Cancer cells have abnormal gene expression profiles; however, to what degree these are chaotic or driven by structured gene regulatory networks is often not known. Here we studied a model of Ras-driven invasive tumorigenesis in Drosophila epithelial tissues and combined in vivo genetics with next-generation sequencing and computational modeling to decipher the regulatory logic of tumor cells. Surprisingly, we discovered that the bulk of the tumor-specific gene expression is controlled by an ectopic network of a few transcription factors that are overexpressed and/or hyperactivated in tumor cells. These factors are Stat, AP-1, the bHLH proteins Myc and AP-4, the nuclear hormone receptor Ftz-f1, the nuclear receptor coactivator Taiman/SRC3, and Mef2. Notably, many of these transcription factors also are hyperactivated in human tumors. Bioinformatic analysis predicted that these factors directly regulate the majority of the tumor-specific gene expression, that they are interconnected by extensive cross-regulation, and that they show a high degree of co-regulation of target genes. Indeed, the factors of this network were required in multiple epithelia for tumor growth and invasiveness, and knockdown of several factors caused a reversion of the tumor-specific expression profile but had no observable effect on normal tissues. We further found that the Hippo pathway effector Yorkie was strongly activated in tumor cells and initiated cellular reprogramming by activating several transcription factors of this network. Thus, modeling regulatory networks identified an ectopic and ordered network of master regulators that control a large part of tumor cell-specific gene expression.