I copied the entire list of IP blocklists and pasted it into my favorite AI (which i have configured my own custom settings that I wrote to improve its accuracy and asked it to generate a table with one line for each blocklist, and a column for aggressiveness, false positives, typical use case, and recommendations. It went out and collected info for each and I used the table to update my configuration of which ones i enabled. Here is a link to an Excel spreadsheet in case anybody wants to download and save a copy. It has a tab with the blocklists analysis, and a tab with my custom instructions. My custom instructions include rules for English grammar extracted and condensed from a PDF for a popular American English grammar book, The Elements of Style. That might be helpful for all of the members that aren’t fluent in English. Nethsecurity Threat Shield IP Blocklist recommendations.xlsx
Its actually a pretty useful way to compare blocklists. but i would suggest just be careful about trusting the AI recommendations blindly, especially for false positive rates. Also if you do so test the more aggressive lists for a while and check the logs before enabling too many at once.
I agree completely. I’m particularly sensitive to this because my career in computer science included being exposed to AI back in the 1970s, and I’m following the AI developments pretty closely. If you look at the second tab on my spreadsheet, I have specific custom instructions which help accuracy. I also find that Grok, although far from perfect, is less susceptible to hallucinations than the others. On the tab with the block lists, you can see that Grok opined on the point that you make about the more aggressive block lists, The ones I highlighted in yellow are the ones I’m currently using. And I have a pretty large list of white listed domains because of the false positives.