Ranodeep Saha

Complex problem-solver with analytical and driven mindset. Dedicated to achieving demanding development objectives according to tight schedules while producing impeccable code.

• Analyzation of propagation of network vulnerabilities through MulVal framework

      Set up MulVAL and creating an attack graph of a given input file. Set up MySQL server fetching vulnerabilities from NVD website and scanning a host using OVAL/nessus scanner. Analyzed the propagation and impact of vulnerabilities of a system. Generated interaction rules from CVE descriptions.

• Generating adversarial examples for text-classification model

      Gathered a dataset of sarcastic and non-sarcastic newspaper headlines and transformed the words into word vectors using pre-trained word2vec model and build a model of Word-CNN to classify them. Reduced the accuracy of the model by generating new examples using methods like FGSM. 

• SQL injection detection through machine learning methods

      Gathered a labeled dataset of malicious and non-malicious SQL queries from Kaggle website. Extract features from SQL queries using various pre-processing methods like feature extraction, word embedding etc. Performed a literature survey on the existing methods to detect SQL injection using machine learning methods. 

• Online degree verification using Blockchain

       Made a decentralized app to store degree information of IIT Bombay students. Implemented a decentralized database on Ethereum blockchain with the help of solidity and truffle. Implemented various users with different access policies to enter, verify and request student information. 

• Verification and Robustness Analysis of Neural Networks using DeepPoly

       Verified and analyzed the classification property of NNs by applying a small perturbation to the image. Used ETH Robustness Analyzer for NN with the DeepPoly domain for the verification of NN. 

• Verifying Concurrent Programs Under Sequential Consistency

       Build a tool to produce all the valid traces of sequential execution of an input program consisting of read and write instructions of concurrent processes and check the satisfiability of the assertion statement. Identified read and write operations and made fr, rf and ws edges in all the traces to identify valid traces. 

• Performance Analysis and Benchmarking of various cryptographic methods using Cryptographic Libraries

       Explored 3 Libraries Open SSL, Crypto++, Bouncy Castle. Compared the time taken for encryption and decryption of different symmetric and asymmetric cryptographic algorithms by changing their native library, rounds , and key length.

• Fire prediction using Linear Regression

       Built a Linear Regression model from scratch by performing feature selection and scaling on data-set. Selection of proper basis function and computing mean squared error by using both analytical and gradient descent iterative solution. Analyzed the accuracy by using proper regularization and stopping criteria. 

• Classification of unbalanced data

       Solved binary classification problem by applying various machine learning models. Classified an imbalanced dataset by applying algorithms like Isolation-forest and Local outlier factor. Applied different oversampling and undersampling methods like SMOTE,Tomeklink etc. to balance the dataset.