Solidity Security Vulnerabilities: Floating Points and Numerical Precision

Solidity does not support floating point types and as such any such calculation needs to be simulated carefully with uint. This however can produce a vulnerability if not done right.

Lets look at the contract FunWithNumbers from https://github.com/sigp/solidity-security-blog#15-floating-points-and-precision

This line below is converting ETH to some token.

uint tokens = msg.value/weiPerEth*tokensPerEth;

The problem is in the division – if msg.value is less than 1 ETH (10^18 Wei) then the result of division will be rounded off to zero as only integer types are supported. As a result, “tokens” will always rounded down to the nearest integer. So, if msg.value = 0.5 ETH then tokens = 0 and if msg.value = 1.9 ETH then tokens = 1 ETH.