Backdoor Attacks on Contrastive Continual Learning for IoT Systems

TL;DR

This work investigates backdoor threats in IoT systems that employ Contrastive Continual Learning (CCL) to cope with non-stationary environments. It formalizes embedding-level attack objectives and analyzes persistence mechanisms created by replay and stability constraints, showing that backdoors can entrench themselves in the representation space across increments. A layered taxonomy aligned with IoT architecture is developed, covering system layers, learning stages, trigger realizations, and attack objectives, along with a comparison to standard CL and static SL. The paper also surveys defense strategies spanning data-space, embedding-space, and federated defenses, highlighting IoT-specific constraints such as limited memory and edge computing, and outlines open research directions for certifiable robustness and efficient replay-aware protection. Overall, while CCL enhances adaptive IoT intelligence, it can raise long-lived representation-level threats if not secured with embedding-centric defenses and robust federated mechanisms.

Abstract

The Internet of Things (IoT) systems increasingly depend on continual learning to adapt to non-stationary environments. These environments can include factors such as sensor drift, changing user behavior, device aging, and adversarial dynamics. Contrastive continual learning (CCL) combines contrastive representation learning with incremental adaptation, enabling robust feature reuse across tasks and domains. However, the geometric nature of contrastive objectives, when paired with replay-based rehearsal and stability-preserving regularization, introduces new security vulnerabilities. Notably, backdoor attacks can exploit embedding alignment and replay reinforcement, enabling the implantation of persistent malicious behaviors that endure through updates and deployment cycles. This paper provides a comprehensive analysis of backdoor attacks on CCL within IoT systems. We formalize the objectives of embedding-level attacks, examine persistence mechanisms unique to IoT deployments, and develop a layered taxonomy tailored to IoT. Additionally, we compare vulnerabilities across various learning paradigms and evaluate defense strategies under IoT constraints, including limited memory, edge computing, and federated aggregation. Our findings indicate that while CCL is effective for enhancing adaptive IoT intelligence, it may also elevate long-lived representation-level threats if not adequately secured.