Thales Paiva

We introduce a new decryption algorithm for BIKE, a code-based key encapsulation mechanism selected by the NIST as an alternate candidate for post-quantum cryptography. Furthermore, we showed a fundamental problem with BIKE's current decryption algorithm BGF that may affect its security. Our constant-time implementation outperforms BGF, providing speedups of 1.26, 1.36, and 1.51 for security levels 128, 192, and 256, respectively. As the main author, I designed and implemented the decoder, ran the experiments, and wrote the paper with input from my advisor.

We investigated a vulnerability in the binary setting of the permuted kernel problem (PKP). PKP is a hard combinatorial problem, conjectured to be hard even for quantum computers, and is used by PKP-DSS, a post-quantum signature algorithm. As the main author, I conceived and implemented the idea of the attack and wrote the paper with input from my advisor.

HQC is a code-based encryption scheme selected by the NIST as an alternate candidate for post-quantum cryptography. We showed that when a non-constant-time algorithm is used for decoding a BCH codeword, it is possible to mount a key-recovery attack assisted by timing the decryption of ciphertexts. As the main author, I noticed the information leakage from timing, exploited it, and wrote the paper with input from my advisor.

GEDAE-LaB is an open source software to help sports physiologists estimate the energy expenditure and energy system contributions during exercise. My contribution to the project was mainly in the Java back end, where I implemented the mathematical modeling and optimization algorithms.

In this paper, we explored new features to classify BGP anomalies. As one of the main authors, I was responsible for the feature selection, building the extractors for these features in Python, and implementing the LSTM-based classifier in Python with Keras and TensorFlow with my colleagues.