Social Media Perceptions of 51% Attacks on Proof-of-Work Cryptocurrencies: A Natural Language Processing Approach

TL;DR

The study tackles how $51\%$ attacks on proof-of-work cryptocurrencies shape user perceptions on social media. It combines a manually constructed timeline of attacks with Twitter-based sentiment and emotion profiling using lexicon methods (VADER and Text2Emotion) to derive SP and EI/EV metrics across datasets. Key contributions include a comprehensive timeline of $31$ attack events across $20$ coins, a reduced Twitter-focused timeline for $17$ events, and the construction of three datasets per event to compare attack versus non-attack contexts, with open code and data. Findings show attacks trigger predominantly negative sentiment and heightened fear, with cross-coin patterns indicating potential for real-time attack signaling, while highlighting the value of transparent, open-risk analytics for regulation and market risk assessment.

Abstract

This work is the first study on the effects of attacks on cryptocurrencies as expressed in the sentiments and emotions of social media users. Our goals are to design the methodologies for the study including data collection, conduct volumetric and temporal analyses of the data, and profile the sentiments and emotions that emerge from the data. As a first step, we have created a first-of-its-kind comprehensive list of 31 events of 51% attacks on various PoW cryptocurrencies, showing that these events are quite common contrary to the general perception. We have gathered Twitter data on the events as well as benchmark data during normal times for comparison. We have defined parameters for profiling the datasets based on their sentiments and emotions. We have studied the variation of these sentiment and emotion profiles when a cryptocurrency is under attack and the benchmark otherwise, between multiple attack events of the same cryptocurrency, and between different cryptocurrencies. Our results confirm some expected overall behaviour and reactions while providing nuanced insights that may not be obvious or may even be considered surprising. Our code and datasets are publicly accessible.

Figures (13)

• Figure 1: ${\mathtt{SP}}({\mathcal{W}}_i)$: Sentiment profiles of the whole datasets ${\mathcal{W}}_i$
• Figure 2: ${\mathtt{SP}}({\mathcal{A}}_i)$: Sentiment profiles of attack datasets ${\mathcal{A}}_i$
• Figure 3: ${\mathtt{SP}}({\mathcal{B}}_i)$: Sentiment profiles of the benchmark datasets ${\mathcal{B}}_i$
• Figure 4: Heat map of percentages of tweets carrying different emotions and their combinations in ${\mathcal{W}}_i$ datasets
• Figure 5: Heat map of percentages of tweets carrying different emotions and their combinations in ${\mathcal{A}}_i$ datasets

Theorems & Definitions (2)

• Definition 3.1
• Definition 3.2