This paper considers unmanned aerial vehicle (UAV)-carried intelligent reflecting surface (IRS) for secure data collection in wireless sensor networks. An eavesdropper (Eve) lurks within the vicinity of the main receiver (Bob) while several randomly placed sensor nodes beamform collaboratively to the UAV-carried IRS that reflects the signal to the main receiver (Bob). The design objective is to maximise the achievable secrecy rate in the noisy communication channel by jointly optimizing the collaborative beamforming weights of the sensor nodes, the trajectory of the UAV and the reflection coefficients of the IRS elements. By designing the IRS reflection coefficients with and without the knowledge of the eavesdropper's channel, we develop a non-iterative sub-optimal solution for the secrecy rate maximization problem. It has been shown analytically that the UAV flight time and the randomness in the distribution of the sensor nodes, obtained by varying the sensor distribution area, can greatly affect secrecy performance. In addition, the maximum allowable number of IRS elements as well as a bound on the attainable average secrecy rate of the IRS aided noisy communication channel have also been derived. Extensive simulation results demonstrate the superior performance of the proposed algorithms compared to the existing schemes.