We present a quantum image encryption protocol that harnesses discrete-time quantum walks (DTQWs) on cycles and explicitly examines the role of the Parrondo paradox in security. Using the NEQR representation, a DTQW-generated probability mask is transformed into a quantum key image and applied via CNOT to encrypt grayscale images. We adopt an efficient circuit realization of DTQWs based on QFT-diagonalization and coin-conditioned phase layers, yielding low depth for N=2^n positions and t steps. On 64x64 benchmark images, the scheme suppresses adjacent-pixel correlations to near zero after encryption, produces nearly uniform histograms, and achieves high ciphertext entropy close to the 8-bit ideal. Differential analyses further indicate strong diffusion and confusion: NPCR exceeds 94% and UACI is around 30%, consistent with robust sensitivity to small plaintext changes. Crucially, we identify parameter regimes in which alternating coin operations induce the Parrondo paradox and degrade security-raising correlations, lowering entropy, and reducing NPCR/UACI, thereby constituting practical failure modes. Our results provide both a performant DTQW-based quantum image cipher and clear guidance on coin/message parameter selection to avoid paradox-dominated regimes. We discuss implications for hardware implementations and extensions to higher-dimensional walks.