Python
Python has become one of the most widely used languages in cybersecurity — from writing exploit scripts and automation tools to building security scanners and processing large datasets. Its readability, extensive standard library, and rich ecosystem of security-focused packages make it the go-to language for security professionals.
In application security, Python appears on both sides: as the language used to build web applications (Django, Flask, FastAPI) and as the primary tool for testing them. Common Python security concerns include unsafe deserialization with pickle, command injection through os.system() and subprocess, SSTI in Jinja2 templates, path traversal in file operations, and SSRF in HTTP libraries like requests.
For offensive security, Python powers many essential tools — from Burp extensions and custom fuzzers to reverse shells and exploit proof-of-concepts. Libraries like pwntools, scapy, and impacket are staples in penetration testing. For defensive security, Python is used to build SIEM integrations, log analyzers, and automated incident response workflows.
This page collects Python security resources covering both secure coding practices for Python applications and Python-based security tooling for offensive and defensive work.
python.org
Python sits on both sides of this library
Two distinct bodies of work end up under this tag, and it helps to know which one you are reading. The first is Python as the language security work is written in: request handling, parsing, concurrency, and the small details that separate a script that works once from a tool you can point at a thousand hosts. The second is Python as an attack surface — the features and libraries that introduce vulnerabilities into applications written in it.
On the second: Python's dangerous edges are mostly about evaluation and
deserialization. pickle executes code by design, which makes any
pickle from an untrusted source equivalent to remote code execution, and it
still turns up in caches, session stores and message queues.
eval and exec need no explanation;
yaml.load without a safe loader is the same bug wearing a
configuration file. Format-string and f-string constructs that touch attacker
input can reach object internals. Subprocess calls with
shell=True are command injection whenever the argument is not
fully controlled. Jinja2 SSTI is common enough that it has its own topic on
this site, and the escape chains there run through exactly these primitives.
Packaging deserves its own attention. Installing a package runs its build code, dependency confusion against internal index names is a live technique rather than a theoretical one, and typosquatted names on PyPI are a recurring event. If you write Python tooling, pinning and hash-checking your own dependencies is security work, not hygiene theatre.
For tool building, the material here leans practical: async and threading
for scanning at scale, working with requests and raw sockets when
you need control over what goes on the wire, parsing hostile input without
crashing, and building Burp extensions or standalone utilities that fit an
existing workflow. If you are new, start with the tooling pieces — writing a
small scanner teaches the protocol details faster than reading about them.
| Date Added | Link | Excerpt |
|---|---|---|
| 2026-11-19 NEW 2026 | Microsoft agent-framework PyPI Package Compromised news 8 min read Supply Chain | Library versions 1.26.1 and 1.26.2 of Microsoft's agent-framework package on PyPI were compromised with credential-stealing malware. The payload, delivered via .pth startup hooks, harvests environment variables, credential files, and scrapes other processes' environments. It also attempts to mint identity tokens from GitHub Actions OIDC, Azure DevOps OIDC, and Azure managed identity endpoints, exfiltrating data to a webhook. The attacker demonstrated rapid re-publication after the initial quarantine. → stepsecurity.io |
| 2026-11-19 NEW 2026 | Microsoft’s agent-framework PyPI package Compromised with Credential Stealer news 3 min read AI Supply Chain | Library that, when compromised in agent-framework versions 1.26.1 and 1.26.2, utilizes Python `.pth` startup hooks to exfiltrate Azure, GitHub Actions, and CI/CD credentials. The malicious code silently harvests environment variables, cloud identity tokens, and configuration files, then transmits them to a webhook.site endpoint. → socket.dev |
| 2026-11-19 NEW 2026 | A Python Startup Hook, a Hijacked Microsoft Package, and 41 Minutes on PyPI intermediate Supply Chain | On November 17, 2026, a malicious actor uploaded a compromised version of a Microsoft open-source Python package to the Python Package Index (PyPI). This compromised package contained a startup hook that, when installed, could execute arbitrary code on users' systems. The malicious package remained on PyPI for 41 minutes before being discovered and removed. This incident highlights the vulnerability of supply chains and the importance of vigilant security monitoring for open-source repositories. |
| 2026-11-18 NEW 2026 | Microsoft agent-framework Compromised: Credential Stealer Targets Azure and CI Pipelines news 4 min read AI Secrets Supply Chain | Library containing a credential stealer targeting Azure and CI pipelines, uploaded to PyPI as malicious versions 1.26.1 and 1.26.2 of Microsoft's agent-framework. The malware, disguised as telemetry, drops .pth files to execute on interpreter start, stealing environment variables, credential files, and Azure/GitHub OIDC tokens, exfiltrating data to webhook[.]site. → wiz.io |
| 2026-11-18 NEW 2026 | InIt CTF: Super Pickle intermediate 3 min read Deser | Library that demonstrates a Python `pickle` deserialization vulnerability exploited in the InIt CTF challenge. The challenge leverages `pickle.loads()` to achieve arbitrary code execution by abusing the `__reduce__` method and executing attacker-controlled functions. The writeup details bypassing keyword filters that blocked common RCE gadgets like `os.system` and `subprocess`, instead using less obvious standard library functions such as `pty.spawn`, `glob.glob`, and `linecache.getlines` to gain file read access and ultimately retrieve the flag from `/proc/self/environ`. |
| 2026-11-18 NEW 2026 | LetMeOut: a three-letter Jinja jail escape advanced 1 min read SSTI | Writeup of LetMeOut, a Jinja2 sandbox jail escape. The challenge restricts templates to ASCII, 170 characters, and three unique letters. Attackers exploit a normalization bypass using a fullwidth underscore (U+FF3F) to access `__globals__` from the `jinja2.sandbox` module and then call `breakpoint()` to achieve shell access, ultimately leading to the flag. |
| 2026-11-18 NEW 2026 | Pytition IDOR intermediate IDOR | Library for Pytition, addressing an Insecure Direct Object Reference (IDOR) vulnerability in versions prior to 2.10. The `show_signatures` view allowed authenticated users to delete signatures from any petition by submitting foreign signature IDs sequentially. This could lead to the permanent deletion of user-owned signatures and trigger confirmation emails to other signers. The vulnerability is identified by CVE-2026-130872. |
| 2026-11-17 NEW 2026 | 5 JWT Authentication Bugs I Keep Seeing in Django REST Framework Projects intermediate AuthN JWT | This content highlights five common JWT authentication bugs found in Django REST Framework projects. Drawing on extensive debugging experience, the author details issues that frequently arise in production environments. The article aims to provide practical, actionable fixes for developers to implement immediately, improving the security and reliability of their authentication systems. |
| 2026-11-17 NEW 2026 | HackTheBox - Nimbus intermediate SSRF Talks | The HackTheBox Nimbus machine involves exploring a system with a health check that reveals an AWS Nimbus job scheduler. The health check initially returns an "InvalidClientToken" error, suggesting a potential avenue for further investigation into the Nimbus scheduler's authentication mechanisms. The content marks the beginning of an nmap scan and then focuses on the initial findings related to the Nimbus job scheduler and its health check behavior. |
| 2026-11-17 NEW 2026 | Introducing Wake 5: agent-written Solidity fuzzing in Python advanced 3 min read AI Fuzzing | Framework for fuzzing Solidity smart contracts, Wake 5 integrates an EVM directly into the Python process, enabling faster testing and cross-chain capabilities. It introduces an MCP server to provide agents with detailed static analysis data, allowing them to generate fuzz test harnesses from prompts. Agent-written tests demonstrated superior efficiency compared to manual methods and tools like Echidna and Foundry on the Daedaluzz benchmark, reaching all assertions significantly faster. Generated Python types enhance developer experience with autocompletion and type checking. |
| 2026-11-17 NEW 2026 | CTF@CIT 2026 Temporary Destruction: Jinja SSTI blacklist bypass intermediate 1 min read RCE SSTI | Writeup detailing a Jinja Server-Side Template Injection (SSTI) vulnerability in the Temporary Destruction CTF challenge. The application attempts to filter dunder methods using a regex, but this blacklist is bypassed by escaping underscores. The exploit leverages this by using `cycler.__init__.__globals__` with escaped dunders to access `os.popen` and read the flag file. |
| 2026-11-17 NEW 2026 | django-tenants SQL injection in PostgreSQL backend news SQLi | Writeup of GHSA-pg64-4qv2-75g5, an SQL injection vulnerability in `django-tenants` (versions 3.4.8 before 4.0.1) affecting its PostgreSQL backend. Attackers can exploit unescaped tenant schema names to inject SQL, potentially leading to unauthorized data access or execution of SQL commands as the application's database user. This affects statements like SET search_path and schema creation/modification. |
| 2026-11-17 NEW 2026 | Frappe Framework SSRF news SSRF | Advisory for a Server-Side Request Forgery (SSRF) vulnerability in Frappe Framework versions before 15.123.0 and 16.x before 16.50.0. The vulnerability exists in the `frappe.email.email_body.get_email_html` method, allowing authenticated users to inject stylesheet link tags into unescaped email templates. This triggers Premailer to fetch content from internal or external hosts, as identified in CVE-2026-130069. |
| 2026-11-16 NEW 2026 | From a broken error message to full server RCE: Jinja2 SSTI writeup ($9,000) intermediate Bug Bounty RCE SSTI | A security researcher discovered a Server-Side Template Injection (SSTI) vulnerability in Jinja2, leading to full Remote Code Execution (RCE). The vulnerability was initially hinted at by a "broken error message" that exposed internal details. This exploit allowed the researcher to gain complete control over the server. The company awarded a $9,000 bounty for this critical finding. |
| 2026-11-16 NEW 2026 | Authlib accepts OpenSSH keys/certs as HMAC secrets news AuthN JWT | Library update addresses vulnerabilities in `authlib.jose` where OpenSSH keys and certificates were incorrectly accepted as HMAC secrets, allowing for the forging of HS256 tokens (CVE-2024-37568, CVE-2026-114615). Applications using `jwt.decode` or `JsonWebSignature` without algorithm restrictions are vulnerable if an attacker knows the public key. The fix in version 1.9.5 prevents OctKey from loading OpenSSH public keys or certificates as HMAC secrets. Workarounds include algorithm restriction or using typed keys. |
| 2026-11-16 NEW 2026 | eMudhra / emSign CA Compromise: What Python, pip, and CI/CD Users Need to Update news 7 min read Supply Chain | Library updates are critical following the eMudhra/emSign CA compromise, which involved eight intermediate CA certificates with private keys generated using insufficient randomness. This vulnerability allowed for the creation of impersonating certificates, posing a risk to dependency downloads and CI/CD pipelines. Users must update Python packaging tools, including `certifi` to version 2026.10.19 or later, and `pip` to 26.2.2, ensuring custom certificate bundles and distributor backports are also addressed. Verification of local certificate trust configurations is essential, as updating a single package may not resolve trust issues across all system and application-level certificate stores. → stepsecurity.io |
| 2026-11-16 NEW 2026 | litellm 1.107.0 Compromised on PyPI and Docker: Credential Stealer and Hourly Remote Code Loader news 22 min read Secrets Supply Chain | Library for Python developers detailing the compromise of litellm version 1.107.0 on PyPI and Docker. The malicious release contained a credential stealer named `privcollect` and a remote code execution payload. The attack involved a lookalike GitHub repository, a stolen GitHub organization owner's token, and manipulation of the build process via `maturin`'s `data` option. Clean releases have since resumed. → stepsecurity.io |
| 2026-11-16 NEW 2026 | fast-jwt empty-key unsigned-accept model (CVE-2026-107720) news 3 min read JWT | Writeup of CVE-2026-107720 in fast-jwt, providing an offline, deterministic Python model of its createVerifier empty-key unsigned-accept vulnerability. The model re-implements the pre- and post-fix verifier gating logic, demonstrating the bug path and fix without executing fast-jwt or performing cryptographic verification. It includes 18 test cases and independently derives expected outcomes, producing CSV output and SHA256 checksums. |
| 2026-11-15 NEW 2026 | SUSE-SU-2026:5064-1: PyJWT HMAC forgery and JWKS key-substitution fixes news 4 min read AuthN JWT | Update for python-PyJWT addresses eight vulnerabilities, including CVE-2026-101917 concerning JWKS endpoint amplification and CVE-2026-102267 for forwarded credential disclosure or key substitution. It also fixes HMAC forgery issues identified in CVE-2026-102268 and CVE-2026-102271, raw-token revocation bypass in CVE-2026-102269, denial-of-service in CVE-2026-102270, and stolen token misuse in CVE-2026-102275. |
| 2026-11-15 NEW 2026 | livesub: subfinder + httpx wrapper for live-host recon reports intermediate 2 min read Recon | Tool wrapping ProjectDiscovery's subfinder and httpx, livesub automates live-host reconnaissance for authorized pentesting and bug bounty engagements. It generates grouped text and JSON reports, categorizing live subdomains by HTTP status codes, titles, and detected technologies. Installation is straightforward via pip or direct source execution, with no Python runtime dependencies and clear instructions for specifying the httpx binary. |
| 2026-11-15 NEW 2026 | Frappe india-payroll missing authorization news AuthZ | Writeup of GHSA-xfwp-rff6-r369 impacting Frappe india-payroll before v16.0.4, detailing a missing authorization vulnerability within `tax_regime_selector.py`. This flaw allows authenticated, low-privileged users to view other employees' payroll data, including base pay and earning/deduction components, by supplying alternative employee identifiers. Additionally, attackers can alter tax regimes on draft Salary Structure Assignments, impacting tax withholding. The vulnerability is linked to CVE-2026-129429. |
| 2026-11-14 NEW 2026 | ZDI-26-951: FAIRChem deserialization of untrusted data RCE news Deser RCE | Writeup of ZDI-26-951, a deserialization of untrusted data remote code execution vulnerability in Meta FAIRChem. This flaw, exploitable via user interaction with malicious files or pages, allows attackers to execute arbitrary code in the context of the current process due to insufficient validation during checkpoint file parsing. Mitigation involves restricting interaction with the affected product. → zerodayinitiative.com |
| 2026-11-14 NEW 2026 | BBOT 3.1.0: faster DNS brute-forcing and new passive subdomain sources intermediate 5 min read OSINT Recon | Library release BBOT 3.1.0 introduces faster DNS brute-forcing via blastdns 2.0, new passive subdomain sources like otilabs and scanmalware, and WAF bypass detection with the nowafpls module. Enhancements include improved custom module loading, hardened configuration validation, and optimized DNS concurrency. Notable module updates involve telerik, nuclei, and wayback, alongside new favicon and js_unpacker capabilities. This release also removes massdns and DeepDiff dependencies, updating others to improve performance and security. |
| 2026-11-14 NEW 2026 | ZDI-26-908: Meta Detectron2 deserialization of untrusted data RCE news Deser RCE | Advisory ZDI-26-908 details a deserialization of untrusted data remote code execution vulnerability in Meta Detectron2. This flaw arises from improper validation during pickle file parsing, allowing attackers to execute arbitrary code within the affected process by tricking users into opening malicious files or visiting compromised pages. Mitigation involves restricting user interaction with the product. → zerodayinitiative.com |
| 2026-11-14 NEW 2026 | ZDI-26-940: fairseq deserialization of untrusted data RCE news 1 min read AI Deser RCE | Writeup of ZDI-26-940, a deserialization of untrusted data RCE vulnerability in Meta fairseq's ApplyKmeans class within the dump_km_label module. This flaw, reported by Mat Powell of TrendAI Zero Day Initiative, allows remote attackers to execute arbitrary code by tricking users into visiting a malicious page or opening a file. As the fairseq repository is no longer maintained, the only mitigation is to restrict interaction with the product. → zerodayinitiative.com |
| 2026-11-14 NEW 2026 | joserfc: list-valued jti accepted by exact-value JWT claim validation intermediate 2 min read JWT | Writeup on joserfc GHSA-mcm9-p458-pjmc reveals a flaw where JWTClaimsRegistry accepted list-valued `jti` claims when an exact string match was configured. This vulnerability, present in versions before 1.7.6, allowed an array containing the expected `jti` value to bypass exact-value validation, treating it as a list-membership check. The issue was fixed in joserfc 1.7.6, reinforcing RFC 7519's requirement for `jti` to be a string. |
| 2026-11-14 NEW 2026 | Keep 0.34.10–0.55.0 blind SQL injection in CEL-to-SQL intermediate SQLi | Advisory for GHSA-vq4h-g7mx-jcq2 details a blind SQL injection vulnerability in Keep versions 0.34.10–0.55.0. This vulnerability exists within the CEL-to-SQL converter, where unescaped bracket-indexed JSON keys from CEL filters are inserted into JSON path expressions for SQLite, PostgreSQL, and MySQL. Authenticated users with read:alert permission can exploit this by submitting crafted CEL expressions to specific API endpoints, using result counts as a boolean oracle to extract sensitive data like other tenants' alerts and password hashes. The CVE ID CVE-2026-129166 is associated with this issue. |
| 2026-11-14 NEW 2026 | Schemathesis 4.33.0: GraphQL coverage phase and unary gRPC fuzzing intermediate 4 min read API Sec Fuzzing GraphQL | Library updates Schemathesis to version 4.33.0, introducing a coverage phase for GraphQL schemas and fuzzing for unary gRPC methods. Enhancements include improved XML and cookie parameter handling, faster pytest collection, and numerous bug fixes across data generation, request serialization, GraphQL processing, schema loading, checks, authentication, timeouts, test execution, and reporting. Notably, it addresses issues with response schema conformance, OpenAPI 3.2 specifications, and handling of $ref and XML namespaces. |
| 2026-11-14 NEW 2026 | Arize Phoenix: format string injection in applyChatTemplate GraphQL query leaks secrets / JWT signing key news 2 min read AI GraphQL JWT Secrets | Library advisory details a format string injection in Arize Phoenix's `applyChatTemplate` GraphQL endpoint. Exploitation, even without authentication, allows attackers to read server environment variables, including the JWT signing key (`PHOENIX_SECRET`), enabling session token forgery and decryption of stored secrets. Affected versions range from v13.0.0 to v14.8.0. |
| 2026-11-14 NEW 2026 | DjangoBlog stored XSS news XSS | Writeup of stored XSS in DjangoBlog, affecting versions 2.0.0.0 through 2.1.0.0. The vulnerability stems from the `class_filter` comment sanitizer, which fails to properly sanitize event-handler attributes on specific HTML tags like `span`, `div`, `pre`, and `code`. This allows authenticated commenters to inject malicious JavaScript, such as `autofocus` and `onfocus` attributes within a `div` tag, which executes for all visitors without user interaction. The issue is tracked as CVE-2026-129155. |
| 2026-11-13 2026 | social-auth-core login CSRF intermediate AuthN CSRF | Library for social-auth-core, addressing a login cross-site request forgery vulnerability in versions prior to 7.0.0. This SAML backend flaw allows attackers to craft malicious IdP responses and RelayState parameters, enabling session hijacking and data capture. The vulnerability, referenced as GHSA-33vc-vcr3-vj5p and CVE-2026-126775, is patched in version 7.0.0. |
| 2026-11-13 2026 | PyJWT HMAC forgery with known public key (CVE-2026-102268) security update news 2 min read JWT | Update for python-PyJWT addresses CVE-2026-102268, an HMAC forgery vulnerability allowing attackers with the public key to forge tokens, and CVE-2026-102270, a denial-of-service flaw causing CPU consumption. This security update is available for various SUSE Linux Enterprise products. |
| 2026-11-13 2026 | DataGerry 3.3.0 SSTI via DocAPI template rendering (CVE-2026-57853) news RCE SSTI | Advisory for CVE-2026-57853 details a Server-Side Template Injection (SSTI) vulnerability in DataGerry versions prior to 3.3.0. This flaw allows authenticated attackers with base.docapi.template.add permissions to inject arbitrary Jinja2 payloads through POST requests to `/rest/docapi/template`. The vulnerability is triggered when templates are rendered via GET requests to `/rest/docapi/template/<id>/render/<object_id>`. Attackers can leverage the `cycler.__init__.__globals__.os` gadget chain to achieve remote code execution by executing operating system commands as the DataGerry web process user, potentially root in default Docker setups. → vulncheck.com |
| 2026-11-13 2026 | datamodel-code-generator before 0.84.0 code injection news RCE | Advisory for `datamodel-code-generator` before version 0.84.0 detailing a code injection vulnerability. Attackers can achieve arbitrary code execution by submitting malicious JSON Schemas with crafted `x-python-type` annotations containing `ast.Constant` nodes within subscript slices, bypassing validation. This allows embedding Python expressions, leading to execution via Pydantic v2's `model_rebuild()` when generated models are imported at runtime. CVE-2026-127208 is associated with this vulnerability. |
| 2026-11-12 2026 | Pickle-Me-Not: CTF pickle deserialization writeup intermediate 2 min read Deser | Writeup of the "Pickle-Me-Not" CTF challenge from Qiangwang NEST 2026, detailing a "malicious model" that requires static analysis rather than execution. The solve involves bypassing a gateway's path-policy with double URL-encoding, reading diagnostic logs to reconstruct a SHA-256 keystream using mhkdf-sha256-v1 parameters, and then XORing this keystream with interleaved ciphertext bytes from a strided tensor view. The key takeaway emphasizes that executing untrusted code, like a pickle, can be the intended trap, and understanding tensor view layouts is crucial for data recovery. |
| 2026-11-12 2026 | JWT verification bugs: pin the algorithm, then prove the fix intermediate 9 min read AuthN JWT | Guide for developers on fixing common JWT verification bugs, including allowing the token to dictate the algorithm or skipping signature checks. It details how to configure PyJWT, jsonwebtoken, and jose to enforce security rules, emphasizes the importance of a test suite to prove verifier effectiveness, and covers essential checks like signature validity, algorithm pinning, claim validation (exp, iss, aud), and secure key management to prevent vulnerabilities like those described in PortSwigger's JWT research. |
| 2026-11-12 2026 | Opik SDKs render Jinja2/Nunjucks prompt templates without a sandbox intermediate 1 min read AI RCE SSTI | Library advisory for Opik SDKs addresses CVE-2023-XXXX, detailing how versions prior to 2.2.123 render Jinja2/Nunjucks prompt templates without a sandbox, enabling arbitrary code execution. This vulnerability impacts applications formatting user-controlled Jinja2 prompts, with potential for access to environment variables, credentials, and network resources. Patches are available in version 2.2.123, which implements sandboxing for template rendering, or users can adopt Mustache prompts or restrict prompt editing. |
| 2026-11-12 2026 | Python Just Got a New Kind of String. It Could End a Whole Class of Security Bugs. advanced | Python has introduced a new type of string, "Stringify," designed to prevent a common class of security vulnerabilities. This innovation aims to mitigate risks associated with improper handling of strings, particularly in web development and data processing. The new string type offers enhanced security features, making it more difficult for attackers to exploit vulnerabilities that arise from string manipulation errors. This development represents a significant step towards more secure Python applications. → levelup.gitconnected.com |
| 2026-11-12 2026 | Dell OpenManage Ansible Modules XXE intermediate XXE | Dell OpenManage Ansible Modules versions 2.0.4 through 2.1.2 contain a critical XML External Entity (XXE) vulnerability. This flaw allows an attacker to inject malicious XML, potentially leading to sensitive data disclosure, denial of service, or even remote code execution on the affected systems. Users are strongly advised to update to the patched versions, 2.0.5 and 2.1.3 respectively, to mitigate this significant security risk. |
| 2026-11-11 2026 | CVE-2026-123198: unauthenticated Jinja SSTI in LMFDB render_knowl intermediate RCE SSTI | LMFDB, prior to commit 01829bd, is vulnerable to CVE-2026-123198, an unauthenticated server-side template injection (SSTI) flaw within the `render_knowl` function. Attackers can exploit this by crafting malicious query-string values, potentially leading to unauthorized access or code execution. → insightcve.com |
| 2026-11-11 2026 | Malicious PyPI dynamic-mcp, reqnest and friends: reverse shells, OAuth relays, Telegram stealers intermediate 14 min read Supply Chain | Library for identifying malicious Python packages like `dynamic-mcp`, `labelearthengine-api`, `qzc690382`, and `reqnest`. These packages exploit various vulnerabilities including CWE-506 (Introduction of Malicious Code), CWE-306 (Authentication Bypass), CWE-829 (Inclusion of Functionality from External Source), CWE-912 (Non-Tamper-Proof Update of Operating System Software), and CWE-200 (Exposure of Sensitive Information). The analysis details how `dynamic-mcp` establishes reverse shells, `labelearthengine-api` relays OAuth tokens, and other packages exfiltrate data or execute commands. |
| 2026-11-11 2026 | Wake 5: a Python framework for Solidity testing, fuzzing and static analysis intermediate 2 min read Fuzzing | Framework for Solidity testing, fuzzing, and static analysis, Wake 5, offers an in-process EVM and Model Context Protocol server for language-model agents. It achieves high transaction throughput, comparable to Foundry and significantly faster than Echidna. Wake 5's analysis demonstrates that integer generator priors, not throughput, bound stateful smart-contract fuzzing once shallow assertions are exhausted. Harnesses leveraging constraint solvers and language models achieved full assertion coverage efficiently. → arxiv.org |
| 2026-11-11 2026 | Intel BigDL before 2.5.0b1 deserialization of untrusted data news AI Deser | Library for Intel BigDL, versions prior to 2.5.0b1, vulnerable to deserialization of untrusted data. This vulnerability, found within Ring 3: User Applications, could allow an unprivileged, unauthenticated attacker with low complexity to escalate privileges via network access, impacting confidentiality, integrity, and availability to a low degree. |
| 2026-11-11 2026 | LetMeOut: HICyberCon CTF 2026 Jinja2 sandbox escape writeup intermediate 2 min read SSTI | Writeup detailing a Jinja2 sandbox escape vulnerability in the LetMeOut CTF challenge. The exploit leverages NFKC normalization to smuggle attribute names past the `is_safe_attribute` guard, specifically targeting the `__class__` attribute. A secondary vulnerability in `getitem` allows for the execution of unsandboxed code, leading to access to `pdb` and subsequently reading flag files by abusing octal escapes for payload construction. |
| 2026-11-10 2026 | PyPI makes release attestations easier to find with new security tab news 2 min read Supply Chain | Library adds a security tab to PyPI project pages, making release attestations easier to find. This tab displays whether release files have digital attestations, such as PyPI Publish or SLSA Build Provenance, which verify their origin and build process. Users can review attestation details and linked transparency logs, or identify warnings for incomplete provenance, differing repositories, or workflow changes that may indicate potential compromise. → helpnetsecurity.com |
| 2026-11-10 2026 | Beyond single-command recon: async multi-stage pipelines intermediate 2 min read Recon | Pipeline for automated multi-stage reconnaissance, transforming target domains into structured, triage-ready telemetry. This approach chains an async crawler like `katana` to discover endpoints and JS bundles, followed by parameter discovery using `arjun`. A context and reflection engine powered by `dalfox` or `XSStrike-go` then probes reflections and parses DOM sinks. The output is a machine-readable `recon_report.json` adhering to a Pydantic schema, enabling methodology over luck in bug finding. |
| 2026-11-10 2026 | SUBX: subdomain recon and asset management framework intermediate OSINT Recon | SUBX is a framework designed for subdomain reconnaissance and asset management. It aims to streamline the process of identifying and organizing subdomains associated with an organization's assets. This tool likely assists security professionals and researchers in mapping out an organization's digital footprint, which is crucial for vulnerability assessment and overall security posture. |
| 2026-11-10 2026 | ZDI-26-850: Microsoft Qlib MLflowRecorder deserialization RCE news 9 min read AI Deser RCE | Reference details ZDI-26-850, a high-severity deserialization vulnerability in Microsoft Qlib's MLflowRecorder. This flaw allows remote code execution via user interaction, enabling attackers to leverage unsafe Python object reconstruction with tools like `pickle` to compromise data science workstations and research servers running quantitative analysis or ML experimentation. The entry provides Sigma rules for detecting Python processes spawning shells and establishing unexpected outbound connections, crucial hunt logic for defenders. |
| 2026-11-10 2026 | Public PoC for Hugging Face Datasets pickle deserialization RCE (CVE-2026-57845) news 3 min read AI Deser RCE | Writeup detailing CVE-2026-57845, a critical RCE vulnerability in Hugging Face datasets affecting versions up to 5.1.0. The flaw, discovered by Sai Teja Erukude, allows arbitrary Python code execution via crafted pickle files loaded through a deprecated pandas builder. This bypasses security checks, enabling attackers to steal credentials and sensitive data, especially in machine learning pipelines handling untrusted datasets. No patch is available, and mitigation involves avoiding the pandas builder. |
| 2026-11-10 2026 | From SSTI to a Django secret key (XBEN-056) intermediate 4 min read Secrets SSTI | Writeup detailing XBEN-056, a vulnerability discovered during a three-step registration flow where a user-supplied name field was compiled as template source. This allowed for template injection, bypassing a block on percent signs and utilizing double-brace expressions to access the application's signing key material from Django's SECRET_KEY. The exploit successfully retrieved the secret value, demonstrating a critical path from user input to exposed signing credentials. |
Frequently Asked Questions
- What are common Python security vulnerabilities?
- Common Python security issues include unsafe deserialization with pickle, command injection through os.system() and subprocess with shell=True, Server-Side Template Injection in Jinja2, path traversal in file operations, SSRF in the requests library, and SQL injection when using string formatting instead of parameterized queries.
- Why is Python popular in cybersecurity?
- Python's readability, extensive standard library, and rich ecosystem of security packages make it ideal for exploit development, automation, and tool building. Libraries like pwntools, scapy, impacket, and requests are widely used in penetration testing. Python is also the primary language for Burp Suite extensions (via Jython) and many security scanners.
- How do you write secure Python code?
- Use parameterized queries for database access, avoid pickle for untrusted data (use JSON instead), never use eval() or exec() on user input, use subprocess with shell=False and explicit argument lists, validate and sanitize file paths to prevent traversal, and keep dependencies updated to patch known vulnerabilities.
Weekly AppSec Digest
Get new resources delivered every Monday.