The Double Exposure: AI as Character and Cinematic Lens for Network Security
Conference
Kaur, N, Kaur, G, Prabhakar, US et al. (2026). The Double Exposure: AI as Character and Cinematic Lens for Network Security
. 1599 LNEE 494-506. 10.1007/978-3-032-20318-2_45
Kaur, N, Kaur, G, Prabhakar, US et al. (2026). The Double Exposure: AI as Character and Cinematic Lens for Network Security
. 1599 LNEE 494-506. 10.1007/978-3-032-20318-2_45
Artificial Intelligence (AI) is widely utilized in network security not as an defensive agent but also as an active agent, as it helps to analyse potential threats, launch attacks, adapt changing conditions and even study how humans respond to it. Building upon this metaphor of the double exposure, in this paper, the research studies published in the last 6 years (2015–2025) in the fields of intrusion detection systems (IDS), adversarial machine learning (AML), explainable AI (XAI), and the new security role of large language models (LLMs) will be presented in a systematic manner. Representative literature that we publish includes peer reviewed articles and technical reports to find out what can be learned by AI about attack surfaces (like adversarial examples and prompt injection) and what defence (like behavioural analytics, triage with XAI) is least vulnerable to adversarial examples. The review identifies five focal areas—DL-IDS, AML, XAI, LLMs, and organizational/policy implications—supported by trend analyses and summary figures. To sum up our final recommendations emphasize the need to focus on real-world AML testing, XAI for high-stakes triage, and architectures resilient to generative adversaries.
