scholarly journals Inference attacks against differentially private query results from genomic datasets including dependent tuples

2020 ◽  
Vol 36 (Supplement_1) ◽  
pp. i136-i145
Author(s):  
Nour Almadhoun ◽  
Erman Ayday ◽  
Özgür Ulusoy

Abstract Motivation The rapid decrease in the sequencing technology costs leads to a revolution in medical research and clinical care. Today, researchers have access to large genomic datasets to study associations between variants and complex traits. However, availability of such genomic datasets also results in new privacy concerns about personal information of the participants in genomic studies. Differential privacy (DP) is one of the rigorous privacy concepts, which received widespread interest for sharing summary statistics from genomic datasets while protecting the privacy of participants against inference attacks. However, DP has a known drawback as it does not consider the correlation between dataset tuples. Therefore, privacy guarantees of DP-based mechanisms may degrade if the dataset includes dependent tuples, which is a common situation for genomic datasets due to the inherent correlations between genomes of family members. Results In this article, using two real-life genomic datasets, we show that exploiting the correlation between the dataset participants results in significant information leak from differentially private results of complex queries. We formulate this as an attribute inference attack and show the privacy loss in minor allele frequency (MAF) and chi-square queries. Our results show that using the results of differentially private MAF queries and utilizing the dependency between tuples, an adversary can reveal up to 50% more sensitive information about the genome of a target (compared to original privacy guarantees of standard DP-based mechanisms), while differentially privacy chi-square queries can reveal up to 40% more sensitive information. Furthermore, we show that the adversary can use the inferred genomic data obtained from the attribute inference attack to infer the membership of a target in another genomic dataset (e.g. associated with a sensitive trait). Using a log-likelihood-ratio test, our results also show that the inference power of the adversary can be significantly high in such an attack even using inferred (and hence partially incorrect) genomes. Availability and implementation https://github.com/nourmadhoun/Inference-Attacks-Differential-Privacy

2020 ◽  
Vol 10 (1) ◽  
pp. 137-152
Author(s):  
Tosin A. Adesuyi ◽  
Byeong Man Kim

AbstractData is the key to information mining that unveils hidden knowledge. The ability to revealed knowledge relies on the extractable features of a dataset and likewise the depth of the mining model. Conversely, several of these datasets embed sensitive information that can engender privacy violation and are subsequently used to build deep neural network (DNN) models. Recent approaches to enact privacy and protect data sensitivity in DNN models does decline accuracy, thus, giving rise to significant accuracy disparity between a non-private DNN and a privacy preserving DNN model. This accuracy gap is due to the enormous uncalculated noise flooding and the inability to quantify the right level of noise required to perturb distinct neurons in the DNN model, hence, a dent in accuracy. Consequently, this has hindered the use of privacy protected DNN models in real life applications. In this paper, we present a neuron noise-injection technique based on layer-wise buffered contribution ratio forwarding and ϵ-differential privacy technique to preserve privacy in a DNN model. We adapt a layer-wise relevance propagation technique to compute contribution ratio for each neuron in our network at the pre-training phase. Based on the proportion of each neuron’s contribution ratio, we generate a noise-tuple via the Laplace mechanism, and this helps to eliminate unwanted noise flooding. The noise-tuple is subsequently injected into the training network through its neurons to preserve privacy of the training dataset in a differentially private manner. Hence, each neuron receives right proportion of noise as estimated via contribution ratio, and as a result, unquantifiable noise that drops accuracy of privacy preserving DNN models is avoided. Extensive experiments were conducted based on three real-world datasets and their results show that our approach was able to narrow down the existing accuracy gap to a close proximity, as well outperforms the state-of-the-art approaches in this context.


2019 ◽  
Author(s):  
Nour Almadhoun ◽  
Erman Ayday ◽  
Özgür Ulusoy

Abstract Motivation The rapid progress in genome sequencing has led to high availability of genomic data. However, due to growing privacy concerns about the participant’s sensitive information, accessing results and data of genomic studies is restricted to only trusted individuals. On the other hand, paving the way to biomedical discoveries requires granting open access to genomic databases. Privacy-preserving mechanisms can be a solution for granting wider access to such data while protecting their owners. In particular, there has been growing interest in applying the concept of differential privacy (DP) while sharing summary statistics about genomic data. DP provides a mathematically rigorous approach but it does not consider the dependence between tuples in a database, which may degrade the privacy guarantees offered by the DP. Results In this work, focusing on genomic databases, we show this drawback of DP and we propose techniques to mitigate it. First, using a real-world genomic dataset, we demonstrate the feasibility of an inference attack on differentially private query results by utilizing the correlations between the tuples in the dataset. The results show that the adversary can infer sensitive genomic data about a user from the differentially private query results by exploiting correlations between genomes of family members. Second, we propose a mechanism for privacy-preserving sharing of statistics from genomic datasets to attain privacy guarantees while taking into consideration the dependence between tuples. By evaluating our mechanism on different genomic datasets, we empirically demonstrate that our proposed mechanism can achieve up to 50% better privacy than traditional DP-based solutions. Availability https://github.com/nourmadhoun/Differential-privacy-genomic-inference-attack. Supplementary information Supplementary data are available at Bioinformatics online.


2020 ◽  
pp. 104687812097274
Author(s):  
Syretta Spears ◽  
Gabriel M. Díaz ◽  
Desiree A. Diaz

Background. Escape rooms have been utilized to incorporate teamwork, communication, policy, and procedures reinforcement, as well as clinical care. Often themed, escape rooms challenge a team of players to discover clues and sequentially solve puzzles with the ultimate task of accomplishing a specific goal in a limited time frame, fostering critical thinking. Method. This learning activity incorporates legal implications for prelicensure nursing students in a pediatric community summer camp/after school care environment. Care and legal implications for a minor experiencing respiratory distress was the premise of this escape room. Outcomes. While this was not a research study, surveys were distributed for quality improvement and a deeper needs assessment related to the content and delivery of an escape room. Learners discussed the positive aspects of this activity through journal entries and a survey, noting the need for increased content related to legal implications for the medical team. This room engaged the learner in the forward-thinking that is needed related to emergency care, Good Samaritan Act, liability for cost, and permission to treat in a time-pressured environment. Future Plans. Testing the knowledge pre-post survey related to legal implications with implied care will be explored.


2021 ◽  
pp. 1-13
Author(s):  
Shen Li ◽  
Shi Yu Chan ◽  
Amy Higgins ◽  
Mei-Hua Hall

Abstract Background Diminished sensory gating (SG) is a robust finding in psychotic disorders, but studies of early psychosis (EP) are rare. It is unknown whether SG deficit leads to poor neurocognitive, social, and/or real-world functioning. This study aimed to explore the longitudinal relationships between SG and these variables. Methods Seventy-nine EP patients and 88 healthy controls (HCs) were recruited at baseline. Thirty-three and 20 EP patients completed 12-month and 24-month follow-up, respectively. SG was measured using the auditory dual-click (S1 & S2) paradigm and quantified as P50 ratio (S2/S1) and difference (S1-S2). Cognition, real-life functioning, and symptoms were assessed using the MATRICS Consensus Cognitive Battery, Global Functioning: Social (GFS) and Role (GFR), Multnomah Community Ability Scale (MCAS), Awareness of Social Inference Test (TASIT), and the Positive and Negative Syndrome Scale (PANSS). Analysis of variance (ANOVA), chi-square, mixed model, correlation and regression analyses were used for group comparisons and relationships among variables controlling for potential confounding variables. Results In EP patients, P50 ratio (p < 0.05) and difference (p < 0.001) at 24-month showed significant differences compared with that at baseline. At baseline, P50 indices (ratio, S1-S2 difference, S1) were independently associated with GFR in HCs (all p < 0.05); in EP patients, S2 amplitude was independently associated with GFS (p = 0.037). At 12-month and 24-month, P50 indices (ratio, S1, S2) was independently associated with MCAS (all p < 0.05). S1-S2 difference was a trending predictor of future function (GFS or MCAS). Conclusions SG showed progressive reduction in EP patients. P50 indices were related to real-life functioning.


Author(s):  
Poushali Sengupta ◽  
Sudipta Paul ◽  
Subhankar Mishra

The leakage of data might have an extreme effect on the personal level if it contains sensitive information. Common prevention methods like encryption-decryption, endpoint protection, intrusion detection systems are prone to leakage. Differential privacy comes to the rescue with a proper promise of protection against leakage, as it uses a randomized response technique at the time of collection of the data which promises strong privacy with better utility. Differential privacy allows one to access the forest of data by describing their pattern of groups without disclosing any individual trees. The current adaption of differential privacy by leading tech companies and academia encourages authors to explore the topic in detail. The different aspects of differential privacy, its application in privacy protection and leakage of information, a comparative discussion on the current research approaches in this field, its utility in the real world as well as the trade-offs will be discussed.


Author(s):  
Shuo Han ◽  
George J. Pappas

Many modern dynamical systems, such as smart grids and traffic networks, rely on user data for efficient operation. These data often contain sensitive information that the participating users do not wish to reveal to the public. One major challenge is to protect the privacy of participating users when utilizing user data. Over the past decade, differential privacy has emerged as a mathematically rigorous approach that provides strong privacy guarantees. In particular, differential privacy has several useful properties, including resistance to both postprocessing and the use of side information by adversaries. Although differential privacy was first proposed for static-database applications, this review focuses on its use in the context of control systems, in which the data under processing often take the form of data streams. Through two major applications—filtering and optimization algorithms—we illustrate the use of mathematical tools from control and optimization to convert a nonprivate algorithm to its private counterpart. These tools also enable us to quantify the trade-offs between privacy and system performance.


Author(s):  
Usman Naeem ◽  
Richard Anthony ◽  
Abdel-Rahman Tawil ◽  
Muhammad Awais Azam ◽  
David Preston

We live in a ubiquitous world where we are surrounded by context sensitive information and smart devices that are able to capture information about our surroundings unobtrusively. Making use of such rich information can enable recognition of activities conducted by elderly users, and in turn can allow the possibility of tracking any functional decline. This chapter highlights the current methods for unobtrusively recognising activities of daily living within a home environment for people with physical or cognitive disabilities. A main group for which this is important for are Alzheimer's patients. The chapter also bases the discussion of what makes a successful environment for carrying out accurate activity recognition, which is then followed by a proposed taxonomy of the key characteristics that are required for robust activity recognition within a smart environment, contextualised with real-life scenarios.


2020 ◽  
Vol 46 (Supplement_1) ◽  
pp. S85-S86
Author(s):  
Maija Walta ◽  
Heikki Laurikainen ◽  
Reetta-Liina Armio ◽  
Tiina From ◽  
Raimo K R Salokangas ◽  
...  

Abstract Background Attrition rates and sampling bias in controlled clinical studies are a concern when evaluating the relevance of the results to a specific patient population in a real-life clinical / treatment setting. Dropout rates in studies on psychotic disorders are high and many eligibility criteria may lead to bias in study samples. We wanted to analyze how representative are the patient samples typically included in first-episode psychosis studies such as the Turku Early Psychosis (TEPS) study by using a platform of 3772 consecutive admissions to clinical psychiatric services of Turku Psychiatry. Methods TEPS study was started in 2011 as a part of a larger study on psychosis treatment processes in Turku Psychiatric services. Each patient, inpatient and outpatient, went through initial clinical screening by the treatment group which was followed by a structured evaluation if the screen for first-episode psychosis was positive. Between Oct 2011 and June 2016 there were 195 patients with first-episode psychosis (FEP) suitable to the TEPS study. Of them 102 were willing and 93 were not willing to participate or were not reached in a baseline structured evaluation. Using patient records, we compared if these two groups differed in terms of clinical variables, treatment or prognosis during a 1-year follow-up. Time of hospital stay, involuntary vs. voluntary admission, coercive measures during the hospital care, re-hospitalizations and drop-out from the clinical care during the follow-up were used as outcomes. Results Non-participating (NTP) group had higher rate of involuntary care than participating (TP) group (70 % vs 62 %) as well as higher rate of coercion during the treatment and higher rate of re-admissions during the follow-up than the TP group (36 % vs 22 % and 41 % vs 34 %, respectively) but these differences did not reach statistical significance. During the one-year follow-up NTP group had a significantly higher rate of dropping out from the clinical care than participating TP group (48 % vs 30 %, p=0.01). NTP group had also higher rate of dropping out of clinical treatment mainly because of patient non-adherence (33 % vs 16 %, p=0.03). Discussion Nearly half (47 %) of the intent-to-study FEP patients were not reached or declined to participate in our study. Non-participating patients had a slightly more severe illness and poorer treatment adherence during one-year follow-up. The clinical differences were not as marked as we expected. E.g. involuntary care, inpatient care and more coercion during the follow-up were not significantly different between NTP and TP groups. Nevertheless, the data suggest considerable differences between participating and non-participating patients with first-episode psychosis which should be taken in to account when evaluating the generalizability of the results for an unselected group of psychotic patients in ‘real-life’ clinical care.


2018 ◽  
Vol 3 (3) ◽  
pp. 2473011418S0025
Author(s):  
Jeff Houck ◽  
Jillian Santer ◽  
Kostantinos Vasalos ◽  
Judith Baumhauer

Category: Other Introduction/Purpose: New instruments like the Patient Reported Outcome Information System (PROMIS) minimize the burden to patients and providers addressing significant barriers to adoption. Despite these advances provider adoption remains lackluster. Models of technology adoption suggest adoption is more likely to occur when PRO’s directly improve patient care (performance expectancy) and it’s easy to implement (effort expectancy). Problems with effort expectancy are dealt with by training and improving logistics (i.e. eHR presentation, alerts), where performance expectancy is addressed through research (i.e. validation of thresholds). The purposes of this study were to: 1) evaluate the proportion of orthopedic rehabilitation providers who use PRO’s and how they use them; And, 2) to determine if performance expectancy, effort expectancy or provider burnout are related to provider use. Methods: Fifty rehabilitation providers (physical therapist and athletic trainers) anonymously completed the electronic PRO Adoption Survey. Participants were 23.4±5.8 years old and 54% were female. The purpose of the PRO Adoption Survey is to track adoption across health systems. The first section of the PRO Adoption survey includes whether providers use PRO’s and asks them to detail how they use them (Table 1). A factor analysis supported the use of sets of questions to determine performance expectancy and effort expectancy (Table 1). Performance expectancy captures the health benefits the provider expects to experience. Effort expectancy captures the provider’s expectations of how easy it will be to implement PRO tools. The validated Maslach-2 burnout scale (BO) was included as another a factor that may influence adoption. Proportions and chi square tests were used to describe provider use of PRO’s and its relationship with performance expectancy, effort expectancy, and burnout. Results: The profile of PRO use by rehabilitation professionals is that a majority know about PRO’s (86%) however only 34% utilize PRO’s during clinic visits (Table 1). The most common PRO used is PROMIS (83%), followed by generic measures (41%) and disease specific (29%) measures. Type of use indicated the most common use was to make clinical decisions (71%) with relatively few using it for research (12%). Interestingly, 47% of PRO users review data with patients. The average responses for performance expectancy were 3.9 ± 0.1. The average responses for effort expectancy were 3.2 ± 0.2 or “neutral”. The average BO score was 4.6 ± 1.0. Chi square analysis suggested performance expectancy, effort expectancy, and burn out were not significantly associated with provider use. Conclusion: PROMIS scales are currently available in the electronic medical record(eMR) leading to high use (86%) by current PRO users (34%). High performance expectancy scores (~4/5) and low BO suggest providers can be motivated to use PRO’s. However, providers are neutral (~3/5) on how easy PRO’s would be to implement. Also, lower scores for performance expectancy associated with “aggregate” PRO data (only 54% marked “Agree” for this item) suggests training on specific uses of aggregate data are also indicated. These data detail the real issues providers need addressed to effectively capitalize on the benefits of PRO’s to improve clinical care.


2013 ◽  
Vol 9 (3-4) ◽  
pp. 131-146 ◽  
Author(s):  
J.M. Williams ◽  
D.M. Marlin ◽  
N. Langley ◽  
T.D. Parkin ◽  
H. Randle

The Grand National (GN) attracts high profile press and subsequent public attention. This study aimed to establish if factors influential to non-completion, horse-falls and specific fence risk in the GN supported the measures implemented by the British Horseracing Authority (BHA) to improve equine welfare in the GN. Horse, jockey, trainer and race related factors associated with non-completion, horse-falls and horse-falls at specific fences of the GN were collated over a 22 year period from 1990 to 2012. Descriptive analysis calculated non-completion rates per year, according to age and reason for non-completion. The distribution of fallers during the race in relation to fence number, design and key feature fences were also determined. Univariable analysis informed multivariable model building to identify factors associated with non-completion (n=840) and horse-falls (n=514) in the GN. Two final logistic regression models were refined through a backward stepwise process with variables retained if likelihood ratio test P-values were <0.05. Chi-square goodness of fit analyses evaluated fall risk at fence level. During the period investigated 347 horses completed the GN; the probability of a horse falling in the race was 0.24. The first fence, Becher's brook and drop fences increased the risk of falling compared to plain fences. Good-soft going increased the number of horses that completed the race and reduced the number of fallers suggesting this is the optimal ground condition for the race. GNs run at a faster than average speed increased the risk of horses not completing and falling. Inexperienced horses and jockeys show a greater risk of not completing and falling. Our work supports BHA measures implemented to improve safety in the GN; controlling speed, modifying fence design, promoting race experience and ground maintenance to produce good-soft going can increase completions and reduce falls, therefore enhancing equine welfare.


Sign in / Sign up

Export Citation Format

Share Document