01AI security / LLM
LLM Firewall & RAG Security Lab
A defensive gateway for prompt injection, canary DLP, output inspection, structured-response validation, and bounded tool authorization.
View repository ↗Cybersecurity systems · AI/ML security · cloud
Making complex attack surfaces understandable.
I’m Niket, a cybersecurity postgraduate building practical defenses across SOC detection, LLM security, cloud privilege paths, software supply chains, and identity analytics.
5defensive security labs
AI/MLwith explainable decisions
2027M.Tech graduation target
DETECTIONevidence first
AI SECURITYbounded by policy
GRAPH PATHSrisk made visible
NKsecure
by design
by design
01 / PROFILE
Security builder with a systems mindset.
My work lives at the intersection of defensive engineering, machine learning, and the systems that make security decisions trustworthy.
I’m currently pursuing an M.Tech in Cyber Security Systems and Networks at Amrita School of Engineering. I like taking a difficult security question, turning it into a testable model, and making the result useful to the person who has to act on it.
“
Good security tooling should explain what it saw, why it matters, what it cannot know, and what a human should do next.
— working principle8.38CGPA / 10 in M.Tech
5public defensive repositories
18+validated SOC test cases
3active research builds
02 / SELECTED WORK
Small labs. Serious questions.
Every project is defensive, reproducible, and explicit about its limits. The goal is not to make a louder alert—it is to make a better decision.
02SOC / machine learning
Explainable AI SOC Detection
Hybrid detection using transparent telemetry rules, Isolation Forest anomaly scoring, evidence explanations, ATT&CK tags, and incident correlation.
View repository ↗03Cloud security / graph
Cloud Attack-Path Prioritizer
Read-only graph analysis that connects public exposure, trust relationships, privilege transitions, and sensitive assets into explainable remediation paths.
View repository ↗04AppSec / DevSecOps
SBOM Supply-Chain Intelligence
Dependency and vulnerability analysis with direct/transitive context, provenance, policy decisions, and SARIF output for developer workflows.
View repository ↗05Identity / behavioral ML
Identity Compromise Detector
Privacy-conscious anomaly detection for suspicious login behavior with behavioral signals, analyst explanations, and pseudonymous reporting.
View repository ↗RESEARCH IN PROGRESS
Three current builds pushing beyond the portfolio labs.
Render-aware prompt-injection defense
PyMuPDF, DistilBERT, XGBoost, SHAP, and PyTorch fused to detect hidden PDF instructions and semantic injection in enterprise RAG pipelines.
Research repository ↗MedTrust zero-trust healthcare platform
Cloud-native healthcare security with FastAPI, PostgreSQL RLS, Keycloak, OPA, mutual TLS, behavioral analytics, and post-quantum identity research.
More on GitHub ↗Cross-layer privilege escalation analysis
NetworkX-based AWS IAM and Kubernetes RBAC analysis modeling IRSA attack paths from a compromised pod toward administrator access.
Research repository ↗03 / TOOLKIT
Tools are useful. Context is the skill.
Security engineering
Detection engineering, SOC triage, incident correlation, ATT&CK mapping, threat modeling, digital forensics, and secure API design.
PythonFastAPIRESTpytest
AI & ML security
Isolation Forest, feature engineering, explainability, prompt-injection defense, RAG trust boundaries, DLP, SHAP, and model evaluation.
PyTorchDistilBERTXGBoostSHAP
Cloud & platform
AWS IAM, Kubernetes RBAC, IRSA, NetworkX graph modeling, Docker, PostgreSQL, Redis, JWT, OPA, and Keycloak.
AWSKubernetesDockerNetworkX
Ways of working
Reproducible fixtures, policy-as-code thinking, incremental testing, privacy-aware analytics, clear limitations, and human-readable evidence.
GitLinuxCI/CDSecurity docs
04 / BACKGROUND
Building the foundation, one layer at a time.
2025—2027
M.Tech · Cyber Security Systems and Networks
Amrita School of Engineering, Amritapuri
CGPA 8.38 / 102021—2025
B.Tech · Electrical and Electronics Engineering
Government Engineering College Barton Hill, Thiruvananthapuram
77.1%Languages
English · Malayalam · Tamil · Hindi
05 / NEXT CONVERSATION
Let’s talk about the next hard security problem.
I’m open to conversations about cybersecurity engineering, AI security, cloud security, research opportunities, and roles where careful systems thinking matters.