Khalil El‐Khatib

Parsa Memarzadehsaghezi, Pooria Madani, Khalil El-Khatib

Large code language models (CodeLLMs) can generate and rewrite programs, enabling functionality-preserving code mutation that may be used to create diverse malware variants and evade signature-based detection. A key security question is whether this mutation capability survives model compression, which would make deployment feasible under limited hardware budgets. We propose SecRL-Prune, a structured pruning framework for CodeLLMs that operates on feed-forward (MLP/FFN) channels. Starting from a pretrained teacher, it learns a layer-wise pruning policy with reinforcement learning using a teacher-student KL-divergence reward. To improve efficiency, we cache the teacher's top-P predictions once and compare the pruned student against this compact target, avoiding simultaneous teacher-student residency in GPU memory. We evaluate SecRL-Prune on HumanEval using pass@k for execution correctness and var@k for code diversity across three 7B CodeLLMs at 10-30% compression. SecRL-Prune consistently preserves higher pass@k and var@k than recent structured pruning baselines under aggressive pruning. In a case study on real malware samples, semantics-preserving mutations from 20%-pruned models substantially reduced detections. These results show that code mutation capability can survive significant structured pruning, highlighting the security relevance of compressed CodeLLMs.

Einstein Rivas Pizarro, Wajiha Zaheer, Li Yang et al.

Radiation Detection Systems (RDSs) play a vital role in ensuring public safety across various settings, from nuclear facilities to medical environments. However, these systems are increasingly vulnerable to cyber-attacks such as data injection, man-in-the-middle (MITM) attacks, ICMP floods, botnet attacks, privilege escalation, and distributed denial-of-service (DDoS) attacks. Such threats could compromise the integrity and reliability of radiation measurements, posing significant public health and safety risks. This paper presents a new synthetic radiation dataset and an Intrusion Detection System (IDS) tailored for resource-constrained environments, bringing Machine Learning (ML) predictive capabilities closer to the sensing edge layer of critical infrastructure. Leveraging TinyML techniques, the proposed IDS employs an optimized XGBoost model enhanced with pruning, quantization, feature selection, and sampling. These TinyML techniques significantly reduce the size of the model and computational demands, enabling real-time intrusion detection on low-resource devices while maintaining a reasonable balance between efficiency and accuracy.

Brandon Laughlin, Christopher Collins, Karthik Sankaranarayanan et al.

This paper presents a framework which enables a user to more easily make corrections to adversarial texts. While attack algorithms have been demonstrated to automatically build adversaries, changes made by the algorithms can often have poor semantics or syntax. Our framework is designed to facilitate human intervention by aiding users in making corrections. The framework extends existing attack algorithms to work within an evolutionary attack process paired with a visual analytics loop. Using an interactive dashboard a user is able to review the generation process in real time and receive suggestions from the system for edits to be made. The adversaries can be used to both diagnose robustness issues within a single classifier or to compare various classifier options. With the weaknesses identified, the framework can also be used as a first step in mitigating adversarial threats. The framework can be used as part of further research into defense methods in which the adversarial examples are used to evaluate new countermeasures. We demonstrate the framework with a word swapping attack for the task of sentiment classification.

Youssef Gahi, Mouhcine Guennoun, Khalil El-Khatib

Cloud computing emerges as an attractive solution that can be delegated to store and process confidential data. However, several security risks are encountered with such a system as the securely encrypted data should be decrypted before processing them. Therefore, the decrypted data is susceptible to reading and alterations. As a result, processing encrypted data has been a research subject since the publication of the RSA encryption scheme in 1978. In this paper we present a relational database system based on homomorphic encryption schemes to preserve the integrity and confidentiality of the data. Our system executes SQL queries over encrypted data. We tested our system with a recently developed homomorphic scheme that enables the execution of arithmetic operations on ciphertexts. We show that the proposed system performs accurate SQL operations, yet its performance discourages a practical implementation of this system.

Youssef Gahi, Mouhcine Guennoun, Zouhair Guennoun et al.

The trend towards delegating data processing to a remote party raises major concerns related to privacy violations for both end-users and service providers. These concerns have attracted the attention of the research community, and several techniques have been proposed to protect against malicious parties by providing secure communication protocols. Most of the proposed techniques, however, require the involvement of a third party, and this by itself can be viewed as another security concern. These security breaches can be avoided by following a new approach that depends on data sorted, managed, and stored in encrypted form at the remote servers. To realize such an approach, the encryption cryptosystem must support algebraic operations over encrypted data. This cryptosystem can be effective in protecting data and supporting the construction of programs that can process encrypted input and produce encrypted output. In fact, the latter programs do not decrypt the input, and therefore, they can be run by an un-trusted party without revealing their data and internal states. Furthermore, such programs prove to be practical in situations where we need to outsource private computations, especially in the context of cloud computing. Homomorphic cryptosystems are perfectly aligned with these objectives as they are a strong foundation for schemes that allow a blind processing of encrypted data without the need to decrypt them. In this dissertation we rely on homomorphic encryption schemes to secure cloud computing, services and routing protocols. We design several circuits that allow for the blind processing and management of data such that malicious parties are denied access to sensitive information. We select five areas to apply our models to. These models are easily customized for many other areas. We also provide prototypes that we use to study the performance and robustness of our models.