This research proposes an image encryption scheme based on two different types of chaotic oscillators which are autonomous and non-autonomous chaotic oscillators. While both chaotic oscillators can work perfectly as random number generators, the non-autonomous oscillator might be subject to a clock injection attack since it depends on external input. The research highlights the risk of using non-autonomous chaotic oscillators as a source of entropy to construct the encryption key. The non-autonomous chaotic oscillator is implemented on FPGA. We inject clock glitches to affect the randomness of the system as a robustness test of the system. We attack 5 possible positions of the oscillator. We conduct NIST randomness test on those 5 streams to allocate the best position to attack. In our case using clocked van der pol oscillator, position 5 passed the least amount of NIST tests, which makes it best position to attack the system among others. The encryption scheme performance metrics for the ciphered image includes Histogram uniformity, Key space analysis, key sensitivity, test analysis of correlation. The encryption scheme is effective in small and large images. The thesis examines 6 different sizes of images in regard to the computational time, correlation between adjacent pixels, and the entropy of the ciphered image. In an effort to reduce the computational time we create a dynamic loop to analyze the number of iterations required to perform the shuffling phase. The results enhanced our initial computational time up to 10 times.