The file fragment classification problem : a combined neural network and linear programming discriminant model approach

摘要: The increased use of digital media to store legal, as well illegal data, has created the need for specialized tools that can monitor, control and even recover this data. An important task in computer forensics security is identify true file type which a or fragment belongs. File identification traditionally done by means metadata, such extensions header footer signatures. As result, traditional metadata-based object techniques work cases where required metadata available unaltered. However, approaches are not reliable when integrity guaranteed unavailable. an alternative, any pattern content be used determine associated type. This called content-based identification. Supervised learning infer classifier exploiting some unique underlies type’s common structure. study builds on existing literature regarding supervised identification, explores combined multilayer perceptron neural network classifiers linear programming-based discriminant solution multiple class problem. purpose was investigate compare single classifier, ensemble these field ability each individual accurately predict belongs were tested empirically. found both (used round robin) seemed perform solving results combining better than those optimized classifiers.