python pickle deserialization exploits, detailing how attackers can achieve arbitrary code execution by crafting malicious pickle payloads, often leveraging the `__reduce__` method for command injection via functions like `os.system`. The resource highlights common attack vectors such as insecure file uploads and vulnerable API endpoints that blindly deserialize untrusted data, leading to real-world impacts like remote code execution and data breaches. |
| 2026-04-16 2026 | Insecure Deserialization in Python: Attack Techniques and Secure Coding intermediate 8 min read | Library for understanding and mitigating insecure deserialization vulnerabilities in Python. It details attack techniques using Python's `pickle`, `PyYAML` (specifically `yaml.load`), `jsonpickle`, `shelve`, and `marshal`. The resource covers how attackers fingerprint applications, craft payloads like reverse shells and OS command injection, and deliver them via HTTP, and includes steps for post-exploitation such as privilege escalation and lateral movement. Secure coding practices, including using `yaml.safe_load`, are also discussed. |
| 2026-04-16 2026 | The Complete Guide on Python for Cyber Security beginner 9 min read | Library for integrating Python into cybersecurity workflows, aiding penetration testers, security analysts, incident responders, researchers, and network security engineers. It leverages Python's readability, extensive libraries like Requests, Scapy, Pandas, PyTorch, pwntools, Atheris, CrowdStrike FalconPy, vt-py, YARA, pySigma, and PyMISP, and its versatility to automate tasks, build exploits, analyze data, detect anomalies, and manage infrastructure, supporting tools like Nmap, Metasploit, and Burp Suite. |
| 2026-04-13 2026 | Critical flaw in Marimo Python notebook exploited within 10 hours of disclosure news 3 min read | Writeup of CVE-2026-39987, a critical pre-authentication RCE vulnerability in Marimo Python notebooks, which allows unauthenticated attackers to gain a full shell and execute arbitrary commands. Exploited within 10 hours of disclosure, this flaw affects Marimo versions prior to 0.23.0 and enables credential theft in under three minutes. The vulnerability stems from an unauthenticated terminal WebSocket endpoint, highlighting risks in AI-adjacent developer tools like MLflow and Langflow. → csoonline.com |
| 2026-04-10 2026 | This Python notebook flaw shows how fast hackers are acting on advisories news | This Python notebook flaw shows how fast hackers are acting on advisories https://ift.tt/U56juBE → cybernews.com |
| 2026-04-10 2026 | Python CVE Details beginner | Python CVE Details |
| 2026-04-10 2026 | Python Security Vulnerabilities CVE Database beginner | Python Security Vulnerabilities CVE Database |
| 2026-04-10 2026 | Picklescan Allows RCE via Malicious Pickle File intermediate 1 min read | Advisory GHSA-655q-fx9r-782v details a remote code execution (RCE) vulnerability in Python's pickle module. Attackers can craft malicious pickle files that bypass static analysis tools like Picklescan by leveraging `pip.main()` for installation of a compromised package. This allows arbitrary code execution during deserialization, impacting systems that process untrusted pickle data and enabling supply chain attacks. |
| 2026-04-10 2026 | CVE-2025-56005: PLY RCE Vulnerability news 4 min read | Library vulnerability analysis of CVE-2025-56005 in Dabeaz PLY version 3.11, detailing an insecure deserialization flaw within an undocumented `picklefile` parameter of the `yacc()` function. This allows Remote Code Execution (RCE) through malicious pickle files, a risk amplified by the parameter's obscurity. The analysis includes technical details on the attack vector, root cause (CWE-502), detection methods, and mitigation strategies, while noting ongoing disputes regarding the CVE's validity. → sentinelone.com |
| 2026-04-10 2026 | Multi-Stage Malware Attack on Python Package Index advanced 4 min read | Tool for verifying Python Package Index (PyPI) dependencies, specifically addressing the chimera-sandbox-extensions malware that harvested developer credentials and environment variables. It emphasizes implementing curated package registries, software composition analysis within CI/CD pipelines, lock file usage, and hash-based verification to prevent supply chain attacks. The tool supports techniques like static and dynamic analysis to detect credential harvesting and DGA calls, alongside runtime sandboxing and secret management to mitigate risks from compromised dependencies. |
| 2026-04-10 2026 | CVE-2025-1716 Sonatype Security Advisory news 2 min read | Advisory detailing CVE-2025-1716, an unsafe deserialization vulnerability in Python's `pickle` module, allowing bypass of static analysis tools like `picklescan`. An attacker can craft a malicious model using `pickle` to execute `pip.main()` and install a compromised PyPI package, leading to remote code execution. The vulnerability, CWE-184, stems from `pip` not being treated as an unsafe global by `picklescan` before version 0.0.21. Sonatype recommends upgrading to version 0.0.22 or higher for mitigation. → sonatype.com |
| 2026-04-10 2026 | Picklescan Fails to Detect Unsafe Globals Advisory intermediate 1 min read | Library advisory details a bypass of Picklescan by an unsafe deserialization vulnerability in Python's pickle module. Attackers can leverage `pip.main()` during unpickling to install a malicious package, achieving remote code execution (RCE) via the package's `setup.py` or entry points. This technique allows for silent exploitation and supply chain attacks, as the use of `pip` may not trigger typical security alerts. |
| 2026-04-10 2026 | CVE-2025-1716: Picklescan Analysis Bypass RCE news 1 min read | Writeup of CVE-2025-1716, detailing an unsafe deserialization vulnerability in Python's pickle module. Attackers can bypass static analysis tools like Picklescan by exploiting `pip.main()` during deserialization, leading to the installation of malicious packages and subsequent remote code execution (RCE). The exploit leverages `pip install` to fetch and run arbitrary code from setup.py or post-install hooks, making it a potent supply chain attack vector. |
| 2026-04-10 2026 | CVE-2025-56005: Python PLY Flaw Enables Remote Code Execution news 3 min read | Library for hardening Python applications against the CVE-2025-56005 remote code execution vulnerability in the PLY (Python Lex-Yacc) library. This flaw exploits unsafe pickle deserialization when loading cached parser tables via the undocumented `picklefile` parameter, allowing arbitrary code execution during application startup before traditional security controls are active. The library addresses this by promoting secure deserialization practices, filesystem hardening for parser cache locations, and pipeline protections to prevent artifact poisoning. → esecurityplanet.com |
| 2026-04-10 2026 | CVE Search: Python beginner | CVE Search: Python |
| 2026-04-10 2026 | Python CVE Details beginner | Python CVE Details |
| 2026-04-10 2026 | Python Security Vulnerabilities & Risk Score beginner | Library providing a comprehensive security risk assessment for Python, analyzing 349 vulnerabilities with EPSS scores, exploitation status, and remediation availability. It identifies specific weaknesses like `shutil.unpack_archive()` handling of Windows absolute paths, `BaseCookie.js_output()` character neutralization, and out-of-bounds writes in `asyncio.ProacterEventLoop.sock_recvfrom_into()`. The data includes CVEs such as CVE-2026-5713 for mote debugging, CVE-2026-4786 for command injection via `webbrowser.open()`, CVE-2026-6100 for use-after-free in decompression, and CVE-2026-1502 for HTTP client proxy tunnel validation issues, alongside Pillow's vulnerability to FITS GZIP decompression bombs. |
| 2026-04-10 2026 | Python Security Vulnerabilities in 2026 beginner 19 min read | Survey of Python security vulnerabilities impacting versions up to 3.15.0, detailing CVEs such as CVE-2026-6019 (XSS via Morsel.js_output), CVE-2026-3298 (OOB Buffer Write in ProactorEventLoop), and CVE-2026-5713 (Privileged Memory Access via Profiling/Asyncio Introspection). The analysis also highlights issues like command injection in webbrowser.open (CVE-2026-4786), CRLF injection in http.client, and quadratic complexity DoS vulnerabilities in xml.dom.minidom and HTMLParser. The resource also touches upon resource exhaustion in plistlib and various tarfile module vulnerabilities including filter bypass, arbitrary filesystem writes, and infinite loops. |
| 2026-04-10 2026 | RCE With Modern AI/ML Formats and Python Libraries intermediate 13 min read | Library vulnerabilities in NVIDIA's NeMo, Salesforce's Uni2TS, and Apple/ETH Zurich's FlexTok allow for remote code execution (RCE) when malicious metadata is loaded. These PyTorch-based AI/ML libraries, widely used on HuggingFace, leverage Hydra's `instantiate()` function to load configurations, inadvertently executing arbitrary code embedded in metadata. CVE-2025-23304 (NeMo) and CVE-2026-22584 (Uni2TS) have been assigned, with fixes released by the respective vendors. → unit42.paloaltonetworks.com |
| 2026-04-10 2026 | Critical PickleScan Vulnerabilities Expose AI Model Supply Chains news 1 min read | Writeup of CVE-2025-10155, CVE-2025-10156, and CVE-2025-10157, three critical vulnerabilities in PickleScan. These flaws enable attackers to bypass model scanning safeguards and distribute malicious AI models by exploiting file extension misclassifications, divergent ZIP archive handling between PickleScan and PyTorch, and evasion of dangerous import blacklisting through subclassing. The vulnerabilities, with a CVSS score of 9.3, underscore risks in AI supply chains and highlight the need for layered defenses and safer formats like Safetensors. → infosecurity-magazine.com |
| 2026-04-10 2026 | How a Poisoned Security Scanner Backdoored LiteLLM intermediate 9 min read | Library that suffered a supply chain attack via Trivy and Checkmarx KICS, resulting in malicious versions (1.82.7 and 1.82.8) of the litellm Python package being published to PyPI. The attack involved credential harvesting through a compromised GitHub Action and the use of .pth files for persistence, enabling data exfiltration and lateral movement within Kubernetes environments. → snyk.io |