The crash risk assessment of mixed traffic has been understudied due to a lack of pertinent data. In recent years, proactive methods have gained significant attention in transportation safety analysis because of their numerous advantages. In this study, we model and evaluate the effect of speed difference on the sideswipe crash risk of mixed traffic using a novel proactive safety indicator called Anticipated Collision Time (ACT). For analysis, detailed trajectory data were collected using an unmanned aerial vehicle from four-lane and six-lane rural highways. The crash risk was derived from the observed conflict risk and then used to assess the highway system's safety performance under study. Using the Extreme Value Theory (EVT), the conflict risk was mapped to the crash risk. The extreme events were identified using the Block Maxima (BM) approach. Later, the Generalized Extreme Value Distribution (GEV) models were developed for each location by extracting the sideswipe conflicts from the vehicle trajectories. The findings suggest that sideswipe conflicts are more of a safety concern than rear-end conflicts due to the frequent nature of lane changes or passing operations. Also, significant speed difference exists among different vehicle modes present in mixed traffic, and the sideswipe crash risk increases with the increase in the maximum speed difference. The analyses of speed differences also indicate that compared to the four-lane highway, the available safety margin is limited on a six-lane highway due to the higher maximum speed difference. Hence, any driver error may lead to sideswipe crashes. In light of the findings of this study, we strongly suggest the need to provide speed control strategies and restrict the frequent risky lane-changing or passing operations that mainly give rise to sideswipe conflicts on a six-lane highway. Further, this study found that the sideswipe crash risk decreases with the increase in vehicle size on both four-lane and six-lane highways. Hence, we suggest developing separate crash risk models for different vehicle modes in mixed traffic on multilane rural highways. Otherwise, the crash risk mitigation strategies may not be appropriate for mixed traffic conditions. © 2023 Elsevier Ltd