Convolutional Neural Network-Based Cryptography Ransomware Detection for Low-End Embedded Processors

摘要: A crypto-ransomware has the process to encrypt victim’s files. Afterward, requests a ransom for password of encrypted files victims. In this paper, we present novel approach prevent by detecting block cipher algorithms Internet Things (IoT) platforms. We extract sequence and frequency characteristics from opcode binary 8-bit Alf Vegard’s RISC (AVR) processor microcontroller. other words, late fusion method is used two features one source data, learn through each network, integrate them. classify virus or harmless software proposed method. The general AVR packages implementations written in C language lightweight library (i.e., Fair Evaluation Lightweight Cryptographic Systems (FELICS)) are trained deep learning network evaluated. successfully classified training functions Furthermore, detect codes that file using ciphers. detection rate evaluated terms F-measure, which harmonic mean precision recall. not only achieved 97% success but also 80% classification cryptographic algorithm benign firmware. addition, Substitution-Permutation-Network (SPN) structure, Addition-Rotation-eXclusive-or structures (ARX) firmware 95%.