XGBoost outperforms Transformer models in Ethereum Sybil bot detection

A new research paper published on arXiv evaluates different machine learning models for detecting Sybil bots on the Ethereum blockchain. The study introduces a leakage-aware evaluation framework and a "Transaction Grammar" to represent wallet behavior, aiming to provide more accurate and practical detection methods. Results indicate that XGBoost, a tree-based model, outperforms Transformer-based sequence models like Transformer++ and BiLSTM when label leakage is accounted for, while also offering lower latency and energy consumption. AI

IMPACT This research suggests that simpler, more efficient models like XGBoost may be sufficient for real-time blockchain analytics, potentially reducing computational costs and improving deployment practicality.

RANK_REASON Research paper published on arXiv detailing model evaluation. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• arXiv
• BiLSTM
• Ethereum
• Ethereum Virtual Machine
• support vector machine
• Sybil bots
• Transformer++
• XGBoost

• paper
• model release
• infra

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →