Related papers: UNSAIL: Thwarting Oracle-Less Machine Learning Att…

Large language models (LLMs) employ safety mechanisms to prevent harmful outputs, yet these defenses primarily rely on semantic pattern matching. We show that encoding harmful prompts as coherent mathematical problems -- using formalisms…

Recent work has shown that, while large language models (LLMs) demonstrate strong word translation or bilingual lexicon induction (BLI) capabilities in few-shot setups, they still cannot match the performance of 'traditional' mapping-based…

Decompositional jailbreaks pose a critical threat to large language models (LLMs) by allowing adversaries to fragment a malicious objective into a sequence of individually benign queries that collectively reconstruct prohibited content. In…

The evolution of Internet and its related communication technologies have consistently increased the risk of cyber-attacks. In this context, a crucial role is played by Intrusion Detection Systems (IDSs), which are security devices designed…

We propose SANSCrypt, a novel sequential logic encryption scheme to protect integrated circuits against reverse engineering. Previous sequential encryption methods focus on modifying the circuit state machine such that the correct…

Backdoor attacks compromise the integrity and reliability of machine learning models by embedding a hidden trigger during the training process, which can later be activated to cause unintended misbehavior. We propose a novel backdoor…

Industrial control system (ICS) operations use trusted endpoints like human machine interfaces (HMIs) and workstations to relay commands to programmable logic controllers (PLCs). Because most PLCs lack layered defenses, compromise of a…

Layout camouflaging can protect the intellectual property of modern circuits. Most prior art, however, incurs excessive layout overheads and necessitates customization of active-device manufacturing processes, i.e., the front-end-of-line…

State-of-the-art attacks against cyclic logic obfuscation use satisfiability solvers that are equipped with a set of cycle avoidance clauses. These cycle avoidance clauses are generated in a pre-processing step and define various key…

Fault-tolerant quantum error correction provides a strategy to protect information processed by a quantum computer against noise which would otherwise corrupt the data. A fault-tolerant universal quantum computer must implement a universal…

Mechanistic interpretability reveals that safety-critical behaviors (e.g., alignment, jailbreak, backdoor) in Large Language Models (LLMs) are grounded in specialized functional components. However, existing safety attribution methods…

Few-Shot Class-Incremental Learning (FSCIL) enables machine learning systems to expand their inference capabilities to new classes using only a few labeled examples, without forgetting the previously learned classes. Classical…

Sound Source Localization (SSL) enabling technology for applications such as surveillance and robotics. While traditional Signal Processing (SP)-based SSL methods provide analytic solutions under specific signal and noise assumptions,…

Protecting intellectual property (IP) has become a serious challenge for chip designers. Most countermeasures are tailored for CMOS integration and tend to incur excessive overheads, resulting from additional circuitry or device-level…

Large Language Models (LLMs) offer transformative capabilities for hardware design automation, particularly in Verilog code generation. However, they also pose significant data security challenges, including Verilog evaluation data…

Ensuring the reliability and safety of machine learning models in open-world deployment is a central challenge in AI safety. This thesis develops both algorithmic and theoretical foundations to address key reliability issues arising from…

The LLM unlearning technique has recently been introduced to comply with data regulations and address the safety and ethical concerns of LLMs by removing the undesired data-model influence. However, state-of-the-art unlearning methods face…

Split manufacturing was introduced as an effective countermeasure against hardware-level threats such as IP piracy, overbuilding, and insertion of hardware Trojans. Nevertheless, the security promise of split manufacturing has been…

Backdoor attacks pose a persistent security risk to deep neural networks (DNNs) due to their stealth and durability. While recent research has explored leveraging model unlearning mechanisms to enhance backdoor concealment, existing attack…

The security of logic locking has been called into question by various attacks, especially a Boolean satisfiability (SAT) based attack, that exploits scan access in a working chip. Among other techniques, a robust design-for-security (DFS)…