In this paper, we propose a lightweight trust management system (TMS) for fog-enabled cyber physical systems (Fog-CPS). Trust computation is based on multi-factor and multi-dimensional parameters, and formulated as a statistical regression problem which is solved by employing random forest regression model. Additionally, as the Fog-CPS systems could be deployed in open and unprotected environments, the CPS devices and fog nodes are vulnerable to numerous attacks namely, collusion, self-promotion, badmouthing, ballot-stuffing, and opportunistic service. The compromised entities can impact the accuracy of trust computation model by increasing/decreasing the trust of other nodes. These challenges are addressed by designing a generic trust credibility model which can countermeasures the compromise of both CPS devices and fog nodes. The credibility of each newly computed trust value is evaluated and subsequently adjusted by correlating it with a standard deviation threshold. The standard deviation is quantified by computing the trust in two configurations of hostile environments and subsequently comparing it with the trust value in a legitimate/normal environment. Our results demonstrate that credibility model successfully countermeasures the malicious behaviour of all Fog-CPS entities. The multi-factor trust assessment and credibility evaluation enable accurate and precise trust computation and guarantee a dependable Fog-CPS system.