Recognition of electro-magnetic leakage information from computer radiation with SVM

This paper focuses on the far-field reception of electromagnetic (EM) radiation and the recognition of letters recovered from the EM leakage. EM radiation captured by a wideband antenna is strengthened by a pre-manipulation system with amplifiers and filters, and the useful information is extracted. After being recovered from the EM radiation with signal processing method, the text image is further recognized by support vector machine (SVM) algorithm. Here, a two-layered SVM network with 60 SVMs is constructed to recognize the letters. In the process, the text image is cut apart into single letters to meet the input requirement of letter-based SVMs. To handle those exceptions of stroke connections, an interactive method and an automatic method are proposed. The received text image is usually obtained in low resolution with characteristics of large stroke distortion, font variation and variable size. In our two applications, however, the recognition accuracy reached 99.2% for the larger font size texts and 96.4% for the smaller size texts. From this, we may draw a conclusion that the proposed SVMs network works well in recognizing textual information and emphasize the potential risk of information leakage for computer system.