CCTV cameras, often known as surveillance cameras, are among the most sophisticated security systems now available. Even though surveillance cameras (CCTV) are a security tool, it is common for them to be targeted to conceal a crime caught on camera. Video injection is one of the methods used to compromise surveillance cameras (CCTV). Video injection attacks insert live video feeds, resulting in a loss of data integrity that can impede or even alter the absolute truth. This paper employs the ensemble learning approach to recognize video injection attempts on security cameras. Ensemble learning utilized here is random forest and support vector machine (SVM) estimators. The random forest estimator-based model yields a f1-score value of 91% and an accuracy of 93% with a total dataset of 600 data, while the support vector machine (SVM) estimator yields a f1-score value of 84% and an accuracy of 87% with a total dataset of 12000 data. The accuracy of random forest is fairly high and may be used to identify video injection attacks.
Keywords: CCTV, video injection, random forest, support vector machine (SVM)