Intelligent reflecting surface (IRS) is widely used to enhance physical layer security recently. However, the double-fading attenuation caused by IRS reflection often limits its security performance. The employment of active IRS could improve this situation but incurring more energy consumption. In this correspondence, we consider an active IRS-assisted multiple input single output (MISO) downlink system and aim to maximize the secrecy energy efficiency (SEE) to achieve the suboptimal trade-off between the secrecy rate and energy consumption. To solve this non-convex joint beamforming problem, an alternating optimization algorithm is proposed, which jointly optimizes the reflecting matrix of active IRS and beamforming vector at the base station using the successive convex approximation and dual Lagrange method, respectively. Simulation results show the superior SEE performance of active IRS, especially when the eavesdropping channel is stronger.