Abstract. Side-Channel Analysis (SCA) represents a serious threat to the security of millions of smart devices that form part of the so-called Internet of Things (IoT). Choosing the "right" cryptographic primitive for the IoT is a highly challenging task due to the resource constraints of IoT devices and the variety of primitives. An important criterion to assess the suitability of a lightweight cipher with respect to SCA is the amount of leakage available to an adversary. In this paper, we analyze the efficiency of different selection functions that are commonly used in Correlation Power Analysis (CPA) attacks on symmetric primitives. To this end, we attacked implementations of the lightweight block ciphers AES, Fantomas, LBlock, Piccolo, PRINCE, RC5, Simon, and Speck on an 8-bit AVR processor. By exploring the relation between the nonlinearity of the studied selection functions and the measured leakages, we discovered some imperfections when using nonlinearity to quantify the resilience against CPA. Then, we applied these findings in an evaluation of the "intrinsic" CPA-resistance of unprotected implementations of the eight mentioned ciphers. We show that certain implementation aspects can influence the leakage level and try to explain why. Our results shed new light on the resilience of basic operations executed by these ciphers against CPA and help to bridge the gap between theory and practice.