Secure Forecast in the Encrypted Domain Tikaram Sanyashi, Sharad Kumar, and Bernard Menezes Indian Institute of Technology, Bombay, Mumbai, India {tikaram,sharadkmr18,bernard}@cse.iitb.ac.in Abstract. Forecast plays a vital role in a business model. Forecasting inventory data is widely used in real-life applications to predict future market needs and prepare for them. However, for small retailers fore- casting sales data may not be feasible due to a lack of computational and storage resources. In such a scenario, they have an option to pay and use cloud computing facilities. However, storing sales data in the public cloud may result in loss of data privacy; thus, storing data in the encrypted form is necessary. Forecasting data in encrypted form is not an easy task. This paper aims to show different available approaches for cloud computation and use them for forecast computation. We have also designed a lightweight symmetric key encryption scheme for secure single-party cloud computation and used it for secure forecast computa- tion. The efficiency of the proposed encryption scheme is also compared with that of existing cloud computation techniques such as the SEAL library of homomorphic encryption and the SCALE-MAMBA library of secure multiparty computation. Keywords: Secure forecast computation, Time series forecast, Homo- morphic encryption, SEAL library, Multiparty computation, SCALE- MAMBA library 1 Introduction Forecasting plays a vital role in the business model. For keeping track of out-of- stock stuff, sell information is stored, and its forecast data is used to order things from the vendor. Vendor Managed Inventory (VMI) is a business model where a product’s buyer provides information to a product vendor. The vendor takes full responsibility for maintaining an agreed inventory of the material, usually at the buyer’s consumption location. A third-party logistics provider may be involved to ensure that the buyer has the required inventory level by adjusting the demand and supply gaps. The inventory information makes it less likely that a business will unintentionally become out of stock of goods and reduces inventory in the supply chain. Forecast of inventory data is widely used in various industries, from manufac- turing to utilities, healthcare, education, government, and much more. It is not wrong to say that forecast information streamlines and centralizes the process for controlling the flow and maintenance of inventory to ensure that the right amount of stock is available at the right time and of the right quality and at