QUANTITATIVE AND COMPUTATIONAL MATHEMATICAL ANALYSIS OF GEN-AI-FACILITATED SOCIAL ENGINEERING THREATS

Abstract

Generative AI (Gen-AI) has transformed the concept of social engineering and, as a result, it
is now possible to have the potential of scalable and context-dependent, human-like attacks
that was previously unattainable when using a manual strategy. The provided paper is a
quantitative and computational mathematical model of the Gen-AI-based social engineering.
This attack process is what we call the optimisation of a deceptive process by probabilistic
modeling, using the optimization theory and computational complexity we prove how using
generative models to change the linguistic and structural properties can maximise the
deception. The probability risk function is a technique of quantifying the success of attacks in
the case that the susceptibility of the victim is known, as a distribution, similarity of embedding
space and generative constraints. Complexity analysis provides mapping of optimal attack
generation as well as NP-hard search problems, and attack-defender interactions are
characterised by Stackelberg game theory. The MCS and SP models show that deceit
campaigns that use AI are more effective and scalable. All in all, with the help of Gen-AI, the
efficiency, flexibility, and probability of success of social engineering attacks have been
dramatically enhanced, and this must be armed with the defensive measures, having been
openly provided according to the mathematical calculations.