Some food for thought: it is relatively easy to profile site visitors and narrow down the IP of the scraper servers. For example:
1) Google analytics. Simple. Just find all incoming referrals and next hop. Kissmanga or their image server / domains might be one of them, assuming they use a in-browser scraper. If you don't want to leverage the googs you can do it yourself with some cookie work.
2) Use timing attacks, track all IPs that do an img GET request in the time window a decoy manga was up then taken down, do this multiple times and filter by common IP until the pool narrows down.
3) Lazy approach is to look at site usage and which visitor has the most transferred material/request, but you'll need to verify the results obviously.
4) Or we can go full counterintel with stenganography and add a unique digital watermark to every image, generated from using the IP address as the seed. Like how physical printers do it, change certain pixels to be tiny bit more darker/lighter that the human eye cannot see. Cross-reference uploaded manga images and narrow down the pool of candidates.
After you get the IP addresses, just block them or pull some obfuscation tricks, for example:
1) randomly give them images from other manga series or non-manga images
2) give an error for every image and change the comments section to a fake one where people are complaining about images not loading so any human intervention is likely just to wait for the images to be fixed - but its never fixed
3) give them the images but randomly drop some
4) automatically splash a huge watermark on every single image
5) or if you're evil enough and know which parser they use you can try to engineer shit like an infinite file stream or RFI attacks or even this: https://en.wikipedia.org/wiki/Zip_bomb
Eh these are just some basic counterops suggestions not really much of an expert.