About
I'm Aritra
Final year M.Tech CSE student at IIT Bombay, with strong interests in Hardware Security, System Software, and Software Engineering.
- Dept. Rank 1 – CSE, RCCIIT
- Knight on LeetCode (1800+ rating)
- Filecoin Track Prize winner – Protocol Labs
- Top 6 – HackNPitch Hackathon (JU E-Cell)
Education
IIT Bombay
M.Tech - Computer Science
2024 – 2026*
CPI: 9.46
RCCIIT (MAKAUT)
B.Tech - Computer Science
2020 – 2024
CPI: 9.43
Projects
Scalable Distributed Key-Value Store with Fault Tolerance
- Built a fault-tolerant in-memory KV store with 2× replication and atomic two-phase commit.
- Ensured strong consistency with efficient request handling across distributed servers.
- Added heartbeat-based failure detection, seamless data migration, and write-through LRU caching.
GrpTorrent : P2P File Sharing System
- Developed group-based P2P sharing with centralized tracker for authentication and metadata.
- Implemented multithreaded socket server enabling concurrent uploads/downloads.
- Used piece-wise SHA-1 hashing for integrity and efficient piece selection.
Gurukul : AI-Enabled Q&A Portal
- Django-based platform with email OTP login, markdown, profanity filter, and user management.
- Implemented CRUD for questions, answers, replies with advanced search and profiles.
- Integrated AI-generated first answers using DeepSeek-R1 via OpenRouter API.
GPU Uncore Based Side & Covert Channels
- Designed a DRAM-frequency covert channel on AMD RDNA3 GPUs achieving ~98.6% accuracy at 3.7 bps.
- Demonstrated feasibility of GPU uncore-level information leakage.
- Proposed extensions to additional uncore components and ML model identification via side channels.
Matrix Multiplication & Embedding Optimization
- Achieved 3× speedup via loop reordering, unrolling, tiling, SIMD and optimized transpose.
- Applied optimizations to accelerate NN embedding operations.
- Boosted throughput using software prefetching and SIMD for large embedding tables.
Flush+Reload Cache Side-Channel Attack
- Implemented Flush+Reload timing attack exploiting shared cache behavior.
- Extracted victim access patterns through precise cache-line reload timing.
- Analyzed microarchitectural leakage conditions affecting attack success.