Cloud Service Providers (CSPs) have recently significantly improved, allowing for outsourcing Machine Learning (ML) training and inference. However, due to the data privacy needs in most of the ML applications, several privacy-preserving technologies, such as Multi-party Computation (MPC), have been proposed to protect the data privacy. MPC offers splitting and exchanging of data among multiple parties, typically managed on cloud environments. Although MPC performs better than other alternatives, it still lags behind regular clear-text ML processing in terms of performance. To reduce the execution time of Privacy-Preserving ML (PPML) via MPC, parallelization of computation and communication among the parties (i.e., pipelined MPC), can be employed. However, the complex nature of these systems makes it challenging to select an optimal network and node configuration for executing a pipelined MPC. To address these challenges, in this paper, we propose a Multi-Objective Optimization (MOO) model focusing on achieving optimal configuration to minimize the MPC execution time along with its costs. We formulate an optimization model and propose two distinct approaches to solve it. Our evaluation clearly demonstrates a reduction in execution time and cost with respect to regular MPC execution. The evaluation results also provide valuable insights into the impact of latency and bandwidth considerations on our system's performance, contributing to PPML optimization.

While Federated Learning (FL) enables training by only sharing model updates rather than data, FL can still be prone to privacy leaks. Therefore, many efforts have been made to adopt homomorphic encryption or differential privacy approaches to prevent this. However, these solutions come with several issues that may limit their widespread adoption in applications that involve sensitive data sitting in silos. Such issues include but are not limited to trust in the aggregation server, the accuracy of the model, potential collusion among clients, and limited aggregation function support. To address these issues, we advocate using secure Multiparty Computation (MPC) to offer privacy-preserving computation. Specifically, we propose an FL framework that enables outsourcing the model aggregation to MPC parties on untrusted cloud environments and offers correctness verification to the model owners. Unlike differential privacy-based solutions, the proposed framework offers the same level of accuracy as models that are trained on the clear and minimize the possibility of collusion among clients and MPC parties. We implemented and evaluated the proposed framework under various conditions. The results showed that our framework can match the accuracy of centralized FL training while maintaining the required level of privacy and security in malicious cross-silo settings.

As drones become more widely available, they find applications in many different fields. One of the most promising applications of drones is to use them in deliveries of items/food within certain distances to offer quick service for the customers. In such cases, we will see many different companies deploying drone ducking stations within a neighborhood and fly drones frequently during the day. However, as more companies get into this domain, this will increase the potential for collisions as several drones will simultaneously fly to destinations that are close to each other. Therefore, there is a need to coordinate their trajectory planning in advance by sharing information about their trajectories and destinations. Nevertheless, since drones belong to different companies sharing this information may violate the privacy of their customers and also expose their business privacy. The sharing needs to be done in a privacy-preserving manner just in time so that collisions can be avoided. In this paper, we propose a secure multi-party computation based trajectory planning among the drones. Specifically, a drone shares information with others via wireless communication within their range and performs local computation to check if there is any potential for collisions using Shamir's secret sharing. If there is, the trajectory can be modified (e.g., by changing the altitude of the drone). As part of the computation, we propose an approach which enables comparison of the trajectories by representing them as matrices and performing addition on these matrices since comparison operation is challenging to achieve in Shamir's secret sharing. We implemented a preliminary prototype of this approach using Raspberry PI devices. We demonstrated the feasibility and overhead of the proposed approach under a variety of conditions.

Techniques to perform computations without disclosing the input values have notably improved in the last decade. One such technology, called Secure Multiparty Computation (MPC), where two or more computation nodes hold secret pieces of private data and jointly execute a protocol to obtain a function output, has proven effective for preserving privacy in many applications (e.g., distributed signing, financial scores, machine learning, and more). Nonetheless, in many cases, the data source and computation nodes are often assumed to be the same, with the existence of a manually preconfigured network before they start the computation. This challenge is typical of many IoT applications where the IoT devices need to collaborate using MPC but do not have the resources, and thus outsource the tasks to powerful MPC nodes. Nonetheless, in such a scenario, the IoT devices do not know the MPC nodes, and vice-versa to manage the overall process. To fill this gap, we propose an MPC management protocol that automates the registration and authentication of a group of clients (i.e., sources and consumers of data) and MPC servers (the private computation providers), the requesting of MPC jobs, and receiving the results thereafter. Our experiments over a cloud environment demonstrate the first protocol that efficiently and securely automates the management of MPC systems on many use cases, which would otherwise take considerable time and effort.

Although recent performance improvements to Secure Multiparty Computation (SMPC) made it a practical solution for complex applications such as privacy-preserving machine learning (ML), other characteristics such as robustness are also critical for its practical viability. For instance, since ML training under SMPC may take longer times (e.g., hours or days in many cases), any interruption of the computation will require restarting the process, which results in more delays and waste of computing resources. While one can maintain exchanged SMPC messages in a separate database, their integrity and authenticity should be guaranteed to be able to re-use them later. Therefore, in this paper, we propose ReplayMPC, an efficient failure recovery mechanism for SMPC based on blockchain technology that enables resuming and re-synchronizing SMPC parties after any type of communication or system failures. Our approach allows SMPC parties to save computation state snapshots they use as restoration points during the recovery and then reproduce the last computation rounds by retrieving information from immutable messages stored on a blockchain. Our experiment results on Algorand blockchain show that recovery is much faster than starting the whole process from scratch, saving time, computation, and networking resources.

Secure multi-party computation (SMPC) allows mutually distrusted parties to evaluate a function jointly without revealing their private inputs. This technique helps organizations collaborate on a common goal without disclosing confidential or protected data. Despite its suitability for privacy-preserving computation, SMPC suffers from network-based performance limitations. Specifically, the SMPC parties perform the techniques in rounds, where they execute a local computation and then share their round output with the other parties. This network interchange creates a bottleneck as parties need to wait until the data propagates before resuming the execution. To reduce the SMPC execution time, we propose a pipelining-like approach for each round’s computation and communication by dividing the data and readjusting the execution order. Targeting deep learning applications, we propose strategies for the case of matrix multiplication, a core component of such applications. Our results on a distributed cloud deployment show a significant reduction in the SMPC execution time.

With the recent developments of communication technologies surrounding vehicles, we will be witnessing the simultaneous availability of multiple on-board communication interfaces on vehicles. While most of the current interfaces already include Bluetooth, WiFi, and LTE, they will be augmented further by IEEE 802.11p and the 5G interfaces, which will serve for safety, maintenance, and infotainment applications. However, dynamic management of interfaces depending on application needs will become a significant issue that can be best addressed by Software Defined Networking (SDN) technology. While SDN-based vehicular networks have been promoted previously, none of these works dealt with their practical challenges. In this paper, we propose and develop a practical framework that will realize SDN-based vehicular networks for a wide range of applications. Through this framework, we demonstrate a platoon example which demonstrates the use of SDN for quick and efficient multi-hop messaging. The route from source vehicle to destination is computed with the help of the SDN Controller to transmit the Beacon Safety Messages through Road Side Units (RSUs) at the MAC layer without relying on IP for proper platooning operations. The results show the efficiency of the SDN-based approach compared to the traditional routing approaches.

With the increasing interest in Secure Multi-Party Computation protocols (MPC), there have been several works such as the SPDZ1protocol that tackled this problem under a malicious security with dishonest majority attack model. However, most of these MPC efforts assume that the nodes running the computations are also supplying the inputs, which is not a realistic assumption for many real-life applications. In this paper, we extend the SPDZ protocol to enable clients outsource data and computation to the clouds while ensuring the correctness of the results, in addition to integrity and confidentiality of the input and output. We guarantee that the computation among nodes is done correctly by verifying their output’s Message Authentication Codes (MACs) at the end. Specifically, we delegate this task to an honest server. Our approach strives to minimize the burden on clients while enabling cheating detection even when assuming a malicious attack model with dishonest majority.

With the proliferation of drones in our daily lives, there is an increasing need for handling their numerous challenges. One of such challenge arises when a swarm-of-drones are deployed to accomplish a specific task which requires coordination and communication among the drones. While this swarm-of-drones is essentially a special form of mobile ad hoc networks (MANETs) which has been studied for many years, there are still some unique requirements of drone applications that necessitates re-visiting MANET approaches. These challenges stem from 3--D environments the drones are deployed in, and their specific way of mobility which adds to the wireless link management challenges among the drones. In this paper, we consider an existing routing standard that is used to enable meshing capability among Wi-Fi enabled nodes, namely IEEE 802.11s and adopt its routing capabilities for swarm-of-drones. Specifically, we propose a link quality metric called SrFTime as an improvement to existing Airtime metric which is the 802.11s default routing metric to enable better network throughput for drone applications. This new metric is designed to fit the link characteristics of drones and enable more efficient routes from drones to their gateway. The evaluations in the actual 802.11s standard indicates that our proposed metric outperforms the existing one consistently under various conditions.