Troga fleet care loonie5/11/2023 ![]() More-complex cyberattacks require that the sequence of system states be used as a context for the detection of a cyberattack. There is an opportunity to insert machine learning here because false sensor readings can be learned and generalized if symbolic learning is used. Benchmarks need to include sensory information, which is typically included in the rewind. A car at a spoofed stop sign is not representative of such a system. This approach works well for cyber-physical systems that can be rewound. The use of benchmarks follows from “syntactic randomization” because, in general, code compiled on different compilers will not be all susceptible to the same cyberattack, which is then detectable and recoverable from. Software may be kept “clean” by storing it in ROM and taking care to insure that the ROM cannot be reprogrammed. The idea is to purge the infection and continue processing from the last successful point, while continuing the random benchmark tests. In such cases, one needs to rewind to the last successful benchmark, restore the context, and execute from there on. If any such runs deviate from the expected output, it follows that the system is infected. The classic approach to detecting and preventing cyberattack is to interrupt executing software at random points and run benchmark programs, whose (intermediary) output is known. It similarly follows that complex software is inherently subject to cyberattack. Arbib, A Programming approach to computability. Software, which is complex enough to be capable of self-reference, cannot be proven valid [ 1 A.J.
0 Comments
Leave a Reply. |