Tommaso Bona

IT & Security Consultant | AI Security Researcher | ADHD & Autism Advocate

Welcome

As a passionate cybersecurity professional and researcher, I currently serve as an IT & Security Consultant at Tralerighe Service and an Independent AI Security Researcher, specializing in LLMs and AI Model vulnerability discovery. My career is built on a foundation of technical leadership and innovation; as the former Co-Founder and CTO of Voyxa, I led a cross-functional team of 20+ PhD students, consultants and engineers to develop our AI-driven vocal Helpdesk solution. Under my technical steering, Voyxa was recognized as a finalist at Digithon (Italy's largest pitch event) and, among other things, i also served as the technical spokesperson, bringing Voyxa from 0 to a 500k$ evaluation, in only 10 months.

My transition from a SOC Analyst and Cybersecurity Specialist, to a CTO, and now a specialized Consultant, has allowed me to blend deep technical expertise with strategic business development. I also worked as a Business Development Advisor and Marketing Strategist for StationX, where I proposed and helped guide the creation of their new white papers platform, all of which started, with me working remotely for StationX as a "simple" Cybersecurity Writer, for a bit more than a year. Beyond the corporate world, I am a volunteer OSINT Analyst for the Guardian Group's PURSUIT® Team, helping fight child exploitation, together with law enforcement, as well as i've had the privilege of contributing to the Python Software Foundation in multiple ways, and i'm a proud advocate for ADHD and Autism awareness, considering neurodiversity as a vital asset for innovation, in the current cybersecurity landscape.

🛠️ My Expertise

Core Skills

AI Security: LLM Red Teaming, Adversarial Machine Learning, and AI Safety Assessments.
OSS Security: Independent Python-focused Security Research on Open-Source Software.
Offensive Security: Web Penetration Testing, Vulnerability Research, and OSINT.
Leadership & Strategy: Technical Team Management, Product Management, and Business Development.
Programming & Scripting: Python, Rust, PowerShell, Bash, and Batchfile (even if deprecated).

Publications & Research

Whitepaper: Anchored Fictional Multilingual Injection (AFMI): A Deterministic Prompt-Based Safety Bypass in Web-Integrated LLM Search-Engine Pipelines (StationX Ltd. - 2025)
Technical Writeup: PySpector: how my own SAST Framework, led me to an RCE in a Major AI Library (SecurityCert s.r.l. - 2025), CVE-2026-31900: How a Permissive Regex Led to RCE in psf/black’s GitHub Action (SecurityCert s.r.l. - 2026)
Academic Paper: Fragmentation, Encryption, and Redundancy in Data Transmission: A Novel Secure UDP-Based Fragmentation Protocol and Its Applications (CERN & OpenAIRE - 2024)
General Cybersecurity Articles (English): SQLmap Tutorial for Beginners 2024 (StationX Ltd. - 2024), XSS vs CSRF: The Difference Between These Popular Attacks (StationX Ltd. - 2024), Best OS for Hacking (StationX Ltd. - 2024), Top 15+ Cyber Security Tools (StationX Ltd. - 2024), Hacking With Python: Unlocking Python's Potential (StationX Ltd. - 2023), SQL Injection Testing: All You Need to Know! (StationX Ltd. - 2023)
General Cybersecurity Articles (Italian): AI Offensive Security per Principianti (EP02) (SecurityCert s.r.l. - 2025), AI Offensive Security per Principianti (EP01) (SecurityCert s.r.l. - 2025)

🛡️ Core Security Contributions

Title	Description	CVE ID	Links
NoSQL injection in Openwhyd `controller(): request.body` leads to Private Information Disclosure	Discovered and responsibly disclosed a Medium severity NoSQL Injection in Openwhyd, leading to Info Disclosure	CVE-2026-123617	CVE.org record GitHub Advisory
Unauthenticated Server-Side Template Injection leading to RCE in LMFDB	Discovered and responsibly disclosed a Critical (CVSS4: 10.0) unauthenticated SSTI (CWE-94, CWE-1336) in the L-functions and Modular Forms Database (LMFDB). The `/knowledge/render/<ID>` endpoint spliced request data, including every query-string key and value, directly into a Jinja template rendered in an unsandboxed environment. A single unauthenticated GET request was enough to achieve OS-level Remote Code Execution, confirmed against the live lmfdb.org production instance with a non-destructive `id` call. Patched in commit `01829bd`.	CVE-2026-123198	CVE.org record GitHub Advisory
Disabled TLS Certificate Verification in OpenAlgo	Found by my `pyspector-monitor` pipeline at SatoriCI. OpenAlgo's SMTP code (versions >= 1.0.0.29, < 2.0.3.1) disabled both hostname and certificate verification (CWE-295), allowing Man-in-the-Middle attacks to capture SMTP credentials and intercept password reset links (account takeover primitive). Medium severity (CVSS4: 6.0). Fixed in v2.0.3.1.	CVE-2026-120250	Read the full writeup CVE.org record GitHub Advisory
Argument Injection in Red Hat's Ansible-core leading to RCE	Found and responsibly disclosed a High severity (CVSS: 7.8) Argument Injection in Red Hat's Ansible-core, leading to Remote Code Execution.	CVE-2026-11332	Read the full writeup NIST NVD Red Hat Advisory
0-click RCE in psf/black's GitHub Action	Discovered and responsibly disclosed a High severity (CVSS4: 8.7) RCE in the official psf/black GitHub Action (2.2B downloads). A permissive regex in `action/main.py` allowed direct URL references to bypass version validation, enabling arbitrary code execution on CI runners via a malicious `pyproject.toml`, triggered by the creation of a new pull request with no maintainer interaction required. Patched in v26.3.0.	CVE-2026-31900	Read the full writeup NIST NVD GitHub Advisory
Remote Code Execution (RCE) in AI Framework	Identified and responsibly disclosed a critical RCE vulnerability in a widely used AI library, by leveraging my own SAST framework, PySpector.	N/A	Research & Proof of Concept
Public Acknowledgement from the U.S. Department of Commerce	Discovered and reported a Critical, High-Impact Error-based SQLi, leading to a full database takeover on a US Department of Commerce subdomain.	N/A	Official acknowledgement
Requests (Python Software Foundation)	Reported what was initially thought to be a minor security flaw in the Requests library, which was ultimately classified as a bug and fixed upstream.	N/A	PR #7308
CPython (Python Programming Language)	Contributed to the CPython core test suite by addressing a Path Traversal vulnerability (minor impact, since it only affected the dev test suite).	N/A	PR #138158
Lead Dev and Maintainer of PySpector	State-of-the-art hybrid Rust/Python graph-based SAST framework, made for CI/CD and large Python codebases. Engineered for high-speed, high-accuracy vulnerability scanning, performing 2.1x faster than Bandit and 3.7x faster than Semgrep.	N/A	GitHub: PySpector

Minor Security Contributions

This list shows the vulnerabilities i found in really small projects, that got fixed and published, though are too small to request a CVE ID for:

📜 Relevant Certifications

CompTIA: SecAI+ (in progress).
ISC2: Certified in Cybersecurity (CC), NIST Cybersecurity Framework 2.0.
EC-Council: E|HE, N|DE, D|FE (former Cybersecurity Essentials Series).
Security Blue Team: Blue Team Junior Analyst (BTJA).
SkillFront: ISO/IEC 27001:2022 Associate.
Other non-vendor-neutral certifications: ESET Managed Client Security Specialist, ESET Managed Cloud Security Specialist, Fortinet Certified Associate in Cybersecurity, IBM Roles Processes and Operating System's Security.