Prodipta Sen Amartya

Interval type-2 fuzzy sets (IT2 FSs) are more popular over type-1 fuzzy sets (T1 FSs) as they capture uncertainty in a better way in many real-world problems. Modelling uncertainty with an IT2 FS is relatively complex and solely depends on how to define its Footprint of Uncertainty (FOU)-an uncertainty region of IT2 FS bounded by its lower and upper membership functions. Existing methods either transform existing T1 FSs to IT2 FSs or design IT2 FSs from input data, captured by interval via constructing T1 FSs. In both cases, they require additional processing step and computational time and thus not useful for time-sensitive applications (e.g., intelligent transportation systems, online recommendation systems). To address this limitation, this paper puts forward a simple approach, termed as ’Interval Creation Approach’ (ICA) to design IT2 FSs directly from the input data. Further, it is designed to work equally well for transforming existing T1 FS into IT2 FS. The new approach skips data preprocessing phase and the creation of T1 FS as an intermediary to define the FOU of the IT2 FS, thus contributing to faster execution. The paper provides a description of the ICA along with a comparison of its effectiveness against the state-of-the-art methods using both synthetic and real-world data sets.

Backdoor Attacks on Large Language Models (LLMs) typically follow a trigger-response pattern, where satisfying a trigger condition activates a single terminal behavior. However, this single-point backdoor activation does not explore attacks in which malicious behavior develops gradually through multiple intermediate stages before reaching the terminal activation stage. We present Cancerous Backdoor Attack, that gradually corrupts an LLM’s reasoning over a sequence of interactions by leveraging invisible tokens and a stage controller. Specifically, invisible tokens provide hidden stages to the model without appearing in the user’s visible request and the stage controller tracks the attack’s progression during inference to realize gradual activation. We evaluate Cancerous Backdoor Attack across several reasoning datasets. The attack consistently produces gradual reasoning corruption across models and tasks. We compare Cancerous Backdoor Attack with prior backdoor attack baselines to demonstrate our proposed attack outperforms the baselines. Our results indicate that reasoning corruption can develop gradually before it becomes visible, making answer accuracy alone insufficient for detecting backdoor attacks.