Application of Embedding Method for both the Study of Binary Codes

Sudhanshu Pandey

Abstract

Binary code analysis is essential for analyzing programmers despite access to the actual source code specific to valuable place. The study of binaries may be difficult due to their wide variations: due to the proliferation of tech developers, the source code is now often optimized for several instruction set architectures (ISAs); although there is no systematic vocabulary that distinguishes among their assembly languages. The complexity of detection has been further exacerbated by a number of developer enhancements and obscured application identifiers. Such minutiae ensures that certain bugs may only be found at a fine-grained stage. Recent steps in deep learning in Natural Language Processing (NLP)—may provide a solution: deep learning methods may process large texts and encode the semantics of individual words into vectors called word embedding that are useful for data visualization and interpretation. By considering assemblies as a phrase and directions as phrases, we use NLP ideas to produce individual instruction embedding’s. Explicitly, we choose to build on existing models that are single-architecture, or that suffer from performance problems while managing several architectures. This research proposes a cross-architectural instruction embedding model that co-encodes instruction semantics from different ISAs where identical instructions are tightly combined within and across architectures. Results demonstrate that our model is effective in removing semantics from binaries on its own and our embedding catch semantic equivalences through multiple architectures. When combined, these instruction embedding can reflect the sense of functions or simple blocks; thus this model may prove useful for cross-architectural bug, ransomware, and plagiarism detection.

Application of Embedding Method for both the Study of Binary Codes

Abstract

Read more of it

Publish your article

Abstract

Related articles

Brain-Inspired Computing: The Rise of Neuromorphic AI

An Energy-Efficient CMOS Analog Front-End for Wearable Biomedical Monitoring Applications

Design and Implementation of 32-Bit High Speed Vedic Multiplier

IMPLEMENTATION OF VOTING MACHINE TO AVOID RIGGING USING VERILOG HDL

Microelectronics: Reach, Functions, and Worldwide Effects

Design of Real-Time Candy Vending Machine Using Verilog HDL

Read more of it

Publish your article