CyberSec Research

Fairness has emerged as a critical consideration in the landscape of machine learning algorithms, particularly as AI continues to transform decision-making across societal domains. To ensure that these algorithms are free from bias and do not discriminate against individuals based on sensitive attributes such as gender and race, the field of algorithmic bias has introduced various fairness concepts, along with methodologies to achieve these notions in different contexts. Despite the rapid advancement, not all sectors have embraced these fairness principles to the same extent. One specific sector that merits attention in this regard is insurance. Within the realm of insurance pricing, fairness is defined through a distinct and specialized framework. Consequently, achieving fairness according to established notions does not automatically ensure fair pricing in insurance. In particular, regulators are increasingly emphasizing transparency in pricing algorithms and imposing constraints on insurance companies on the collection and utilization of sensitive consumer attributes. These factors present additional challenges in the implementation of fairness in pricing algorithms. To address these complexities and comply with regulatory demands, we propose an efficient method for constructing fair models that are tailored to the insurance domain, using only privatized sensitive attributes. Notably, our approach ensures statistical guarantees, does not require direct access to sensitive attributes, and adapts to varying transparency requirements, addressing regulatory demands while ensuring fairness in insurance pricing.

In this study, we propose a novel integrated Generalized Autoregressive Conditional Heteroskedasticity-Gated Recurrent Unit (GARCH-GRU) model for financial volatility modeling and forecasting. The model embeds the GARCH(1,1) formulation directly into the GRU cell architecture, yielding a unified recurrent unit that jointly captures both traditional econometric properties and complex temporal dynamics. This hybrid structure leverages the strengths of GARCH in modeling key stylized facts of financial volatility, such as clustering and persistence, while utilizing the GRU's capacity to learn nonlinear dependencies from sequential data. Compared to the GARCH-LSTM counterpart, the GARCH-GRU model demonstrates superior computational efficiency, requiring significantly less training time, while maintaining and improving forecasting accuracy. Empirical evaluation across multiple financial datasets confirms the model's robust outperformance in terms of mean squared error (MSE) and mean absolute error (MAE) relative to a range of benchmarks, including standard neural networks, alternative hybrid architectures, and classical GARCH-type models. As an application, we compute Value-at-Risk (VaR) using the model's volatility forecasts and observe lower violation ratios, further validating the predictive reliability of the proposed framework in practical risk management settings.

With rapid technological progress reshaping the financial industry, quantum technology plays a critical role in advancing risk management, asset allocation, and financial strategies. Realizing its full potential requires overcoming challenges like quantum hardware limits, algorithmic stability, and implementation barriers. This research explores integrating quantum annealing with portfolio optimization, highlighting quantum methods' ability to enhance investment strategy efficiency and speed. Using hybrid quantum-classical models, the study shows combined approaches effectively handle complex optimization better than classical methods. Empirical results demonstrate a portfolio increase of 200,000 Indian Rupees over the benchmark. Additionally, using rebalancing leads to a portfolio that also surpasses the benchmark value.

We consider an equity market subject to risk from both unhedgeable shocks and default. The novelty of our work is that to partially offset default risk, investors may dynamically trade in a credit default swap (CDS) market. Assuming investment opportunities are driven by functions of an underlying diffusive factor process, we identify the certainty equivalent for a constant absolute risk aversion investor with a semi-linear partial differential equation (PDE) which has quadratic growth in both the function and gradient coefficients. For general model specifications, we prove existence of a solution to the PDE which is also the certainty equivalent. We show the optimal policy in the CDS market covers not only equity losses upon default (as one would expect), but also losses due to restricted future trading opportunities. We use our results to price default dependent claims though the principal of utility indifference, and we show that provided the underlying equity market is complete absent the possibility of default, the equity-CDS market is complete accounting for default. Lastly, through a numerical application, we show the optimal CDS policies are essentially static (and hence easily implementable) and that investing in CDS dramatically increases investor indirect utility.

We study the problem of resource provisioning under stringent reliability or service level requirements, which arise in applications such as power distribution, emergency response, cloud server provisioning, and regulatory risk management. With chance-constrained optimization serving as a natural starting point for modeling this class of problems, our primary contribution is to characterize how the optimal costs and decisions scale for a generic joint chance-constrained model as the target probability of satisfying the service/reliability constraints approaches its maximal level. Beyond providing insights into the behavior of optimal solutions, our scaling framework has three key algorithmic implications. First, in distributionally robust optimization (DRO) modeling of chance constraints, we show that widely used approaches based on KL-divergences, Wasserstein distances, and moments heavily distort the scaling properties of optimal decisions, leading to exponentially higher costs. In contrast, incorporating marginal distributions or using appropriately chosen f-divergence balls preserves the correct scaling, ensuring decisions remain conservative by at most a constant or logarithmic factor. Second, we leverage the scaling framework to quantify the conservativeness of common inner approximations and propose a simple line search to refine their solutions, yielding near-optimal decisions. Finally, given N data samples, we demonstrate how the scaling framework enables the estimation of approximately Pareto-optimal decisions with constraint violation probabilities significantly smaller than the Omega(1/N)-barrier that arises in the absence of parametric assumptions

We study distributionally robust optimization (DRO) problems with uncertainty sets consisting of high dimensional random vectors that are close in the multivariate Wasserstein distance to a reference random vector. We give conditions under which the images of these sets under scalar-valued aggregation functions are equal to or contained in uncertainty sets of univariate random variables defined via a univariate Wasserstein distance. This allows to rewrite or bound high-dimensional DRO problems with simpler DRO problems over the space of univariate random variables. We generalize the results to uncertainty sets defined via the Bregman-Wasserstein divergence and the max-sliced Wasserstein and Bregman-Wasserstein divergence. The max-sliced divergences allow us to jointly model distributional uncertainty around the reference random vector and uncertainty in the aggregation function. Finally, we derive explicit bounds for worst-case risk measures that belong to the class of signed Choquet integrals.

We propose an enhanced deep hedging framework for index option portfolios, grounded in a realistic market simulator that captures the joint dynamics of S&P 500 returns and the full implied volatility surface. Our approach integrates surface-informed decisions with multiple hedging instruments and explicitly accounts for transaction costs. The hedging strategy also considers the variance risk premium embedded in the hedging instruments, enabling more informed and adaptive risk management. In this setting, state-dependent no-trade regions emerge naturally, improving rebalancing efficiency and hedging performance. Tested across simulated and historical data from 1996 to 2020, our method consistently outperforms traditional delta and delta-gamma hedging, demonstrating superior adaptability and risk reduction.

Risk management in finance involves recognizing, evaluating, and addressing financial risks to maintain stability and ensure regulatory compliance. Extracting relevant insights from extensive regulatory documents is a complex challenge requiring advanced retrieval and language models. This paper introduces RiskData, a dataset specifically curated for finetuning embedding models in risk management, and RiskEmbed, a finetuned embedding model designed to improve retrieval accuracy in financial question-answering systems. The dataset is derived from 94 regulatory guidelines published by the Office of the Superintendent of Financial Institutions (OSFI) from 1991 to 2024. We finetune a state-of-the-art sentence BERT embedding model to enhance domain-specific retrieval performance typically for Retrieval-Augmented Generation (RAG) systems. Experimental results demonstrate that RiskEmbed significantly outperforms general-purpose and financial embedding models, achieving substantial improvements in ranking metrics. By open-sourcing both the dataset and the model, we provide a valuable resource for financial institutions and researchers aiming to develop more accurate and efficient risk management AI solutions.

United States (US) IG bonds typically trade at modest spreads over US Treasuries, reflecting the credit risk tied to a corporation's default potential. During market crises, IG spreads often widen and liquidity tends to decrease, likely due to increased credit risk (evidenced by higher IG Credit Default Index spreads) and the necessity for asset holders like mutual funds to liquidate assets, including IG credits, to manage margin calls, bolster cash reserves, or meet redemptions. These credit and liquidity premia occur during market drawdowns and tend to move non-linearly with the market. The research herein refers to this non-linearity (during periods of drawdown) as downside convexity, and shows that this market behavior can effectively be captured through a short position established in IG Exchange Traded Funds (ETFs). The following document details the construction of three signals: Momentum, Liquidity, and Credit, that can be used in combination to signal entries and exits into short IG positions to hedge a typical active bond portfolio (such as PIMIX). A dynamic hedge initiates the short when signals jointly correlate and point to significant future hedged return. The dynamic hedge removes when the short position's predicted hedged return begins to mean revert. This systematic hedge largely avoids IG Credit drawdowns, lowers absolute and downside risk, increases annualised returns and achieves higher Sortino ratios compared to the benchmark funds. The method is best suited to high carry, high active risk funds like PIMIX, though it also generalises to more conservative funds similar to DODIX.

We consider the problem of an agent who faces losses over a finite time horizon and may choose to share some of these losses with a counterparty. The agent is uncertain about the true loss distribution and has multiple models for the losses. Their goal is to optimize a mean-variance type criterion with model ambiguity through risk sharing. We construct such a criterion by adapting the monotone mean-variance preferences of Maccheroni et al. (2009) to the multiple models setting and exploit a dual representation to mitigate time-consistency issues. Assuming a Cram\'er-Lundberg loss model, we fully characterize the optimal risk sharing contract and the agent's wealth process under the optimal strategy. Furthermore, we prove that the strategy we obtain is admissible and prove that the value function satisfies the appropriate verification conditions. Finally, we apply the optimal strategy to an insurance setting using data from a Spanish automobile insurance portfolio, where we obtain differing models using cross-validation and provide numerical illustrations of the results.

We introduce a framework for systemic risk modeling in insurance portfolios using jointly exchangeable arrays, extending classical collective risk models to account for interactions. We establish central limit theorems that asymptotically characterize total portfolio losses, providing a theoretical foundation for approximations in large portfolios and over long time horizons. These approximations are validated through simulation-based numerical experiments. Additionally, we analyze the impact of dependence on portfolio loss distributions, with a particular focus on tail behavior.

We present a methodology for causal risk analysis in a network. Causal dependence is formulated by a max-linear structural equation model, which expresses each node variable as a max-linear function of its parental node variables in a directed acyclic graph and some exogenous innovation. We determine directed~paths~responsible~for extreme risk propagation in the network. We give algorithms for structure learning and parameter estimation and apply them to a network of financial data.

Agricultural price volatility challenges sustainable finance, planning, and policy, driven by market dynamics and meteorological factors such as temperature and precipitation. In India, the Minimum Support Price (MSP) system acts as implicit crop insurance, shielding farmers from price drops without premium payments. We analyze the impact of climate on price volatility for soybean (Madhya Pradesh), rice (Assam), and cotton (Gujarat). Using ERA5-Land reanalysis data from the Copernicus Climate Change Service, we analyze historical climate patterns and evaluate two scenarios: SSP2.4.5 (moderate case) and SSP5.8.5 (severe case). Our findings show that weather conditions strongly influence price fluctuations and that integrating meteorological data into volatility models enhances risk-hedging. Using the Exponential Generalized Autoregressive Conditional Heteroskedasticity (EGARCH) model, we estimate conditional price volatility and identify cross-correlations between weather and price volatility movements. Recognizing MSP's equivalence to a European put option, we apply the Black-Scholes model to estimate its implicit premium, quantifying its fiscal cost. We propose this novel market-based risk-hedging mechanism wherein the government purchases insurance equivalent to MSP, leveraging Black-Scholes for accurate premium estimation. Our results underscore the importance of meteorological data in agricultural risk modeling, supporting targeted insurance and strengthening resilience in agricultural finance. This climate-informed financial framework enhances risk-sharing, stabilizes prices, and informs sustainable agricultural policy under growing climate uncertainty.

Loss Given Default (LGD) is a key risk parameter in determining a bank's regulatory capital. During LGD-estimation, realised recovery cash flows are to be discounted at an appropriate rate. Regulatory guidance mandates that this rate should allow for the time value of money, as well as include a risk premium that reflects the "undiversifiable risk" within these recoveries. Having extensively reviewed earlier methods of determining this rate, we propose a new approach that is inspired by the cost of capital approach from the Solvency II regulatory regime. Our method involves estimating a market-consistent price for a portfolio of defaulted loans, from which an associated discount rate may be inferred. We apply this method to mortgage and personal loans data from a large South African bank. The results reveal the main drivers of the discount rate to be the mean and variance of these recoveries, as well as the bank's cost of capital in excess of the risk-free rate. Our method therefore produces a discount rate that reflects both the undiversifiable risk of recovery recoveries and the time value of money, thereby satisfying regulatory requirements. This work can subsequently enhance the LGD-component within the modelling of both regulatory and economic capital.

We propose to model the records of the maximum Drawdown in capital markets by means a Piecewise Deterministic Markov Process (PDMP). We derive statistical results such as the mean and variance that describes the sequence of maximum Drawdown records. In addition, we developed a simulation study and techniques for estimating the parameters governing the stochastic process, using a practical example in the capital market to illustrate the procedure.

This work analytically characterizes impermanent loss for automated market makers (AMMs) in decentralized markets such as Uniswap or Balancer (CPMM). We derive a static replication formula for the pool's value using a combination of European calls and puts. Furthermore, we establish a result guaranteeing hedging coverage for all final prices within a predefined interval. These theoretical results motivate a numerical example where we illustrate the strangle strategy using real cryptocurrency options data from Deribit, one of the most liquid markets available.

Lending within decentralized finance (DeFi) has facilitated over \$100 billion of loans since 2020. A long-standing inefficiency in DeFi lending protocols such as Aave is the use of static pricing mechanisms for loans. These mechanisms have been shown to maximize neither welfare nor revenue for participants in DeFi lending protocols. Recently, adaptive supply models pioneered by Morpho and Euler have become a popular means of dynamic pricing for loans. This pricing is facilitated by agents known as curators, who bid to match supply and demand. We construct and analyze an online learning model for static and dynamic pricing models within DeFi lending. We show that when loans are small and have a short duration relative to an observation time $T$, adaptive supply models achieve $O(\log T)$ regret, while static models cannot achieve better than $\Omega(\sqrt{T})$ regret. We then study competitive behavior between curators, demonstrating that adaptive supply mechanisms maximize revenue and welfare for both borrowers and lenders.

This study explores the integration of a representative large language model, ChatGPT, into lending decision-making with a focus on credit default prediction. Specifically, we use ChatGPT to analyse and interpret loan assessments written by loan officers and generate refined versions of these texts. Our comparative analysis reveals significant differences between generative artificial intelligence (AI)-refined and human-written texts in terms of text length, semantic similarity, and linguistic representations. Using deep learning techniques, we show that incorporating unstructured text data, particularly ChatGPT-refined texts, alongside conventional structured data significantly enhances credit default predictions. Furthermore, we demonstrate how the contents of both human-written and ChatGPT-refined assessments contribute to the models' prediction and show that the effect of essential words is highly context-dependent. Moreover, we find that ChatGPT's analysis of borrower delinquency contributes the most to improving predictive accuracy. We also evaluate the business impact of the models based on human-written and ChatGPT-refined texts, and find that, in most cases, the latter yields higher profitability than the former. This study provides valuable insights into the transformative potential of generative AI in financial services.