Selected as a 2025 RSA Security Scholar, representing the intersection of AI security, vulnerability research, and offensive red teaming at one of the world's premier security conferences.
- 🔐 AI/ML Infrastructure Security — agentic workflow exploitation, RAG hardening, prompt injection, MCP attack surface, model file deserialization, GPU memory corruption
- 🛠️ Offensive Security — adversary emulation, red teaming, fuzzing, symbolic execution, n-day exploit development, kernel research
- 🤖 Adversarial ML — jailbreaking, attention-head attacks, safety alignment failures, predictive multiplicity (Rashomon attacks)
- 🧠 Cryptography — LLM-driven CTF generation, JWE/JWT bypasses, authenticated encryption pitfalls
- ☁️ Cloud Security & DevSecOps — application security reviews, threat modeling, secure SDLC
- 📜 Standards & Governance — NIST OLIR, OWASP, ISO/IEC 42001
Original vulnerability research across the AI/ML stack: orchestration, agents, inference engines, tokenizers, model serializers, and security tooling itself.
| CVE | Target | Class | Severity | Notes |
|---|---|---|---|---|
| CVE-2026-33017 | Langflow | Unauthenticated RCE | Critical 9.3 | 🚨 CISA KEV · Exploited in the wild within 20 hours of disclosure · Federal remediation deadline · GHSA-vwmf-pq79-vjvx |
| CVE-2026-32628 | AnythingLLM | SQL Injection (Built-in SQL Agent) | High | GHSA-jwjx-mw2p-5wc7 · CWE-89 in agentic tool-call surface |
Valid, technically inarguable findings where the vendor declined to assign a CVE (often through a retroactive documentation shield) but where the underlying issue is real, reproducible, and publicly disclosed.
| Target | Finding | Severity | Status |
|---|---|---|---|
| XGBoost (×5) | Heap OOB in tree node access · UBJSON parser memory corruption · Parallel tree double-free · Unsafe pickle.loads() on network data (CWE-502) · Hardcoded 0xff99 magic number as Rabit tracker auth (CWE-798) |
Critical / High | Vendor declined ("performance + resourcing"). Full technical writeup published on Medium and LinkedIn. |
Google sentencepiece |
Off-by-4 bounds check in DecodePrecompiledCharsMap enabling OOB read via crafted .model files (UBSan-confirmed) |
Medium | Google VRP Report #498463886 · sentencepiece is the tokenizer backbone of Gemma, T5, and PaLM |
Google sentencepiece |
Unvalidated trie values used as piece array indices (heap OOB read, release builds) | High | Google VRP Report #498465599 · Upstream fix landed via PR #1207 |
| vLLM | LoRA adapter SSRF → RCE chain | High | Closed via documentation shield. Public writeup. |
| Body | Contribution | Detail |
|---|---|---|
| NIST | OLIR Catalog Mapping (Live) | First AI security framework mapping in the NIST Online Informative References catalog: OWASP LLM Top 10 v2.0 → NIST CSF 2.0, 169 relationship entries covering 77 of 106 CSF subcategories. Listed Point of Contact on the NIST CSRC website. |
Full speaker profile: sessionize.com/aviral-srivastava
| Conference | Year | Talk |
|---|---|---|
| RSAC 2026 | 2026 | From Prompt to Pager: Preparing for AI-Native Incidents Now |
| ISACA North America 2026 | 2026 | Breaking the Loop: Offensive Testing of RL and Agentic AI Systems |
| CactusCon 14 | 2025 | Agents Under Siege: Live Attacks from RAG to Tool Calls |
| CypherCon 2025 | 2025 | Deceiving the Deceivers: Offensive Security Strategies for Adversarial AI Attacks |
| BSidesSLC 2025 | 2025 | Filling Gaps in AI Governance: How ISO/IEC 42001 Shapes the Future of AI Risk and Compliance |
| BSidesTC 2025 | 2025 | ROP Alchemy: Universal Gadgets via Type Confusion |
| CactusCon 13 | 2025 | Weaponizing AI: Adversarial Attacks, Hallucinations, and the Offensive Security Frontier |
| HOPE XV | 2024 | Invited Talk (details under NDA) |
| BSidesChicago 2024 | 2024 | Hacking Neural Networks: The Hidden Vulnerabilities of AI Systems |
Coverage of CVE-2026-33017 across major cybersecurity publications. Quoted by name as the discoverer in The Hacker News, Help Net Security, Barrack AI, and Cloud Security Alliance research notes.
- The Hacker News — Critical Langflow Flaw CVE-2026-33017 Triggers Attacks within 20 Hours of Disclosure
- Help Net Security — CISA sounds alarm on Langflow RCE after rapid exploitation
- Infosecurity Magazine — Hackers Exploit Critical Langflow Bug in Just 20 Hours
- Sysdig Threat Research — CVE-2026-33017: How attackers compromised Langflow AI pipelines in 20 hours
- Cloud Security Alliance — Research Notes (×2) on the Langflow RCE
- Qualys ThreatProtect — CISA Added Langflow Vulnerability to KEV Catalog
- Barrack AI — Langflow Got Hacked Twice Through the Same exec() Call
- InfoSec Today — Critical Langflow Flaw Triggers Attacks within 20 Hours
- HackerNoon — CVE-2026-33017: Unauthenticated RCE in Langflow's Public Flow Endpoint Explained
- CiberSafety — CVE-2026-33017 in Langflow: Critical vulnerability for RCE without authentication
- 🎯 ZeroDayForge — Full-spectrum adversary emulation and exploit automation framework
- 🧨 Autonomous n-day Linux Kernel Exploit Pipeline — Mac M1 toolchain (radare2, lldb, Ghidra headless) running RECON → AUTO-PICK → 12-POINT SAFETY GATE → PATCH ANALYSIS → ROOT CAUSE → EXPLOIT STRATEGY → COMPILE VERIFY, producing Exploit-DB-format
exploit.cartifacts - 🛡️ Multi-Agent CVE Hunting Pipeline — Claude Code agent team (Recon · Auditor · Exploiter · Reporter) governed by a 14-rule submission framework and a 4-gate filter (unauthenticated · default config · no doc shield · core feature)
- 🧪 LLM-Driven Cryptographic CTF Generator — Automated cryptographic challenge generation system (MS Thesis, Penn State)
- 🔐 Security Engineer (L4) — Amazon Ads Security, Sunnyvale, CA (2025 – Present) AppSec reviews, threat modeling, penetration testing, AI/ML security, agentic workflows, prompt injection, RAG hardening
- 🧪 Security Internships (6×) — Malware reverse engineering, secure DevOps, GRC
- 🧑🏫 Teaching Assistant — Red Teaming, CTFs, Penn State — Led offensive security labs and workshops
- 🎓 MS in Cybersecurity Analytics & Operations — The Pennsylvania State University GPA: 4.0 · Research Assistant · RSA Security Scholar Thesis: AI-Generated Cryptographic CTF Challenges
- 🎓 BTech in Computer Science — Amity University Focus: Cryptography, Secure Systems, Network Security
- 🥇 RSA Security Scholar 2025
- 🏅 Cybersecurity Innovator of the Year — BSides Bangalore 2023
- 📜 ISSN Best International Research Award
- 🌟 Young Researcher Award
- 🎙️ Speaker: RSAC, ISACA North America, HOPE XV, CypherCon, CactusCon (×2), BSides (×3)
- 🌍 HackTheBox PRO HACKER — Top 200 Global · 🇺🇸 Rank #24
· HTB Profile
💬 "The more they secure, the more we exploit. The future belongs to offensive AI."