This article considers a relay-based underlay cognitive radio vehicular network, where a secondary source vehicle transmits information to a secondary destination vehicle with a decode-and-forward relay in the presence of an eavesdropper. Specifically, the transmitting power of secondary network is determined by the maximum transmitting power of the secondary transmit vehicle and the interference temperature constraint imposed on primary communication. Besides, an adaptive time allocation scheme is presented to improve system secure performance. As considering imperfect channel state information and time selective fading channels caused by vehicles' mobility, closed-form expressions of outage probability and secure outage probability are derived under this Internet of Vehicles scenario. Moreover, the impacts of time allocation factor, channel quality, vehicle relative rate, power constraint and channel estimation on system secrecy performance are demonstrated. Finally, simulation results are provided to validate our analytical findings.