The Whisper model, an open-source automatic speech recognition system, is widely adopted for its strong performance across multilingual and zero-shot settings. However, it frequently suffers from hallucination errors, especially under noisy acoustic conditions. Previous works to reduce hallucinations in Whisper-style ASR systems have primarily focused on audio preprocessing or post-processing of transcriptions to filter out erroneous content. However, modifications to the Whisper model itself remain largely unexplored to mitigate hallucinations directly. To address this challenge, we present a two-stage architecture that first enhances encoder robustness through Adaptive Layer Attention (ALA) and further suppresses hallucinations using a multi-objective knowledge distillation (KD) framework. In the first stage, ALA groups encoder layers into semantically coherent blocks via inter-layer correlation analysis. A learnable multi-head attention module then fuses these block representations, enabling the model to jointly exploit low- and high-level features for more robust encoding. In the second stage, our KD framework trains the student model on noisy audio to align its semantic and attention distributions with a teacher model processing clean inputs. Our experiments on noisy speech benchmarks show notable reductions in hallucinations and word error rates, while preserving performance on clean speech. Together, ALA and KD offer a principled strategy to improve Whisper’s reliability under real-world noisy conditions.

In cybersecurity, accurate and timely prediction of attacks plays a crucial role in mitigating the risks and impacts of cyber threats.However, traditional attack prediction methods that rely on training Machine Learning (ML) algorithms directly on raw data often suffer from high false alarm rates and low detection rates, leading to inaccurate and unreliable results.To overcome these limitations, this paper presents a novel approach that integrates attack prediction with self-supervision using variance-invariance-covariance regularization (VICReg).The proposed method harnesses VICReg to enhance raw data and generate representations while leveraging self-supervision to learn meaningful features without supervision.Training classic ML algorithms on these refined representations improves prediction accuracy and enhances the robustness of the learning process.We provide a comprehensive description of the proposed method and present an evaluation of its performance on several benchmark datasets.The experimental results demonstrate the superiority of the proposed method over classic ML algorithms.