ReportWire

Tag: Lancaster University

  • AI Researchers Uncover Key Vulnerabilities in Major LLMs

    AI Researchers Uncover Key Vulnerabilities in Major LLMs

    [ad_1]

    Newswise — Large Language Models (LLMs) such as ChatGPT and Bard have taken the world by storm this year, with companies investing millions to develop these AI tools, and some leading AI chatbots being valued in the billions.

    These LLMs, which are increasingly used within AI chatbots, scrape the entire Internet of information to learn and to inform answers that they provide to user-specified requests, known as ‘prompts’.

    However, computer scientists from the AI security start-up Mindgard and Lancaster University in the UK have demonstrated that chunks of these LLMs can be copied in less than a week for as little as $50, and the information gained can be used to launch targeted attacks.

    The researchers warn that attackers exploiting these vulnerabilities could reveal private confidential information, bypass guardrails, provide incorrect answers, or stage further targeted attacks.

    Detailed in a new paper to be presented at CAMLIS 2023 (Conference on Applied Machine Learning for Information Security) the researchers show that it is possible to copy important aspects of existing LLMs cheaply, and they demonstrate evidence of vulnerabilities being transferred between different models.

    This attack, termed ‘model leeching’, works by talking to LLMs in such a way – asking it a set of targeted prompts – so that the LLMs elicit insightful information giving away how the model works.

    The research team, which focused their study on ChatGPT-3.5-Turbo, then used this knowledge to create their own copy model, which was 100 times smaller but replicated key aspects of the LLM.

    The researchers were then able to use this model copy as a testing ground to work out how to exploit vulnerabilities in ChatGPT without detection. They were then able to use the knowledge gleaned from their model to attack vulnerabilities in ChatGPT with an 11% increased success rate.

    Dr Peter Garraghan of Lancaster University, CEO of Mindgard, and Principal Investigator on the research, said: “What we discovered is scientifically fascinating, but extremely worrying. This is among the very first works to empirically demonstrate that security vulnerabilities can be successfully transferred between closed source and open source Machine Learning models, which is extremely concerning given how much industry relies on publicly available Machine Learning models hosted in places such as HuggingFace.”

    The researchers say their work highlights that although these powerful digital AI technologies have clear uses, there exist hidden weaknesses, and there may even be common vulnerabilities across models.

    Businesses across industry are currently or preparing to invest billions in creating their own LLMs to undertake a wide range of tasks such as smart assistants. Financial services and large enterprises are adopting these technologies but researchers say that these vulnerabilities should be a major concern for all businesses that are planning to build or use third party LLMs.

    Dr Garraghan said: “While LLM technology is potentially transformative, businesses and scientists alike will have to think very carefully on understanding and measuring the cyber risks associated with adopting and deploying LLMs.”

    The paper will be presented at CAMLIS 2023 in Arlington, Virginia USA which is held on October 19 and 20.

    The paper’s authors are Lewis Birch, William Hackett, Stefan Trawicki, and Neeraj Suri of Lancaster University, and Peter Garraghan of Lancaster University and Mindgard.

    [ad_2]

    Lancaster University

    Source link

  • How a high fat diet allows expulsion of intestinal parasite worms

    How a high fat diet allows expulsion of intestinal parasite worms

    [ad_1]

    Newswise — Scientists have discovered that a high-fat diet allows the immune system to eliminate a parasitic worm which is a major cause of death and illness in the developing world.

    Parasitic worms affect up to a billion people, particularly in developing nations with poor sanitation. One of these parasites known as “whipworm” can cause long lasting infections in the large intestine.

    Researchers from Lancaster University and the University of Manchester  in the UK have discovered that a high-fat diet allows the immune system to eliminate the parasite. 

    Lead author Dr Evelyn Funjika, formerly at Manchester and now at the University of Zambia, said: “Just like the UK, the cheapest diets are often high in fat and at-risk communities to whipworm are increasingly utilising these cheap diets. Therefore, how worm infection and western diets interact is a key unknown for developing nations.

    “In order to be able to study how nutrition affects parasite worm infection, we have been using a mouse model, Trichuris muris, closely related to the human whipworm Trichuris trichiura and seeing how a high-fat diet impacts immunity.”

    It has been previously shown that immune responses which expel the parasite rely on white blood cells called T-helper 2 cells, specialised for eliminating gastrointestinal parasites.

    The findings, published in the journal “Mucosal Immunology”, demonstrate how a high-fat diet, rather than obesity itself, increases a molecule on T-helper cells called ST2 and this allows an increased T-helper 2 response which expels the parasite from the large intestinal lining.  

    Dr John Worthington from the Department of Biomedical and Life Science at Lancaster University co-led the research.

    “We were quite surprised by what we found during this study. High-fat diets are mostly associated with increased pathology during disease. However, in the case of whipworm infection this high fat diet licenses the T-helper cells to make the correct immune response to expel the worm.”

    Co-lead Professor Richard Grencis from the University of Manchester said: “Our studies in mice on a standard diet demonstrate that ST2 is not normally triggered when expelling the parasite, but the high-fat diet boosts the levels of ST2 and hence allows expulsion via an alternative pathway”.

    Co-lead Professor David Thornton from the University of Manchester added: “It was really fascinating that simply altering the diet completely switched the immune response in the gut from one that fails to expel the parasite, to one that brings about all the correct mechanisms to eliminate it.”

    However, Dr Worthington added caution to the findings.

    “Before you order that extra take-away, we have previously published that weight loss can aid the expulsion of a different gut parasite worm. So these results may be context specific, but what is really exciting is the demonstration of how diet can profoundly alter the capacity to generate protective immunity and this may give us new clues for treatments for the millions who suffer from intestinal parasitic infections worldwide.”

    The research was funded by the Commonwealth Scholarship Commission, The Wellcome Trust and EPSRC (Engineering and Physical Sciences Research Council).

    [ad_2]

    Lancaster University

    Source link

  • A win, win, win for dairy production in East Africa

    A win, win, win for dairy production in East Africa

    [ad_1]

    Newswise — Adopting high yield dairy cattle breeds and improving feed would allow Tanzania to increase milk production, while reducing planet warming greenhouse gas (GHG) emissions and alleviating poverty, a new study reveals.

    Tanzania has the second largest dairy herd in East Africa with 28 million cows. However, its dairy sector is poorly developed with mainly small-scale farms stocked with low-yielding breeds, using poor quality feeds. This, along with other supply chain problems around handling and refrigeration, results in poor productivity and the need to import processed dairy products leading to a $23 million trade deficit.

    A new research paper from an international team of researchers led by Lancaster University scientists and published in Nature Food is the first to find evidence that breeding higher yielding dairy cattle offers significant potential to help Tanzania to reduce its dependency on foreign food imports and at the same time help meet its climate commitments.

    The findings show that two key targets of Tanzanian government policy – becoming self-sufficient in milk and cutting GHG emissions by a third – can be achieved simultaneously while increasing income in farming communities.

    Researchers carried out a household survey of 1,200 dairy farmers in Tanzania which was used as a baseline for a sectoral modelling analysis. The survey, which extended across four districts and two agro-ecological zones, was used to estimate milk production, yields from different cattle breeds and how the cattle are managed – for instance what they are fed and how disease is managed.

    Tanzania’s local cattle cope well with high temperatures but produce little milk. New breeds, which cross local cattle with high yielding European cows, produce three times as much milk, while still coping well with heat.

    The study takes as its starting point the Tanzanian Dairy Development Roadmap (DDR), a Government plan which, with support from stakeholders, aims to achieve dairy self-sufficiency by 2030. Currently, Tanzania’s low cost-competitiveness with trading partners results in the import of roughly 23 Million USD per year in dairy, and the DDR aims to replace these imports with domestic production. The researchers model how the Roadmap could be delivered through farmers changing from local to improved breeds of cattle, and feeding their cattle more nutritious, locally produced feed. Crucially, it assumes land that is already used in agriculture, mainly local pasture, is converted to grow feed crops, so farmers do not rely on imported feed and no forest needs to be cut down to grow it.

    “The idea was to model the Tanzanian Government’s planned interventions to increase milk production and also their targets for improved dairy breeds and feeding practices,” said Dr James Hawkins, an environmental economist from the Lancaster Environment Centre, and lead author of the study.

    “What is very important is understanding the interactions between cattle management and productivity because the carbon footprint is strongly related to the productivity of dairy cows.”

    The combination of more nutritious feed and more productive cattle means that production can be increased while reducing herd size, and cutting the amount of land needed to support the cattle, the study found. Better feed can increase the milk yield for local cows by up to 179% and for the higher yielding breeds by up to 130%. 

    The study modelled a series of scenarios, with different levels of milk production and adoption of new breeds and feeds. All the scenarios showed increases in production and a decrease in GHG emissions. The analysis showed that fulfilling the DDR targets for adopting improved breeds would enable Tanzania to meet 70% of the target milk production level while also fulfilling the country’s ambition to reduce GHG emissions from dairy by a third. The main driver of emissions reductions was from avoided land use change. While the model showed improving feed requires more cropland, a much larger decline in grasslands would reduce carbon dioxide emissions from forest clearance.

    “This is a win, win, win for Tanzania,” said Professor Mariana Rufino, from the Lancaster Environment Centre, principal investigator in the study, who has been researching dairy production in Africa for almost twenty years.

    “There have been a lot of studies showing how to mitigate emissions from the livestock sector that tell low-income countries what they should do, that they shouldn’t have livestock etc.  This study is special because we take Tanzania’s own ambitions, a country level target, and work out how they can achieve it, and more.

    “This Tanzanian policy only aims for food security, we find a way they can also improve incomes and meet their climate target at the same time.

    “Dairy is very good for poor communities. It generates daily cash, instead of farmers having to wait for a crop to be harvested once a year. There is a market for feed and lots of small businesses develop around dairy, so it generates income and alleviates poverty. Drinking milk can make a big difference to children in poor communities, providing a little bit of protein and concentrated micronutrients which they cannot get in other foods. So dairy can have a very important societal impact.”

    The income benefits are not equally distributed, the study shows, with farmers who do not have the resources to invest in, feed and care for higher yielding cattle losing out. The authors warn that while overall incomes increase, these farmers’ incomes could fall, especially if increasing production lowers the price they can charge for milk. Support policies should create safeguards for this key food sector, the researchers say.

    Dr Amos Omore, Tanzania country representative for the International Livestock Research Institute, said: “The findings of this paper have huge implications. The same quantity of milk being produced in smallholder dairy farms that dominate in eastern Africa can easily be produced with less than a quarter the number of animals currently, given the large yield gaps. What is required is more investment in sustainable animal productivity in smallholder farms – a clear win-win for better lives and greener planet.”

    The findings are outlined in the paper ‘High yield dairy cattle breeds improve farmer incomes, curtail greenhouse gas emissions and reduce dairy import dependency in Tanzania’

    Partners in the study include the Center for International Forestry Research (CIFOR) and the International Livestock Research Institute (ILRI), both in Kenya, and the Universities of Reading, Queensland and Wisconsin-Madison.

    [ad_2]

    Lancaster University

    Source link