Large models adaptation through Federated Learning (FL) addresses a wide range of use cases and is enabled by Parameter-Efficient Fine-Tuning techniques such as Low-Rank Adaptation (LoRA). However, this distributed learning paradigm faces several security threats, particularly to its integrity, such as backdoor attacks that aim to inject malicious behavior during the local training steps of certain clients. We present the first analysis of the influence of LoRA on state-of-the-art backdoor attacks targeting model adaptation in FL. Specifically, we focus on backdoor lifespan, a critical characteristic in FL, that can vary depending on the attack scenario and the attacker's ability to effectively inject the backdoor. A key finding in our experiments is that for an optimally injected backdoor, the backdoor persistence after the attack is longer when the LoRA's rank is lower. Importantly, our work highlights evaluation issues of backdoor attacks against FL and contributes to the development of more robust and fair evaluations of backdoor attacks, enhancing the reliability of risk assessments for critical FL systems. Our code is publicly available.

The newly discovered P. mikazuki is a tribute the famous warrior Date Masamune. Breakthroughs, discoveries, and DIY tips sent every weekday. A team of university students in Japan identified an entirely new species of the mighty Portuguese Man O'War . Described in a study recently published in the journal, the creature's distinct features and fearsome venom have earned it a name that honors a famous 16th century samurai warrior. It's easy to mistake the Portuguese Man O'War () for a jellyfish .

Environment Conservation Ocean Dolphins may be getting an Alzheimer's-like disease due to this neurotoxin The neurotoxins, found in algal blooms, primarily affect the body's nervous system. Breakthroughs, discoveries, and DIY tips sent every weekday. For marine biologists, dolphins are often viewed as sentinel species, or animals that shed light on the health of the ocean . Along with whales, porpoises, and other cetacean species, dolphins are one way that researchers know to sound the alarm about environmental hazards that might affect the ocean as a whole and potentially humans. In this context, researchers have connected neurotoxins from algal blooms to brain changes associated with an Alzheimer's-like disease in dolphins in Florida.

As shown in Table 3, IBA attains consistent performance on the Tiny ImageNet. Even against difficult defenses like RLR and Krum, IBA can obtain BAs of approximately 85%.

Each year, snake bites kill upwards of 100,000 people and permanently disable hundreds of thousands more, according to estimates from the World Health Organization. Promising new science, enabled by state-of-the-art technology, could help quell the threat. Researchers have successfully designed two proteins to neutralize some of the most lethal venom toxins, using a suite of artificial intelligence tools, per a study published January 15 in the journal Nature. These "de novo" proteins–molecules not found anywhere in nature–protected 100% of mice from certain death when mixed with the deadly snake compounds and administered in lab experiments. "I think we could revolutionize the treatment [of snake bites]," says Susana Vázquez Torres, lead study author and a biochemist who completed this research as part of her doctoral thesis in David Baker's lab at the University of Washington.

Federated Learning is a promising approach for training machine learning models while preserving data privacy, but its distributed nature makes it vulnerable to backdoor attacks, particularly in NLP tasks while related research remains limited. This paper introduces SDBA, a novel backdoor attack mechanism designed for NLP tasks in FL environments. Our systematic analysis across LSTM and GPT-2 models identifies the most vulnerable layers for backdoor injection and achieves both stealth and long-lasting durability through layer-wise gradient masking and top-k% gradient masking within these layers. Experiments on next token prediction and sentiment analysis tasks show that SDBA outperforms existing backdoors in durability and effectively bypasses representative defense mechanisms, with notable performance in LLM such as GPT-2. These results underscore the need for robust defense strategies in NLP-based FL systems.