Love Kush Pranu lutherleo

$ whoami

name:        Love Kush Pranu
role:        Cybersecurity Engineer · Research Coordinator & Infrastructure Lead
focus:       kernel fuzzing, AI-driven vulnerability discovery, post-quantum crypto
based_in:    Brooklyn, NY · lovekushpranu.com
currently:   NYU OSIRIS Lab — coordinating 9 security research projects, leading RaceFuzz
also:        MS Cybersecurity @ NYU Tandon ('26) · GPA 3.9/4.0
credentials: CompTIA Security+ · CompTIA CySA+ · CCNA (in progress)
ctf:         CSAW CTF 2025 challenge author · OSIRIS CTF team
reachable:   lp2989@nyu.edu · linkedin.com/in/lovekush-pranu

❯ What I build

Bug-Hunting Machines

_{Fuzzers and agentic scanners that find real bugs — kernel races, shell-injection RCEs — and prove them with working exploits.}

Hardened Low-Level Systems

_{Kernels, sandboxes, emulators and secure channels in C, Go and Rust — ASLR, W^X, post-quantum handshakes.}

Security Research Infrastructure

_{Serverless AWS platforms for 5,000+ participants, Slurm-backed GPU grading and CTF hosting at global scale.}

❯ How a bug moves through my pipeline

%%{init: {"theme":"base","themeVariables":{"darkMode":true,"background":"#0F172A","git0":"#10B981","gitBranchLabel0":"#0F172A","primaryColor":"#111827","primaryBorderColor":"#10B981","primaryTextColor":"#E2E8F0","secondaryColor":"#064E3B","secondaryBorderColor":"#34D399","secondaryTextColor":"#ECFDF5","tertiaryColor":"#0B1220","tertiaryBorderColor":"#06B6D4","tertiaryTextColor":"#E2E8F0","mainBkg":"#111827","nodeBorder":"#10B981","clusterBkg":"#0B1220","clusterBorder":"#1E293B","lineColor":"#06B6D4","textColor":"#A7F3D0","edgeLabelBackground":"#0F172A","actorBkg":"#111827","actorBorder":"#10B981","actorTextColor":"#E2E8F0","actorLineColor":"#334155","signalColor":"#06B6D4","signalTextColor":"#A7F3D0","labelBoxBkgColor":"#111827","labelBoxBorderColor":"#10B981","labelTextColor":"#E2E8F0","loopTextColor":"#A7F3D0","noteBkgColor":"#064E3B","noteBorderColor":"#34D399","noteTextColor":"#ECFDF5","activationBkgColor":"#047857","activationBorderColor":"#34D399","sequenceNumberColor":"#0F172A","cScale0":"#064E3B","cScaleLabel0":"#ECFDF5","cScalePeer0":"#34D399","cScale1":"#0E7490","cScaleLabel1":"#ECFDF5","cScalePeer1":"#34D399","cScale2":"#047857","cScaleLabel2":"#ECFDF5","cScalePeer2":"#34D399","cScale3":"#155E75","cScaleLabel3":"#ECFDF5","cScalePeer3":"#34D399","cScale4":"#065F46","cScaleLabel4":"#ECFDF5","cScalePeer4":"#34D399","cScale5":"#0369A1","cScaleLabel5":"#ECFDF5","cScalePeer5":"#34D399","cScale6":"#064E3B","cScaleLabel6":"#ECFDF5","cScalePeer6":"#34D399","cScale7":"#0E7490","cScaleLabel7":"#ECFDF5","cScalePeer7":"#34D399","cScale8":"#047857","cScaleLabel8":"#ECFDF5","cScalePeer8":"#34D399","cScale9":"#155E75","cScaleLabel9":"#ECFDF5","cScalePeer9":"#34D399","cScale10":"#065F46","cScaleLabel10":"#ECFDF5","cScalePeer10":"#34D399","cScale11":"#0369A1","cScaleLabel11":"#ECFDF5","cScalePeer11":"#34D399"}}}%%
sequenceDiagram
    autonumber
    participant T as Target · repo / kernel
    participant G as Code graph · Joern · Neo4j
    participant A as Agents · LLM fleet via MCP
    participant V as Verifier · independent agent
    participant P as Exploit · PoC
    participant R as Report · disclosure
    T->>G: build code property graph
    G->>A: taint paths & suspicious sinks
    A->>V: candidate vulnerability
    V-->>A: false positives rejected
    V->>P: confirmed finding
    P->>R: working proof-of-concept
    Note over G,V: Orion · 14 of 15 detections, zero false positives on NodeGoat

❯ Engineering principles

Break it to build it better. Every system I ship gets attacked by me first — my own protocols, my own kernels, my own challenges.

A finding without a PoC is a rumor. Detections are verified independently and backed by a working exploit before they count.

Go one layer deeper. From CAN bus and firmware to the Linux network stack — the interesting bugs live below the abstraction.

❯ Technology ecosystem

%%{init: {"theme":"base","themeVariables":{"darkMode":true,"background":"#0F172A","git0":"#10B981","gitBranchLabel0":"#0F172A","primaryColor":"#111827","primaryBorderColor":"#10B981","primaryTextColor":"#E2E8F0","secondaryColor":"#064E3B","secondaryBorderColor":"#34D399","secondaryTextColor":"#ECFDF5","tertiaryColor":"#0B1220","tertiaryBorderColor":"#06B6D4","tertiaryTextColor":"#E2E8F0","mainBkg":"#111827","nodeBorder":"#10B981","clusterBkg":"#0B1220","clusterBorder":"#1E293B","lineColor":"#06B6D4","textColor":"#A7F3D0","edgeLabelBackground":"#0F172A","actorBkg":"#111827","actorBorder":"#10B981","actorTextColor":"#E2E8F0","actorLineColor":"#334155","signalColor":"#06B6D4","signalTextColor":"#A7F3D0","labelBoxBkgColor":"#111827","labelBoxBorderColor":"#10B981","labelTextColor":"#E2E8F0","loopTextColor":"#A7F3D0","noteBkgColor":"#064E3B","noteBorderColor":"#34D399","noteTextColor":"#ECFDF5","activationBkgColor":"#047857","activationBorderColor":"#34D399","sequenceNumberColor":"#0F172A","cScale0":"#064E3B","cScaleLabel0":"#ECFDF5","cScalePeer0":"#34D399","cScale1":"#0E7490","cScaleLabel1":"#ECFDF5","cScalePeer1":"#34D399","cScale2":"#047857","cScaleLabel2":"#ECFDF5","cScalePeer2":"#34D399","cScale3":"#155E75","cScaleLabel3":"#ECFDF5","cScalePeer3":"#34D399","cScale4":"#065F46","cScaleLabel4":"#ECFDF5","cScalePeer4":"#34D399","cScale5":"#0369A1","cScaleLabel5":"#ECFDF5","cScalePeer5":"#34D399","cScale6":"#064E3B","cScaleLabel6":"#ECFDF5","cScalePeer6":"#34D399","cScale7":"#0E7490","cScaleLabel7":"#ECFDF5","cScalePeer7":"#34D399","cScale8":"#047857","cScaleLabel8":"#ECFDF5","cScalePeer8":"#34D399","cScale9":"#155E75","cScaleLabel9":"#ECFDF5","cScalePeer9":"#34D399","cScale10":"#065F46","cScaleLabel10":"#ECFDF5","cScalePeer10":"#34D399","cScale11":"#0369A1","cScaleLabel11":"#ECFDF5","cScalePeer11":"#34D399"}}}%%
mindmap
  root((Security Engineering))
    Offense
      Kernel fuzzing
      Binary exploitation
      ROP chains
      CTF challenge design
      OSINT
    Systems
      Linux kernel
      xv6
      eBPF · nftables
      Lind sandbox · WebAssembly
      CHIP-8 emulation
    Cryptography
      ML-KEM · Kyber
      ML-DSA · Dilithium
      HQC · SPHINCS+
      liboqs
      AES-GCM · HKDF
    AI Security
      LLM agents
      Model Context Protocol
      Adversarial ML
      QLoRA fine-tuning
      CleverHans · ART
    Infrastructure
      AWS serverless
      DynamoDB
      Slurm GPU clusters
      Docker
    Embedded & Automotive
      CAN bus
      ECU firmware
      TARA · ISO 26262

_{Primary languages → Python · C · Go · Rust · Bash · Also in production → C++ · C# · TypeScript · Perl · PyTorch · TensorFlow · Neo4j · Joern · AWS · Git}

❯ Career journey

%%{init: {"theme":"base","themeVariables":{"darkMode":true,"background":"#0F172A","git0":"#10B981","gitBranchLabel0":"#0F172A","primaryColor":"#111827","primaryBorderColor":"#10B981","primaryTextColor":"#E2E8F0","secondaryColor":"#064E3B","secondaryBorderColor":"#34D399","secondaryTextColor":"#ECFDF5","tertiaryColor":"#0B1220","tertiaryBorderColor":"#06B6D4","tertiaryTextColor":"#E2E8F0","mainBkg":"#111827","nodeBorder":"#10B981","clusterBkg":"#0B1220","clusterBorder":"#1E293B","lineColor":"#06B6D4","textColor":"#A7F3D0","edgeLabelBackground":"#0F172A","actorBkg":"#111827","actorBorder":"#10B981","actorTextColor":"#E2E8F0","actorLineColor":"#334155","signalColor":"#06B6D4","signalTextColor":"#A7F3D0","labelBoxBkgColor":"#111827","labelBoxBorderColor":"#10B981","labelTextColor":"#E2E8F0","loopTextColor":"#A7F3D0","noteBkgColor":"#064E3B","noteBorderColor":"#34D399","noteTextColor":"#ECFDF5","activationBkgColor":"#047857","activationBorderColor":"#34D399","sequenceNumberColor":"#0F172A","cScale0":"#064E3B","cScaleLabel0":"#ECFDF5","cScalePeer0":"#34D399","cScale1":"#0E7490","cScaleLabel1":"#ECFDF5","cScalePeer1":"#34D399","cScale2":"#047857","cScaleLabel2":"#ECFDF5","cScalePeer2":"#34D399","cScale3":"#155E75","cScaleLabel3":"#ECFDF5","cScalePeer3":"#34D399","cScale4":"#065F46","cScaleLabel4":"#ECFDF5","cScalePeer4":"#34D399","cScale5":"#0369A1","cScaleLabel5":"#ECFDF5","cScalePeer5":"#34D399","cScale6":"#064E3B","cScaleLabel6":"#ECFDF5","cScalePeer6":"#34D399","cScale7":"#0E7490","cScaleLabel7":"#ECFDF5","cScalePeer7":"#34D399","cScale8":"#047857","cScaleLabel8":"#ECFDF5","cScalePeer8":"#34D399","cScale9":"#155E75","cScaleLabel9":"#ECFDF5","cScalePeer9":"#34D399","cScale10":"#065F46","cScaleLabel10":"#ECFDF5","cScalePeer10":"#34D399","cScale11":"#0369A1","cScaleLabel11":"#ECFDF5","cScalePeer11":"#34D399"}}}%%
timeline
    2022 : Robert Bosch India — Security Intern : CAN bus · ECU · firmware TARA with Porsche
    2023 : B.Tech EEE, NIT Calicut : Pentadots — Cybersecurity Engineer · zero-trust · PoC exploits
    2024 : MS Cybersecurity, NYU Tandon : OSIRIS Lab — Infrastructure Lead · 5,000+ participants
    2025 : CSAW CTF 2025 challenge author : Lind sandbox research · adversarial ML
    2026 : Research Coordinator, OSIRIS Lab : 9 projects · RaceFuzz · Orion · post-quantum channel

Period	Company	Role	Key impact
Sep 2024 — now	NYU OSIRIS Lab	Research Coordinator & Infrastructure Lead	Serverless AWS architecture for 5,000+ participants · Slurm-backed GPU grading pipeline · 9 research projects coordinated
Sep 2023 — Jan 2025	Pentadots Solutions	Cybersecurity Engineer	PoC exploit code & automation across 5+ production deployments · zero-trust frameworks, 30% defense-efficiency gain
Jun 2022 — Aug 2022	Robert Bosch India Ltd.	E-Drive System Security Intern	CAN bus, ECU & firmware attack surfaces · TARA with Porsche engineering against ISO 26262 · 25% higher validation coverage

❯ Featured engineering projects

Orion — GraphRAG code-security scanner with a fleet of verifying agents · repo

Multi-agent LLM system that reasons over a Joern code property graph in Neo4j through the Model Context Protocol. Discovery agents propose vulnerabilities; independent agents verify them. 14 of 15 detections at zero false positives on NodeGoat, and a shell-injection RCE found in a 160,000-node production TypeScript repository — with proof-of-concept exploits. Covered by 26 unit tests.

%%{init: {"theme":"base","themeVariables":{"darkMode":true,"background":"#0F172A","git0":"#10B981","gitBranchLabel0":"#0F172A","primaryColor":"#111827","primaryBorderColor":"#10B981","primaryTextColor":"#E2E8F0","secondaryColor":"#064E3B","secondaryBorderColor":"#34D399","secondaryTextColor":"#ECFDF5","tertiaryColor":"#0B1220","tertiaryBorderColor":"#06B6D4","tertiaryTextColor":"#E2E8F0","mainBkg":"#111827","nodeBorder":"#10B981","clusterBkg":"#0B1220","clusterBorder":"#1E293B","lineColor":"#06B6D4","textColor":"#A7F3D0","edgeLabelBackground":"#0F172A","actorBkg":"#111827","actorBorder":"#10B981","actorTextColor":"#E2E8F0","actorLineColor":"#334155","signalColor":"#06B6D4","signalTextColor":"#A7F3D0","labelBoxBkgColor":"#111827","labelBoxBorderColor":"#10B981","labelTextColor":"#E2E8F0","loopTextColor":"#A7F3D0","noteBkgColor":"#064E3B","noteBorderColor":"#34D399","noteTextColor":"#ECFDF5","activationBkgColor":"#047857","activationBorderColor":"#34D399","sequenceNumberColor":"#0F172A","cScale0":"#064E3B","cScaleLabel0":"#ECFDF5","cScalePeer0":"#34D399","cScale1":"#0E7490","cScaleLabel1":"#ECFDF5","cScalePeer1":"#34D399","cScale2":"#047857","cScaleLabel2":"#ECFDF5","cScalePeer2":"#34D399","cScale3":"#155E75","cScaleLabel3":"#ECFDF5","cScalePeer3":"#34D399","cScale4":"#065F46","cScaleLabel4":"#ECFDF5","cScalePeer4":"#34D399","cScale5":"#0369A1","cScaleLabel5":"#ECFDF5","cScalePeer5":"#34D399","cScale6":"#064E3B","cScaleLabel6":"#ECFDF5","cScalePeer6":"#34D399","cScale7":"#0E7490","cScaleLabel7":"#ECFDF5","cScalePeer7":"#34D399","cScale8":"#047857","cScaleLabel8":"#ECFDF5","cScalePeer8":"#34D399","cScale9":"#155E75","cScaleLabel9":"#ECFDF5","cScalePeer9":"#34D399","cScale10":"#065F46","cScaleLabel10":"#ECFDF5","cScalePeer10":"#34D399","cScale11":"#0369A1","cScaleLabel11":"#ECFDF5","cScalePeer11":"#34D399"}}}%%
flowchart LR
    A[("Source repo<br>160K-node graph")] --> B["Joern CPG<br>→ Neo4j"]
    B --> C["MCP tools<br>graph queries"]
    C --> D["Discovery agents"]
    D --> E["Verifier agents"]
    E --> F["PoC exploit<br>14/15 · 0 FP"]
    style A fill:#064E3B,stroke:#34D399,color:#ECFDF5
    style F fill:#164E63,stroke:#06B6D4,color:#ECFDF5

_{Python LLM agents MCP Neo4j Joern GraphRAG}

RaceFuzz — race-condition fuzzer for the Linux network stack · OSIRIS Lab project I lead

Semantic fuzzing framework targeting race conditions in the Linux kernel's TCP stack. A Go orchestrator drives a C kernel module over generic netlink, combining coverage-guided mutation, Go-based corpus generation and concurrency analysis. Authored 16 race templates across BPF, nftables and net/sched.

%%{init: {"theme":"base","themeVariables":{"darkMode":true,"background":"#0F172A","git0":"#10B981","gitBranchLabel0":"#0F172A","primaryColor":"#111827","primaryBorderColor":"#10B981","primaryTextColor":"#E2E8F0","secondaryColor":"#064E3B","secondaryBorderColor":"#34D399","secondaryTextColor":"#ECFDF5","tertiaryColor":"#0B1220","tertiaryBorderColor":"#06B6D4","tertiaryTextColor":"#E2E8F0","mainBkg":"#111827","nodeBorder":"#10B981","clusterBkg":"#0B1220","clusterBorder":"#1E293B","lineColor":"#06B6D4","textColor":"#A7F3D0","edgeLabelBackground":"#0F172A","actorBkg":"#111827","actorBorder":"#10B981","actorTextColor":"#E2E8F0","actorLineColor":"#334155","signalColor":"#06B6D4","signalTextColor":"#A7F3D0","labelBoxBkgColor":"#111827","labelBoxBorderColor":"#10B981","labelTextColor":"#E2E8F0","loopTextColor":"#A7F3D0","noteBkgColor":"#064E3B","noteBorderColor":"#34D399","noteTextColor":"#ECFDF5","activationBkgColor":"#047857","activationBorderColor":"#34D399","sequenceNumberColor":"#0F172A","cScale0":"#064E3B","cScaleLabel0":"#ECFDF5","cScalePeer0":"#34D399","cScale1":"#0E7490","cScaleLabel1":"#ECFDF5","cScalePeer1":"#34D399","cScale2":"#047857","cScaleLabel2":"#ECFDF5","cScalePeer2":"#34D399","cScale3":"#155E75","cScaleLabel3":"#ECFDF5","cScalePeer3":"#34D399","cScale4":"#065F46","cScaleLabel4":"#ECFDF5","cScalePeer4":"#34D399","cScale5":"#0369A1","cScaleLabel5":"#ECFDF5","cScalePeer5":"#34D399","cScale6":"#064E3B","cScaleLabel6":"#ECFDF5","cScalePeer6":"#34D399","cScale7":"#0E7490","cScaleLabel7":"#ECFDF5","cScalePeer7":"#34D399","cScale8":"#047857","cScaleLabel8":"#ECFDF5","cScalePeer8":"#34D399","cScale9":"#155E75","cScaleLabel9":"#ECFDF5","cScalePeer9":"#34D399","cScale10":"#065F46","cScaleLabel10":"#ECFDF5","cScalePeer10":"#34D399","cScale11":"#0369A1","cScaleLabel11":"#ECFDF5","cScalePeer11":"#34D399"}}}%%
flowchart LR
    G["Corpus gen<br>Go"] --> M["Semantic mutation<br>coverage-guided"] --> O["Orchestrator<br>Go"] -->|generic netlink| K["Kernel module<br>C"] --> D["Race detection<br>BPF · nftables · net/sched"]
    style D fill:#064E3B,stroke:#34D399,color:#ECFDF5

_{Go C Linux kernel TCP/IP Generic netlink BPF}

Kyber Crystals — one secure channel, two post-quantum families · repo

Research-grade post-quantum secure channel in C, implemented twice behind one pluggable backend: ML-KEM-768 + ML-DSA-65 and lattice-free HQC-192 + SLH-DSA (SPHINCS+). Both sealed with AES-256-GCM and HKDF-SHA256, then attacked and benchmarked in a unified Docker harness at NIST PQ Category 3.

_{C liboqs ML-KEM ML-DSA SPHINCS+ AES-GCM Docker}

Gemma Intrusion Detection — LLM fine-tuning pipeline for network IDS

Machine learning pipeline that fine-tunes Google's Gemma LLM on the UNSW-NB15 intrusion-detection dataset using QLoRA with Unsloth.

_{Python PyTorch Unsloth QLoRA}

xv6 Security Enhancements — hardening a teaching kernel · repo

Kernel-level hardening of xv6: ASLR, W^X memory policies, stack canaries, process isolation and custom system calls for security auditing — alongside OS security research in Rust on the Lind sandbox (WebAssembly-based POSIX sandbox) at NYU Tandon.

_{C xv6 ASLR W^X Rust WebAssembly}

Adversarial Robustness in DNNs — validating Madry et al. (2018)

PGD adversarial training on MNIST, re-engineering TensorFlow 1.x code for 2.x so CleverHans and ART experiments could run. ~90% accuracy against PGD and FGSM, with weaknesses identified under C&W attacks and distribution shift.

_{Python TensorFlow CleverHans ART}

CHIP-8 Emulator — from scratch in C · repo

All 35 opcodes, 64×32 XOR sprite rendering, SDL2 graphics and input, delay/sound timers and a 16-key hex keypad.

_{C SDL2 CPU emulation}

❯ Impact at a glance

14 / 15
_{detections · zero false positives}

160K
_{node repo · RCE found}

5,000+
_{participants on infra I built}

16
_{kernel race templates}

9
_{research projects coordinated}

❯ GitHub

`lovekush@github ~ $ ./contributions.sh`

`lovekush@github ~ $ ./stats.sh`

❯ Currently exploring

→ semantic fuzzing for concurrency bugs in the Linux network stack
→ agentic vulnerability discovery with independent verification
→ post-quantum protocol design beyond lattices

$ reach me
→ email     lp2989@nyu.edu
→ linkedin  linkedin.com/in/lovekush-pranu
→ web       lovekushpranu.com
→ github    github.com/lutherleo

_{Open to security engineering, vulnerability research and systems security roles.}

Navigation Menu