Research
My work sits at the intersection of networked systems, security, and programmable
hardware. The projects below are the main threads of my Ph.D. and industry research;
see the publications page for the papers that came out of them.
Programmable Dataplane Security for Low-Volume and Slow Attacks
2022–2026 · IEEE ICNP 2026, IEEE ICNP 2024, IEEE Cloud Summit 2024, ACM SIGCOMM N2Women 2023
- Analyzed traffic patterns of diverse network attacks, including Slowloris, SSH/FTP brute-force, DDoS, and network scanning.
- Proposed, designed, and validated an efficient approach for detecting attacks in Tbps-scale network traffic.
- Designed a hybrid programmable-switch and SmartNIC architecture for Tbps-scale monitoring, suspicious-flow identification, and mitigation of low-volume attacks.
- Developed memory-efficient in-switch flow tracking and filtering that reduces traffic forwarded to downstream security processors while preserving line-rate performance and low false-positive rates.
- Built multi-stage detection pipelines combining programmable dataplanes, SmartNIC-based inspection, and machine learning for scalable anomalous-traffic detection.
- Evaluated the system on real-world traces and programmable hardware testbeds.
High-Performance Packet Processing for 5G Infrastructure
Nokia Bell Labs and GWU · IEEE INFOCOM Workshops 2023, ACM SIGCOMM 5G-MeMU 2022
- Developed and optimized high-performance packet processing pipelines using DPDK (kernel bypass), improving throughput and reducing communication overhead in the 5G RAN Central Unit.
- Profiled end-to-end dataplane execution with
perf to identify CPU, cache, and I/O bottlenecks, enabling targeted optimizations for low-latency, high-throughput packet processing.
- Performed detailed latency and throughput analysis across the packet processing pipeline, driving system-level optimizations for scalable, real-time traffic handling.
SmartNIC-Based Flow Load Balancer
IEEE NFV-SDN 2021
- Designed and implemented a flow-aware load balancer on SmartNICs using weighted scheduling to improve flow-distribution consistency and maximize throughput in high-speed dataplanes.
- Developed traffic generators with Zipf-distributed workloads to simulate bursty and skewed traffic under realistic, non-uniform conditions.
- Evaluated latency, tail latency, and fairness with wrk2 and pktgen, comparing weighted and randomized scheduling strategies under high load.
Skills
- Languages: Python, Go, C/C++, P4, Shell, SQL
- Networking systems: Linux, DPDK, SmartNICs, programmable switches (Tofino)
- Protocols: TCP/IP, IPv4/IPv6, DNS, UDP, HTTP/HTTPS, BGP, OSPF
- Machine learning: Random Forest, XGBoost, decision trees, model evaluation, imbalanced-data analysis, scikit-learn
- Data engineering: MySQL, MongoDB, Redis, Kafka, large-scale network trace processing
- Tools: tcpdump, Wireshark, perf, GDB, pktgen, wrk2, Git, Docker