Skip to content
View nishankswamy's full-sized avatar

Highlights

  • Pro

Block or report nishankswamy

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
nishankswamy/README.md

Hi, I'm Nishank Daulath 👋

Security engineering & data analytics. I build things I can explain in depth, then measure them until they tell me something I didn't expect.

🌐 Portfolio: nishankswamy.github.io · finishing an MSc in Cyber Security in Berlin.

I recently ran a 30-day build challenge: ten projects across security and data, each shipped with a benchmark, tests, CI, and a written finding that contradicts the obvious assumption. 831 tests · CI on every repo · 2 capstone platforms.

The rule for every project: build the core, do the hard half, then break it and write down the result. A few that came out of that:

  • Forged a MAC without the key. SHA256(secret‖message) is length-extendable — a digest is the hash's internal state. Appended role=admin to a signed message in milliseconds. → applied-cryptography
  • A benchmark that lied by 600x. A "678x cache speedup" was timing the error-fallback path; the real number was 1.1x. Now I distrust results that flatter the design. → url-shortener
  • ML lost to five lines of statistics. Isolation Forest scored worse than an STL+MAD baseline on security telemetry at a realistic base rate. "Use a model" is not a strategy. → anomaly-detection

The two capstones

What it is The finding
🛡️ SIEM platform Multi-source ingest → inline detection + behavioural baselines → entity-first investigation Detects a full attack campaign in 13ms; alert → evidence is one indexed lookup, 26x faster than a scan
📊 Analytics platform Quality-gated ingest → star schema with cubes → self-serve queries with a cost display What breaks at scale is memory, not compute — 530 B/row as dicts vs 50 on disk

All ten projects

Cryptography · detection engineering · a columnar query engine · anomaly detection · network traffic analysis · an exactly-once streaming pipeline · differential privacy — plus the two capstones. Everything is defensive; the only things I attack, I attack in my own code.

github.com/nishankswamy/30-days — the full index, findings, and case studies.

Reach me

📧 Nishankde@gmail.com · 🌐 Portfolio · 💼 LinkedIn

Popular repositories Loading

  1. Weatherapp-by-Nishank Weatherapp-by-Nishank Public

    Java

  2. deepfake-detection-generalization deepfake-detection-generalization Public

    Deepfake detection with EfficientNet-B0 — investigating cross-generator generalization failure via Grad-CAM

    Jupyter Notebook

  3. cloud-nids-msc cloud-nids-msc Public

    Python

  4. forest-fire-dashboard forest-fire-dashboard Public

    Forest fire detection over a LoRa sensor network. Six Raspberry Pis in a two-cluster multi-hop topology with TDMA scheduling, plus a 50-node browser simulator with LEACH clustering and GPSR routing.

    Python

  5. url-shortener url-shortener Public

    Python

  6. applied-cryptography applied-cryptography Public

    Python