read.cash Log in
@liwi more from that month

@poledancertw 在前些篇 https://noise.cash/post/12g8mgxm 裡 提到 Helium 網路覆蓋率證明（Proof-of-Coverage ）的作弊問題 於此，之前瀏覽到這篇用圖神經網路去分析 HNT 網路的網路文章 Exploring the Helium Network with Graph Theory Using blockchain data to extract insights about network coverage and “suspicious” hotspots https://towardsdatascience.com/exploring-the-helium-network-with-graph-theory-66cbb8bffff9 還蠻有趣的。文長，而且很多地方我也沒仔細去看懂（也說不定本來就看不懂）。如果你剛好是（資料）科學家，說不定可以秒讀懂。 文章裡稍微碰觸了作弊問題（spoofing，電子欺騙）： - In the Helium Network, one of the premiere challenges for the engineering team is to identify bad actors who may be spoofing the system to increase their rewards. - This gaming can take many forms, but one of the most popular approaches is to make hotspots appear more spread out then they really are. - It’s possible there is some spoofing happening here, or maybe this is just an artifact of how well radio waves travel over water. In future models, I would like to account for geographic factors, like local elevation and significant bodies of water. - Another phenomena that we see here is some insanely far witness paths, on the order of hundreds of kilometers. Perhaps these hotspots just have extremely powerful antennas and great locations, or maybe this is a sign of suspicious activity. - Using these features and state-of-the-art Graph Neural Network-based models, we can start to tackle the formidable task of identifying “suspicious” activity on the Helium blockchain. 胡亂抓結論的話大概是：作弊有沒有跡象？有，但在判為有作弊嫌疑之時也很難完全排除其他外在條件。之前在這邊的討論也有高手提到了，RF 信號有時候就是很弔詭的。HNT / LoRa 用的應該也不是什麼先進技術，不然就我可以想見，應該有很多多頻或調變技術可以在硬體面用來偵測、防止 PoC 作弊。 另外用圖論跟神經網路來找大概是殺雞用牛刀了。就 @poledancertw 所描述，一堆作弊機幾乎不掩飾的。 真正的難題是如何在維持去中心化 people's network 的同時解決 PoC 問題

15 comments

Log in to join in Reading is open to everyone. Replying needs an account.