works on

From the 1 of 7 linked papers with an AI index.

collaborators

7 papers

cs.LG2026

RAG-HAR+: Towards Cost-Efficient LLM-Based Human Activity Recognition for Edge Deployment

Hansi Karunarathna, Nirhoshan Sivaroopan, Chamara Madarasingha +2

The paper presents RAG-HAR+, a cost‑efficient system that combines retrieval of similar sensor data with large language model (LLM) assistance to recognize human activities on edge…

cs.LG2026

STELLA: Efficient Sensor-to-LLM Translation for On-Device Human Activity Recognition

Nirhoshan Sivaroopan, Albert Zomaya, Kanchana Thilakarathna

HAR is increasingly expected to run continuously on edge devices, yet recent LLM-based methods remain hard to deploy: raw sensor prompts are long, cloud inference adds latency and…

cs.CV2026

RAG-HAR: Retrieval Augmented Generation-based Human Activity Recognition

Nirhoshan Sivaroopan, Hansi Karunarathna, Chamara Madarasingha +2

Human Activity Recognition (HAR) underpins applications in healthcare, rehabilitation, fitness tracking, and smart environments, yet existing deep learning approaches demand datase…

cs.NI2026

A Comprehensive Survey on Network Traffic Synthesis: From Statistical Models to Deep Learning

Nirhoshan Sivaroopan, Kaushitha Silva, Chamara Madarasingha +4

Synthetic network traffic generation has emerged as a promising alternative for various data-driven applications in the networking domain. It enables the creation of synthetic data…

cs.CR2026

SHIELD: An Auto-Healing Agentic Defense Framework for LLM Resource Exhaustion Attacks

Nirhoshan Sivaroopan, Kanchana Thilakarathna, Albert Zomaya +6

Sponge attacks increasingly threaten LLM systems by inducing excessive computation and DoS. Existing defenses either rely on statistical filters that fail on semantically meaningfu…

cs.CR2026

Prompt-Induced Over-Generation as Denial-of-Service: A Black-Box Attack-Side Benchmark

Manu, Yi Guo, Kanchana Thilakarathna +5

Large Language Models (LLMs) can be driven into over-generation, emitting thousands of tokens before producing an end-of-sequence (EOS) token. This degrades answer quality, inflate…