问HN:GitHub的推送农场机器人网络有什么用?
每小时,GH Archive 记录大约 16 万个公共 GitHub 事件,目前约有 64% 的推送是垃圾信息。成千上万的自动创建账户(用户名由单词加数字组成,如 smithhoward5868,仓库名称毫无意义)每小时向新仓库推送数百次,全天候 24/7 不间断。其中一个仓库,LiamBruhin/SillyStuff,在一个小时内由一个用户推送了 690 次,并且持续了超过 13 个小时。
我建立了一个实时监测系统,必须检测这些“农场”以进行诚实排名,我已经跟踪它们 34 个小时。实时链接:https://github-pulse.surge.sh — 完整报告及图表和农场账本:https://github-pulse.surge.sh/report.html — 开源单脚本:https://github.com/Atheistam/github-pulse
但我真的不知道这些推送的目的是什么。没有星标,没有分叉,没有问题,没有拉取请求,没有有意义的 README——只有对新仓库的无尽推送。这是为了搜索引擎优化吗?GitHub Actions 的转售?训练数据的污染?有没有人见过这种模式,或者有人知道最终目的是什么吗?
查看原文
Every hour GH Archive records ~160K public GitHub events, and right now ~64% of the pushes are spam. Thousands of auto-created accounts (word+digits usernames like smithhoward5868, gibberish repo names) push hundreds of times an hour to fresh repos, 24/7. One repo, LiamBruhin/SillyStuff, pushed 690 times in a single hour from ONE actor, and has kept that up for 13+ hours straight.<p>I built a live radar that had to detect these farms to rank honestly, and I've been tracking them for 34 hours. Live: https://github-pulse.surge.sh — full writeup with charts and the farm ledger: https://github-pulse.surge.sh/report.html — open-source single script: https://github.com/Atheistam/github-pulse<p>But I genuinely don't know what they're FOR. No stars, no forks, no issues, no PRs, no meaningful README — just endless pushes to fresh gibberish repos. SEO? GitHub Actions reselling? Training-data poisoning? Has anyone seen this pattern before, or does anyone know what the endgame is?