Active DIAT · DRDO · 2026 Pune, India

Samratth Singh.

 

I study how intelligent systems are made to betray their purpose — across language models, multi-agent coordination, and behavioral biometrics. The work hunts for the structural condition that makes a whole class of failure inevitable.

2
Publications
12
Attack categories
Top 5%
HackTheBox
14th
Nat'l CTF
4+
Pinned repos
Adversarial AI/LLM red teaming/DIAT · DRDO/Multi-agent RL/IEEE Q1 · 2026/Behavioral biometrics/14th national CTF/Drone swarm robustness/ Adversarial AI/LLM red teaming/DIAT · DRDO/Multi-agent RL/IEEE Q1 · 2026/Behavioral biometrics/14th national CTF/Drone swarm robustness/
01 — Operator profile

A researcher who reads systems for the seam.

  ▄▄▄▄▄▄▄  ▄▄▄▄▄▄▄
  █ ▄▄▄ █  █▄ ▄▄▀█
  █ ███ █  ▀▄▀▄▀ █
  █▄▄▄▄▄█  █▀█ ▀▄█
  ▄▄▄ ▄ ▄▄▄ ▄▀ ▄▄▄
  ▀█▄▄  █▀▄▀▄▀  ▄▀
  █▀▄▄▄▄█▀▀  ▄ ▄▀█
  ▄▄▄▄▄▄▄ █▄▀█ ▀▀▄
  █ ▄▄▄ █ ▀▀▀▄  ▀█
  █ ███ █ ▀█▄ ▄▀▄▀
  █▄▄▄▄▄█ ▄ ▄  ▀█▄
Samratth Singh
// Adversarial Systems Researcher
Status◉ Active
AffiliationDIAT · DRDO-affiliated
ProgrammeB.Tech CSE (Cybersecurity)
YearPre-final · 2023–2027
HonorsICEM 2024 · NCTAAI 4.0
TargetIEEE Q1 · 2026
BasedMaharashtra, IN

I work at the intersection of AI security, multi-agent systems, and applied cryptography — three domains sharing one structural question: what makes an entire category of system exploitable, not just a single weak instance?

That question produced a taxonomy of LLM adversarial surfaces (Best Paper, ICEM 2024), an ongoing MARL study of emergent coordination failure in drone swarms at DIAT (Defence Institute of Advanced Technology, a DRDO-affiliated deemed university), and a behavioral-biometric adversarial primitive (MIMIC, targeting IEEE TIFS). CTF results are execution evidence — not the headline.

// Technical stack
Adversarial AI & ML
LLM red teaming
92%
Adversarial ML
88%
Multi-agent RL
85%
Core languages
Python
95%
PowerShell
78%
MySQL / SQL
68%
Offensive tooling
Web exploitation
84%
Reverse engineering
76%
Cryptography
80%
NmapBurp SuiteMetasploit SQLmapHashcatHydra FFUFGobusterOWASP ZAP BeEFMedusaScapy PyTorchNetworkXScapy LaTeX
02 — Research

Published & in motion.

Published / verified
In progress / submission
Published · ICEM 2024 ★ Best Paper ★ Best Presentation

Prompt Injection & Jailbreaking — a comprehensive taxonomy of adversarial attacks on LLMs.

NCTAAI 4.0 · 2024

A structural taxonomy of LLM adversarial attack surfaces — twelve categories across injection, jailbreaking, evasion, and extraction — designed as a map of failure conditions rather than a single exploit chain.

direct injectionindirect injectionjailbreakingmulti-turn evasiontoken smugglingmodel inversionalignment bypasscontext manipulation
In progress · IEEE Q1

Multi-Agent RL for drone swarm jamming avoidance.

DIAT (DRDO-affiliated) · target: IEEE TNNLS

A QMIX-based multi-agent architecture for coordinated jammer avoidance in drone swarms. Phase D results: +13.0 detection points under chase-jammer via KL-anchored Navigator addressing emergent policy decoupling. Pre-publication.

QMIXcooperative MARLjammer avoidancepolicy decouplingKL anchoringdrone swarms
Redirected · IEEE TIFS

MIMIC — hybrid LSTM/Diffusion synthesis of human mouse motion.

Originally: IEEE RA-L (desk-rejected, scope mismatch) · Redirecting: IEEE TIFS

An LSTM/DDPM hybrid synthesizing realistic human mouse-movement trajectories on a 212K+ sample dataset — full bot-detection evasion demonstrated against commercial detection systems.

behavioral biometricsLSTM / DDPM212K+ samplesbot-detection evasionJSD · ADE/MSE
// the unifying question

How is an intelligent system made to betray its purpose — and what is the structural condition for that betrayal across language models, coordinating agents, and cryptographic trust protocols?

03 — Speculative frontier

Held in containment.

Genuine intellectual interest with no published output yet — bounded deliberately so it's never confused with verified research above.

reading interest
QKD trust-model critique

QBER-threshold models in BB84 protocols may be structurally blind to source-layer compromise.

BB84/E91source compromisePQC
reading interest
Hypergraph threat recon

Modelling attack surface as a hypergraph; quantum-walk sampling as traversal for non-obvious pivots.

hypergraphsquantum walkspivot discovery
reading interest
Quantum-adversarial ML

Whether quantum feature maps in variational classifiers offer genuine robustness — or relocate the attack surface.

VQCQML evasionNISQ
04 — Systems built

Things that run.

30+
Vuln rules
AST
Analysis
MIT
License
Python
Core
◉ Active
RAV3N-SEC
// AI-ready Python vulnerability scanner

Fast, local static analyzer combining Regex and AST to detect 30+ Python security vulnerabilities. Severity-graded findings across code execution, command injection, hardcoded secrets, weak crypto, and unsafe deserialization. Rich CLI with syntax-highlighted reports.

// Architecture pipeline
Source .py files AST Parse + Regex scan Rule Engine 30+ patterns Severity grader Rich CLI report
PythonASTTyperRichsecrets detectionSQL injection
github.com/AnonymousSingh-007 →
Scapy
Engine
RT
Dashboard
Voice
Alerts
TCP/UDP
Coverage
◉ Active
SPH1NX
// Network scan detector · JARVIS-style voice alerts

Python-based network scan detector for TCP Null/UDP scans with JARVIS-like voice alerts for critical ports (FTP, Telnet, SSH, etc.) and a real-time UI dashboard. Powered by Scapy for live packet capture and analysis.

// Detection pipeline
Network live packets Scapy packet sniffer Scan detect TCP/UDP null Port triage critical ports Voice alert + UI dash
PythonScapyTCP/UDPreal-time UIvoice alerts
github.com/AnonymousSingh-007/SPH1NX →
Port
Mapper
Svc
Identifier
Python
Core
CLI
Interface
◉ Active
P.R.I.S.M
// Port Response Identifier & Service Mapper

A Python-based port scanning and service identification tool. Maps open ports to running services with banner grabbing and protocol fingerprinting. Designed for fast recon during offensive security assessments.

// Recon pipeline
Target IP / host Port scan TCP connect Banner grab fingerprint Svc map port→service Report CLI output
Pythonsocketbanner grabservice fingerprint
github.com/AnonymousSingh-007/P.R.I.S.M →
PS
PowerShell
Win
Platform
Auto
Enumerate
CLI
Interface
⏳ Ongoing
T.E.M.P.E.S.T
// Threat-surface Enumeration: Modules, Ports, Extensions, Schedules & Tasks

PowerShell-based automated threat-surface enumeration for Windows environments. Covers open ports, scheduled tasks, loaded modules, and startup extensions — generates a structured threat exposure report.

// Enumeration pipeline
Windows target host PS modules port · tasks Extensions startup scan Risk score triage Report structured
PowerShellWMIscheduled tasksport enumWindows
github.com/AnonymousSingh-007/T.E.M.P.E.S.T →
05 — Cyber range

Execution evidence.

14th
National rank
Ciphathon CTF
Top 5%
Global
HackTheBox
Top 7%
Global
TryHackMe
Active
Competitor
SecLeaf
Award winner
ICEM · NCTAAI

HTB focus

  • Web exploitation
  • Reverse engineering
  • Cryptography · forensics
  • ML challenges

Toolchain

  • Burp Suite · SQLmap
  • Nmap · Metasploit
  • Hashcat · Hydra
  • Gobuster · FFUF

SecLeaf

  • Forensics · crypto
  • Reverse engineering
  • Web · pwn
  • Active rotation
06 — Capability matrix

Operational domains.

Adversarial AI

  • LLM red teaming & prompt injection taxonomy
  • Adversarial ML — evasion & poisoning
  • Multi-agent RL robustness analysis
  • Threat modeling & attack-tree construction
  • Behavioral biometric adversarial primitives

Offensive security

  • Web exploitation & network security
  • OSINT & digital footprinting
  • Reverse engineering & forensics
  • Network scan detection (Scapy / SPH1NX)
  • CTF: crypto, DFIR, pwn, web

Offensive toolchain

  • Nmap · Zenmap · Gobuster · FFUF
  • Metasploit · BeEF · Hydra · Medusa
  • Burp Suite · OWASP ZAP · SQLmap
  • Hashcat · John the Ripper
  • Scapy · Wireshark

Engineering

  • Python — primary research & tooling
  • PowerShell — Windows threat enum
  • PyTorch / CUDA · NetworkX · Scapy
  • MySQL · LaTeX — academic writing

Reading interests

  • QKD & post-quantum migration
  • Quantum-adversarial machine learning
  • Hypergraph-based threat reconnaissance
07 — Mission log

Trajectory.

2025 — present
Research Intern · MARL drone swarm coordination
DIAT (Defence Institute of Advanced Technology, DRDO-affiliated) · Pune

QMIX-based multi-agent RL for jammer avoidance. Phase D: +13.0 detection points via KL-anchored Navigator. Building toward IEEE TNNLS submission.

2025 — present
Software Engineer Intern
Dexpert Systems · Pune

Secure infrastructure, VPN architecture, OWASP vulnerability remediation. Zero-trust implementation across production API surfaces.

2024
Research Intern · AI security
ICAR · Baramati

Developed the ICEM adversarial taxonomy. Best Paper and Best Presentation at ICEM 2024 / NCTAAI 4.0.

2023 — ongoing
Independent security researcher
Self-directed · GitHub · CTF circuit

National-ranked CTF competitor. Building RAV3N-SEC, SPH1NX, P.R.I.S.M, T.E.M.P.E.S.T. 4 public repositories, 33 stars.

2023 — 2027
B.Tech CSE (Cybersecurity)
Pre-final year · Maharashtra

Specializing in cybersecurity with active research publications, internships, and CTF circuit involvement alongside coursework.

08 — Credentials

Certifications.

CEH v12
EC-Council
Advanced Network Attacks
Practical cert
Web Application Hacking
Pentest cert
Cryptography Fundamentals
Applied crypto
HTB Pro Hacker
HackTheBox — Top 5%
GitHub Pull Shark
GitHub Achievement
09 — Establish contact

Open channel.

Open to research collaboration, security consulting, CTF team invitations, and speaking — particularly in adversarial AI, multi-agent robustness, and applied cryptographic security.

// Status panel
Response SLA24–48 hours
Preferred channelEmail
Open toResearch · Consulting · CTF
Current focusIEEE Q1 submission
Status◉ Open to work
LocationMaharashtra, IN
GitHub13 repos · 33 stars