We propose a novel end-to-end privacy-preserving framework, instantiated by three efficient protocols for different deployment scenarios, covering both input and output privacy, for the vertically split scenario in federated learning (FL), where features are split across clients and labels are not shared by all parties. We do so by distributing the role of the aggregator in FL into multiple servers and having them run secure multiparty computation (MPC) protocols to perform model and feature aggregation and apply differential privacy (DP) to the final released model. While a naive solution would have the clients delegating the entirety of training to run in MPC between the servers, our optimized solution, which supports purely global and also global-local models updates with privacy-preserving, drastically reduces the amount of computation and communication performed using multiparty computation. The experimental results also show the effectiveness of our protocols.

Secure RAM computation allows a number of parties to evaluate a function represented as a random-access machine (RAM) program in a way that reveals nothing about the private inputs of the parties except from what is already revealed by the function output itself. In this work we present Ramen, which is a new protocol for computing RAM programs securely among three parties, tolerating up to one passive corruption. Ramen provides reasonable asymptotic guarantees and is concretely efficient at the same time. We have implemented our protocol and provide extensive benchmarks for various settings.

Permanent memory, and biases that humans have such as confirmation bias, are not modeled in some of the leading strategies in Iterated Prisoner's Dilemma. This reduces their effectiveness at modeling actual human behaviors. As a solution to this problem we have proposed a framework to model egoistic agents using trust. This framework accounts for such flaws in human behavior, and provides a more accurate representation of human nature by modeling confirmation bias through egoism and trust. The agents of this model divide their neighboring agents into two sets, I and Non-I, and the actions taken against a neighbor are influenced by the set it belongs to. This framework can be used to modify existing strategies such as Tit-For-Tat to make them more realistic. Using this framework, we have also proposed a trust based strategy and evaluated it. Our strategy does better than the other strategies in pairwise comparisons, and also when there is a combination of different agents in the environment.