Cyberfake news is not a single phenomenon which is causing multiple and complex effects on the society, politics and personality. Whereas social media promote misinformation through engagement-bias algorithms, the speed at which false information is processed and disseminated weakens trust, democracy and decision-making. It goes from discrediting efforts in public health emergencies, such as COVID-19 to swaying voters and deepening social cleavages. However, some approaches of automated fake news detection have been developed, current methodologies are limited by critical issues. Previous studies mainly focus on the text analysis while the cases of multimedia misinformation including deep fake are increasingly observed. In addition, the cross-lingual identification of fake news has not received sufficient attention even though cross-lingual approaches to numerous problems are still in their infancy. The other problem recognized throughout the survey is the real-time detection, as most approaches are based on computationally complex processes that cannot be efficiently implemented in high-risk environments. This work points to such gaps and calls for efficient, smart, and robust methods of detection that factor in multimodal approaches, multilingualism, and the use of lightweight algorithms. Thus, overcoming these limitations and fostering the approach argued in this paper, the research contributes to the improvement of the fake news battle and the enhancement of public discourse credibility.