Cytokines are essential components of the immune system and are recognized as significant biomarkers. However, detection of a single cytokine is not precise and reliable enough to satisfy the requirements for diagnosis. Herein, we developed a pattern recognition-based method for the multiplexed sensing of cytokines, which involves three-color-emitting boronic acid-decorated carbon dots (BCDs) and arginine-modified titanium carbide (Ti3C2 MXenes) as the sensor array. Initially, the fluorescence signals of the three BCDs were quenched by Ti3C2 MXenes. In the presence of cytokines, the fluorescence intensity of the BCDs was restored or further quenched by different cytokines. The fluorescence response occurs in two steps: first, boronic acid interacts with cis-diol functional groups of cytokines, and second, arginine headgroup selectively interacts with glycans. By exploiting the different competing binding of the BCDs and the cytokines toward Ti3C2 MXenes, seven cytokines and their mixtures can be effectively discriminated at a concentration of 20 ng mL-1. Furthermore, our sensor array demonstrated an excellent performance in classifying human oral cancer saliva samples from healthy individuals with clinically relevant specificity. The noninvasive method offers a rapid approach to cytokine analysis, benefiting early and timely clinical diagnosis and treatment.