How does Kaspa's GHOSTDAG protocol prevent attackers from rewriting old transaction history?

Kaspa's GHOSTDAG protocol includes a mathematical guarantee called the freeloading bound that severely limits how much an attacker can benefit from honest network work when trying to rewrite old history. In most proof-of-work systems, an attacker trying to reorganize the chain can piggyback on blocks produced by honest miners to make their attack cheaper — this is called freeloading. GHOSTDAG's freeloading bound (proven as Lemma 12 in the research paper) caps the number of honest blocks an attacker can exploit at a constant value of 4k blocks, meaning that beyond a short window, freeloading stops providing any real advantage. For a beginner, this means that the older a Kaspa transaction is, the harder it becomes to reverse, because the math works against any attacker trying to undo it.