• crowdsensing
• cybersecurity
• data poisoning attack
• defense

The prevalence of smart devices like smartphones, tablets, and smartwatches has increased the number of sensors placed within them, such as GPS, accelerometers, and cameras. This makes them an excellent source of information for Mobile CrowdSensing (MCS), a new way to collect data from the physical world quickly and efficiently. In MCS systems, individuals only need to carry the devices responsible for collecting the data relevant to the CrowdSensing application. MCS is a powerful technique for data collection but it also poses several security challenges. The openness and inclusivity of crowdsensing systems make them susceptible to adversarial manipulation. One insidious form of attack is data poisoning where malicious entities inject deceptive or manipulated data into the crowd-sensed information. This is done by creating or recruiting a group of malicious workers who then submit malicious data. The potential consequences of such attacks are far-reaching and can impact decision-making processes and compromise the reliability of the collected data. The dynamic and decentralized nature of MCS environments poses unique challenges in detecting and mitigating data poisoning attacks. The primary objectives of this thesis are: assess the impact of data poisoning attacks on the accuracy, reliability, and trustworthiness of crowd-sensed data; implement and evaluate novel defense mechanisms designed to detect and mitigate data poisoning attacks in real-world CrowdSensing scenarios.