In view of the fact that the strength of wireless signals in complex indoor environments has anomalies due to the influence of the sudden noise in the propagation environment, an algorithm for WiFi fingerprinting localization based on online coarse positioning and random subset sampling was proposed. First，the algorithm utilized the specific coverage characteristics of access point (AP) signals to rapidly and accurately constrain the target's position to a local region, thereby identifying candidate reference points (RP) near the true location. Subsequently, random subset sampling was employed to generate multiple subsets of signal distances, which were then averaged to mitigate the influence of abnormal AP signals on distance measurements. Finally, a weighted K-nearest neighbors algorithm was applied to estimate the target's position. Experiments were conducted using a publicly available dataset collected from real indoor environments to evaluate the accuracy of both online caorse positioning and random subset sampling. The results demonstrate the effectiveness of the proposed algorithm and reveal the impact of parameter selection on localization performance. Comparative studies with existing fingerprint-based localization algorithms show that the online coarse localization method effectively identifies candidate reference points near the user's true position, while the random subset sampling technique significantly reduces large localization errors. Even when AP signals exhibit abnormalities, the proposed algorithm maintains superior positioning accuracy compared to conventional algorithms.