With the wide application of Internet of Things (IoT) devices, security attacks against their firmware often occur, which has attracted more attention from the research community. Firmware is an important part of IoT devices, and attacks against them is one of the main means to destroy IoT devices. Therefore, firmware security is the core of the overall security of devices. At present, most of the firmware vulnerabilities have a small number of related samples, so it is difficult to use machine learning methods to generate detectors for some specific vulnerabilities. Therefore, based on the collected data of related firmware vulnerabilities, this paper proposes a firmware vulnerability homology detection method based on the clonal selection algorithm. We design the numerical and structural characteristics of vulnerability functions, train a detector for each function separately, and improve the recall rate of vulnerability detection. Compared with existing machine learning methods, this method only depends on the affinity between the objective function and the detector, which avoids the requirement of a large number of sample data sets. Finally, relevant experiments are carried out to verify the effectiveness of the method.
- clonal selection,
- detectors,
- feature extraction,
- firmware,
- homology,
- Internet of Things,
- Internet of Things,
- Machine learning,
- security,
- semantics,
- simulated annealing
IR deposit conditions: