Industrial-scale reverse engineering compromises valuable knowledge embodied in industrial software, as cloning is often cheaper than independent development. This dissertation introduces a conceptual model for hardware–software binding that ties intended software execution to intrinsic, unique, and unclonable hardware properties, aiming to hinder the copying and reverse engineering of software. Software is transformed at design and compile time such that its intended form is restored only at run time using hardware properties that act as a hardware fingerprint. As a result, protected software exhibits its intended behavior only on the original device and unintended behavior on cloned or unauthorized hardware. The core contribution comprises two protection mechanisms based on Rowhammer, a hardware fault in DRAM that enables hardware fingerprinting. The first approach, GlueZilla, binds program instructions to specific DRAM chips and requires device-dependent restoration of intended execution at run time. Angluerus complements this mechanism by binding program data. Both approaches employ reliable and device-unique Rowhammer-induced bit flips to dynamically restore intended program behavior during execution. Complementing these mechanisms, the dissertation presents three additional approaches that address restricted yet common industrial scenarios using other hardware fingerprints. Minotaurox derives control-state transitions from a hardware fingerprint to bind program control flow, ensuring safe transitions under an incorrect fingerprint. Sphinxor protects sensitive data by decoding it only when evaluated with the correct hardware fingerprint and otherwise yields safe alternatives. Locki extends the concept to neural network models by decrypting their weights at run time using a hardware-derived key. Comprehensive security analyses and empirical evaluations across several platforms demonstrate the practical extraction of reliable hardware fingerprints without requiring additional hardware and show that the proposed protection mechanisms render reverse engineering uneconomical with acceptable overhead. Overall, this dissertation establishes hardware–software binding as a practical and versatile protection primitive for industrial software, substantially raising the barrier for unauthorized copying and reverse engineering.     «

Industrial-scale reverse engineering compromises valuable knowledge embodied in industrial software, as cloning is often cheaper than independent development. This dissertation introduces a conceptual model for hardware–software binding that ties intended software execution to intrinsic, unique, and unclonable hardware properties, aiming to hinder the copying and reverse engineering of software. Software is transformed at design and compile time such that its intended form is restored only at ru...     »