ALMOST: Adversarial Learning to Mitigate Oracle-less ML Attacks via Synthesis Tuning