This blog post aims to highlight a promising application of secure (multi-party) computation in business software.
Let me first summarize my view of the technical requirements and accomplishments of secure computation. If multiple parties have inputs (requirement A) and these inputs must be kept highly confidential (requirement B), then they can still safely collaborate (accomplishment A). There is a scenario in business operations where these requirements are met and the accomplishment is useful. In fact, the data is so sensitive that collaboration often does not take place in practice because of security concerns. In that way secure computation is an enabler of additional collaborations not practical previously. This scenario is supply chain collaboration (SCC).
What is the fundamental problem of SCC?
Companies produce goods and services (either to order from customers or to a planned stock level). For this they need to order supplies. The current process is as follows: A company determines how much it wants to produce, checks its supply and inventory and then places orders to its suppliers. This simple process proceeds all the way to the top of the supply chain where raw materials are sourced.
What is the fundamental problem with this approach?
It is long known that this mode of operation does not lead to an optimal use of resources. Each companies optimizes (locally) its use of capacity and stock, but the combination of locally optimal plans is rarely a globally optimal plan. In the entire supply chain significant resources are wasted which implies higher costs for consumers. You might have heard of the bull whip effect. The bull whip effect states that is inevitable in this mode of operation that the fluctuation of orders at the top of the supply is much higher than at the bottom of the supply chain. This implies that companies at the top of the supply chain need to maintain much larger capacities which binds capital and incurs significant additional costs.
What can you do to prevent the problem?
Companies along the supply chain need to exchange data. They need to engage in a collaborative planning process. Supply chain management has come up with a variety of such planning methods. They differ in the number of participating parties — two or many — and in the economic quantity to be optimized. A large scale example with multiple parties that optimizes production, warehousing and transportation is supply chain master planning. A medium scale example with two parties that optimizes production and warehousing is collaborative planning, forecasting and replenishment (known as CPFR). A small scale example that optimizes warehousing is the joint economic lot size (JELS).
How can secure computation help?
A common problem in SCC is that partners at not willing to exchange the necessary data, such as costs and capacities, for security reasons. They fear disadvantages in future collaborations, e.g. price negotiations, due to the insight into their price calculation. This is even often true for simple data exchanges, such as in vendor managed inventory. Therefore few of these schemes have found practical adoption so far. Supply chain researcher have come up with their own solutions, e.g. by using negotiation. Yet, these techniques rarely withstand a rigorous security analysis. Secure computation can implement these planning techniques provably without disclosing the input data. Hence, it may just be the technology that makes them acceptable in business practice.
What is the state of the art?
A number of specialized secure computation protocols have been proposed. The first one that initiated the idea was for CPFR (Atallah et al., 2006, M&SOM). A couple others came later, e.g. Pibernik et al. (2011, EJOR), address the problem of inference from the result of a secure computation of JELS. Even an attempt at something like supply chain master planning was undertaken (Kerschbaum et al., 2011, Computer). And, there are more and even more coming. Still, there are a couple of challenges left: First, as always, increasing the performance is a key challenge. Second, identifying the right computation (planning algorithm) to perform and the right computation model (cloud, etc.) to perform it in can be important for adoption in the market. This, of course, has an impact on which protocols are the fastest. Third, all aspects of security, such as malicious inputs or inferences from the result, etc., need to be addressed.
In summary, supply chain collaboration presents a major opportunity for wider adoption of secure computation due to its high confidentiality requirements. There are a number of challenges to be solved by the cryptography and business community and only their collaboration is likely to bring practically viable results.
florian dot kerschbaum (at) sap dot com