Security analysis of blockchain technology is an active domain of research. There has been both cryptographic and game-theoretic security analysis of Proof-of-Work (PoW) blockchains. Prominent work includes the cryptographic security analysis under the Universal Composable framework and Game-theoretic security analysis using Rational Protocol Design. These security analysis models rely on stricter assumptions that might not hold. In this paper, we analyze the security of PoW blockchain protocols. We first show how assumptions made by previous models need not be valid in reality, which attackers can exploit to launch attacks that these models fail to capture. These include Difficulty Alternating Attack, under which forking is possible for an adversary with less than 0.5 mining power, Quick-Fork Attack, a general bound on selfish mining attack and transaction withholding attack. Following this, we argue why previous models for security analysis fail to capture these attacks and propose a more practical framework for security analysis pRPD. We then propose a framework to build PoW blockchains PRAGTHOS, which is secure from the attacks mentioned above. Finally, we argue that PoW blockchains complying with the PRAGTHOS framework are secure against a computationally bounded adversary under certain conditions on the reward scheme.