The invention provides a data fusion tamper-proofing method based on input data compression representation correlation analysis, and the method comprises a network node data feature compression mechanism which compresses original input data input to represent data features, guarantees the data representation precision, and shields the difference of different data types. The method comprises a sub-data-set collection method based on the Boost principle, and the method carries out the reasonable division of a data set with the probability of putting back, and forms an input data vector of a training sub-classifier; the method comprises a sub-classifier based on Adaboost and a tamper detection model construction method. The sub-classifiers are trained by utilizing the divided sub-data sets, finally an AdaBoost classifier is constructed by combining the sub-classifiers to predict a related network node data vector type, a final prediction result is obtained by applying a majority voting principle, a tampered data fusion result is quickly identified by utilizing the mechanism, and a malicious data fusion operation is prevented.