E-learning has undergone a significant transformation, especially highlighted during crises like the COVID-19 pandemic, when it became essential globally. UNESCO authorized various established e-learning platforms as quick solutions, but these were not recommended as long-term fixes due to several challenges affecting learning processes. Recent studies have addressed these challenges using artificial intelligence (AI), deep learning, and blockchain technologies. AI and deep learning focus on enhancing learner performance evaluation, while blockchain and smart contracts help combat issues like fake certificates, result manipulation, and learner activity tracking. Although both technologies demonstrate strong potential, few studies explore their integration within e-learning, prompting this study to propose a smart framework combining blockchain and deep learning to secure and improve e-learning systems by ensuring data security, transparency, and automation. This framework stores learner data securely on the blockchain using the Interplanetary File System (IPFS) for decentralized large-file storage, ensuring data integrity and confidentiality via Ethereum private blockchain wallets. Deep learning models then analyze this secured data to predict academic performance accurately. Smart contracts facilitate certificate issuance by universities, recording them immutably on the blockchain accessible by network nodes, thereby enhancing automation, security, and trust among learners, professors, and employers. Blockchain provides immutable, time-stamped, secure, and transparent data storage within a distributed peer-to-peer network without a centralized authority. Ethereum, second in market capitalization to Bitcoin, supports programmable smart contracts via the Ethereum Virtual Machine (EVM) using Solidity, enabling conditional and automated transactions far beyond Bitcoin’s capabilities. Smart contracts automate contract terms execution once predefined conditions are met, recording all executions immutably on the blockchain. Due to blockchains being unsuitable for large files, off-chain storage solutions such as IPFS, Storj, and FileCoin are used. IPFS is notable for encrypting and distributing large files peer-to-peer, creating content-addressed hashes that verify data integrity and access, although access control remains a challenge. IPFS is important here for storing learners’ extensive data securely while linking to blockchain transactions via hashes. Deep learning, especially artificial neural networks (ANNs) inspired by biological brains, involves multiple layers—input, hidden, and output layers—that learn via forward propagation, error calculation, and backpropagation through multiple epochs. Deep neural networks (DNNs) with multiple hidden layers enhance prediction accuracy. This study employs these models to process learner data stored via blockchain and IPFS, enabling accurate performance predictions. Several prior studies used deep learning for educational outcomes, including predictions of student dropout rates in Massive Open Online Courses (MOOCs) using recurrent and convolutional neural networks, with improved accuracy over traditional methods. Others utilized bidirectional long-short-term memory networks for dropout predictions, and deep learning models have successfully predicted student performance in small, imbalanced datasets with high accuracy. The proposed framework operates in three phases: 1.

**Storing Learner Data on Blockchain:** Learner data from the “Open University Learning Analytics” dataset (32, 593 records of demographics and virtual learning interactions) is encrypted and stored on IPFS, which generates a cryptographic hash. This hash is stored on a private Ethereum blockchain using smart contracts, enabling decentralized, immutable access. Nodes—including university administration, professors, learners, and guests (employers)—register on the blockchain network using wallets containing private and public keys. 2. **Deep Learning-Based Performance Prediction:** The encrypted learner data retrieved via IPFS undergoes preprocessing involving feature selection, missing value replacement (using mode and constants), categorical data encoding, normalization via MinMaxScaler, and dimensionality reduction using Principal Component Analysis (PCA). The dataset is split (90% train, 10% test) and input to a deep neural network comprising an input layer (10 neurons), five hidden layers (each with 500 neurons, ReLU activation), and an output layer with four neurons representing pass, fail, withdraw, and distinction statuses using softmax activation and sparse categorical cross-entropy loss. This model, implemented in Python, Keras, and Sklearn, achieves high accuracy (~91. 29%) and low loss (~0. 18), outperforming previous studies on the same dataset. 3. **Employing Smart Contracts:** Smart contracts deployed via Solidity on Ethereum enable secure interactions among nodes: professors upload assignments to IPFS, sending file hashes via smart contracts to learners; learners submit assignments through smart contracts; universities issue certificates stored immutably on the blockchain; and employers (guests) access learners’ certificates and performance data upon university approval. These processes ensure transparency, security, and automation. Implementation involved: - **Blockchain Architecture:** Developed using Python, Flask, and Postman for mining blocks, validating the chain, and adding transactions. Nodes are decentralized and registered via MyEtherWallet (MEW), which provides wallet management with keys and addresses. - **Integration of Blockchain and Deep Learning:** IPFS stores encrypted datasets with their hashes secured on Ethereum blockchain via smart contracts. The deep learning model predicts learner performance from secured data. Smart contracts manage transactions, including assignment distribution, submission, certificate issuance, and data access. Testing phases validate functions such as node registration and wallet access, hash storage and verification, certificate issuance and retrieval, assignment interactions, transaction validity, and file integrity on the blockchain. Test results confirm the immutability and tamper-resistance of stored data and smooth interactions via smart contracts. The proposed framework demonstrates that integrating blockchain with deep learning can produce a secure, transparent, and automated e-learning system with high prediction accuracy and robust data management, surpassing prior studies in effectiveness. This approach addresses challenges of data security, learner verification, and automated academic processes, establishing a foundation for future smart e-learning platforms.