Issa Baddour
PhD Candidate, Computer Science & Engineering, IIT Jodhpur

Google Scholar || LinkedIn || GitHub || ORCID

Email: baddour.1@iitj.ac.in
Lab 203, CSE Department, IIT Jodhpur, India

Advised by Prof. Somitra Kumar Sanadhya and Prof. Dip Sankar Banerjee

Research | Publications | Projects | Education | Teaching | Awards | Service

My research exposes and mitigates information leakage in GPU systems across three layers: microarchitectural resources, interconnect fabrics (NVLink), and LLM inference. I am now seeking research positions to advance the security of AI infrastructure.

• GPU Security
• AI Security
• Microarchitectural Attacks
• Side Channels & Covert Channels

• Microarchitectural layer: Extended binary contention to four-level signaling on NVIDIA DGX A100. Achieved 44 Mbps (1 GPU) and 340 Mbps (8 GPUs). Behavioral randomization via SFU function combinations reduces ML detectability from 100% to 36%.
• Interconnect layer: First study to expose NVLink contention across V100, A100, and H200. Covert channel: 6.33-9.90 Kbps with negligible error rates. Application fingerprinting via NVLink latency traces: 96.2% accuracy across 25 apps. Temporal and spatial partitioning countermeasures evaluated.
• Application layer (LLM inference): Unprivileged NVML telemetry alone reveals model identity and prompt semantics. Model fingerprinting: 100% at single-tenant, degrading to 83.8% under four-way contention. Prompt-family leakage significant in 18 of 35 tested cells. Telemetry-layer mitigations evaluated.

• (2026) Issa Baddour, Somitra Kumar Sanadhya, Dip Sankar Banerjee, SideLink: Exposing NVLink to Covert and Side-Channel Attacks, Journal of Hardware and Systems Security (Special Issue) DOI
• (2025) Issa Baddour, Somitra Kumar Sanadhya, Dip Sankar Banerjee, An Improved Micro-Architectural Covert-Channel Attack on GPUs, Journal of Hardware and Systems Security DOI
• (2024) Issa Baddour, Somitra Kumar Sanadhya, Dip Sankar Banerjee, SideLink: Exposing NVLink to Covert and Side-Channel Attacks (Work-in-Progress), SPACE 2024 DOI
• (Under Review, AAAI 2027) Issa Baddour, Somitra Kumar Sanadhya, Dip Sankar Banerjee, Side-Channel Leakage in Large Language Model Inference on GPUs

• Execution-Level Covert Channels on GPUs — 44 Mbps (1 GPU), 340 Mbps (8 GPUs), 36% ML detectability after evasion. Reverse-engineered warp/block scheduling on NVIDIA A100 to build a four-level covert signaling scheme with dynamic SFU randomization. Paper
• SideLink: NVLink Side-Channel & Covert-Channel Attacks — 6.33-9.90 Kbps covert channel, 96.2% app fingerprinting across V100/A100/H200. Proposed temporal and spatial partitioning countermeasures. Journal / SPACE 2024
• Side-Channel Leakage in LLM Inference on GPUs — 100% single-tenant model fingerprinting across 7 LLMs using unprivileged NVML telemetry only; prompt-family leakage significant in 18/35 cells. Under review, AAAI 2027.
• FPGA-Accelerated SHA-256 Password Recovery via Partial Reconfiguration — Graduate course project, IIT Jodhpur, 2021. Isolated length-independent logic into a static region to shrink the reconfigured area on the accelerator; implemented in VHDL, unit-tested in Vivado (design and simulation only; supervised by Dr. Somitra Sanadhya).

• PhD, Computer Science & Engineering — Indian Institute of Technology, Jodhpur (2021-Present), GPA 9.25/10.
• M.Tech, Computer Science & Engineering — National Institute of Technology, Warangal (2018-2020)
• B.E., Computer Engineering — Damascus University (2011-2016). Ranked 1st in class (Gold-medalist equivalent).

• Teaching Assistant, Algorithms for Big Data — IIT Jodhpur (2023, 2024)
• Teaching Assistant, Network Applications Programming — Damascus University
• Teaching Assistant, Advanced Programming — Damascus University

• Ranked 1st in B.E. (Gold-medalist equivalent)
• Fully funded international sponsorship for graduate studies
• ICCR Scholarship (Government of India) for M.Tech
• Multiple year-wise academic excellence awards (2012-2016)

• Reviewer: IEEE Transactions on Dependable and Secure Computing (TDSC)
• Sub-reviewer: CF 2025, SPACE 2025, CANS 2024

• Paper presentation — SPACE 2024
• Poster presentation — Cryptology Conclave, IIT Hyderabad, 2026
• Conference participation — Indocrypt 2022/2023, SPACE 2024, Digital Forensics Conf. 2026
• Workshop participation — NWC 2023, NIWC 2026