Highly urbanized regions are faced with serious environmental issues. The construction of ecological security patterns is an important approach to protecting urban ecological security. There are some technical difficulties in identifying ecological source areas in highly urbanized regions, like Shenzhen, the study area of this paper. Reviewing the existing research on recognition methods for important patches, it was found that the commonly used method is relatively simple, based on biodiversity as well as the importance of ecosystem services. However, these methods are biased towards considering the structure and function of the patch itself, while ignoring the importance of patches in the entire matrix structure and human factors. Therefore, this paper attempted to apply a comprehensive recognition method for landscape connectivity and ecological security to recognize the ecological source area. The framework of DPSIR was applied to establish an ecological security assessment index system from three aspects: social, economic, and natural factors. Further, to explore the optimal scale of ecological source areas, a recognition method was established based on multi-scale landscape connectivity. The results are as follows: (1) The ecological security pattern of Shenzhen is excellent in the southeast and northeast, and not good in the central and western regions. The area ratios for different grades of ecological security are 30.34 %, 13.49 %, 9.85 %, 9.15 %, and 37.17 % for grades Ⅴ, Ⅳ, Ⅲ, Ⅱ, and I, respectively. (2) Through principal component analysis of four landscape connectivity indicators, namely, connectance, aggregation, cohesion, and separation, it was found that the landscape connectivity of ecological source areas in Shenzhen appears optimally when 400 m was applied as the landscape component scale and the minimum recognition threshold area of patches was 0.8 km². (3) The total area of the ecological source areas was 753.78 km² based on the high grades of ecological security, accounting for 37.74 % of the total area of Shenzhen. The ecological source areas are mainly composed of forest and water areas. The ecological source areas contained 86.87 % of legally protected areas, and 91.92 % of the ecological source areas fell within the ecological control line. Thus, this recognition method is feasible to some extent. (4) The method of recognition of the source area proposed in this paper provides a certain degree of improvement as it takes not only the natural factors but also the human factors into account. It is demonstrated that the recognition of ecological source areas based on the ecological security pattern and multi-scale landscape connectivity can provide a new approach for the identification of urban ecological sources and further guide urban planning effectively.