Licentiate seminar: Rohon Kundu

Seminarium

From: 2024-05-15 13:15 to 15:00
Place: E-house, E:1406
Contact: rohon [dot] kundu [at] eit [dot] lth [dot] se

Rohon Kundu, ” On Decentralized Cloud Storage Security and an Efficient Post-Quantum Encryption Scheme” Chair: Dr. Paul Stankowski-Wagner Examiner: Professor Erik Larsson (EIT) Opponent: Bernardo David, Associate Professor, ITU, Copenhagen

In this thesis, we address three main security problems related to cryptography and cloud storage. To tackle the challenge posed by a quantum computer, we need encryption that is resistant to quantum computers. This category of cryptography is called post-quantum cryptography. In the first paper, we solve a challenge in one of lattice-based cryptographic protocols called Nth-degree Truncated polynomial Ring Unit (NTRU) namely how to reduce the key size while keeping the security level.  We propose a solution that reduces the key size significantly. Our proposed solution allows a practical implementation of NTRU with fast polynomial multiplications.

Next, we move to solve a long-standing problem arising in any cloud storage namely the reduction of storage cost of redundant data and maintaining security and privacy at the same time. Data deduplication is considered to be a tool that can be used to eliminate redundant data and store only one of its copies. But data deduplication also means that the file cannot go through client-side encryption which opens up new possibilities of adversarial threats. In order to tackle this challenge, we propose a new architecture where we perform client-side deduplication along with dynamic erasure protection by introducing a third-party assistant. We also performed an erasure analysis to quantitatively analyze the probability of loss of a file when a large number of replicas are deleted at random.

Finally, we shift our interest to Decentralized Cloud Storage (DCS). DCS solutions like Filecoin, Storj, and Arweave are gaining more popularity in the Web 3.0 ecosystem. But they are not without challenges. The robustness of the DCS protocols remains a challenging ground. Since the file in a DCS protocol is stored in a decentralized manner among different nodes, a Distributed Denial of Service (DDoS) attack would render the system vulnerable to data loss. Therefore, it is important to analyze the robustness of decentralized architecture against DDoS attacks. In our last paper, we perform a similar erasure analysis to that of the second paper but in a decentralized setup, where the adversary aims to disrupt the system by deleting a file from the network. Storj is one of the leading players in the DCS space. We have created an adversarial model capturing the real Storj network scenario and simulated our model using real-time data obtained from the Storj network. We obtain resource budget figures for DDoS on Storj using our model. Also, we propose a better parametric value for the erasure piece distribution in Storj which suits well when there are more unvetted nodes in the network.