The goal of this study is to propose a methodology for continuous implicit authentication of smartphone users, using the navigation data, in order to improve security and ensure the privacy of sensitive personal data. Privacy and security are 2 interrelated concepts. Privacy refers to the right that the user has regarding the control of his/her personal information and how they are used. Security refers to the way personal data is protected from any unauthorised third-party access or malicious attacks. Smartphones contain a wealth of personal data such as photos, chats, medical data, bank details, personal passwords and information related a person's close circle (contacts, work, hobbies, activities). It is of vital importance to protect the above information from third parties. Protecting personal data using pin codes or biometrics is not always enough. In case the device is stolen or lost, the attacker can bypass the security code in many ways. The violation of biometric authentication, such as face recognition or fingerprint, is difficult but not impossible. The solution to this problem is achieved through continuous implicit authentication. The system processes the user’s behaviour it collects from the sensors, as a background process. If the behaviour does not belong to the owner of the device, the device is locked. This behaviour protects the data and the device. Each user's behaviour is unique. Subsequently, the device remains locked and personal data is protected until the correct behaviour is recognised. Within the context of this study, the accelerometer and gyroscope sensors were selected to model the way a user interacts with its smartphone. The measurements were collected in uncontrolled environment from an application downloaded from the Store. Two machine learning models were trained, one for each sensor and then, the results were combined to produce the final system’s performance. The performance of the final system exceeded the performance of the literature. The One Class SVM algorithm in its best experiment achieved FAR equal to 1.1% and FRR equal to 5.7%, while the Local Outlier Factor algorithm in its best experiments achieved FAR equal to 0.7%, FRR equal to 8.1% and FAR equal to 2.9% with FRR equal to 5%. The proposed system achieved the best percentage of metric FAR compared to other studies, while the metric FRR had one of the best percentages. The results show that the proposed approach provides an additional level of security and privacy and can ensure that 99% of unauthorised users will be denied access to the device and users personal data.