The project aims at improving the predictability of financial risks by exploiting the richness of the information content of high-frequency data. The practical goal is to minimize the possible losses that financial institutions may encounter during turbulent financial times and by which each economy, as a whole, but also each individual, as a tax-payer, is directly affected. The objectives of the project are to understand what causes extreme losses during financial turmoil, such as the previous financial crisis and to identify and analyze the high-frequency trading specific information that is most valuable in forecasting extreme financial risks. The main focus is to analyze how this information can be incorporated in accurately measuring the occurrence probabilities and sizes of extreme events on financial markets. Of particular interest is the analysis within a multivariate context, given that investors face simultaneously many sources of risk.