Evaluating Privacy Leakage and Memorization Attacks on Large Language Models (LLMs) in Generative AI Applications

The recent interest in the deployment of Generative AI applications that use large language models (LLMs) has brought to the forefront significant privacy concerns, notably the leakage of Personally Identifiable Information (PII) and other confidential or protected information that may have been memorized during training, specifically during a fine-tuning or customization process. We describe different black-box attacks from potential adversaries and study their impact on the amount and type of information that may be recovered from commonly used and deployed LLMs. Our research investigates the relationship between PII leakage, memorization, and factors such as model size, architecture, and the nature of attacks employed. The study utilizes two broad categories of attacks: PII leakage-focused attacks (auto-completion and extraction attacks) and memorization-focused attacks (various membership inference attacks). The findings from these investigations are quantified using an array of evaluative metrics, providing a detailed understanding of LLM vulnerabilities and the effectiveness of different attacks.

Peritoneal fluid indocyanine green test for diagnosis of gut leakage in anastomotic leakage rats and colorectal surgery patients

BACKGROUND Application of indocyanine green(ICG)fluorescence has led to new developments in gastrointestinal surgery.However,little is known about the use of ICG for the diagnosis of postoperative gut leakage(GL).In addition,there is a lack of rapid and intuitive methods to definitively diagnose postoperative GL.AIM To investigate the effect of ICG in the diagnosis of anastomotic leakage in a surgical rat GL model and evaluate its diagnostic value in colorectal surgery patients.METHODS Sixteen rats were divided into two groups:GL group(n=8)and sham group(n=8).Approximately 0.5 mL of ICG(2.5 mg/mL)was intravenously injected postoperatively.The peritoneal fluid was collected for the fluorescence test at 24 and 48 h.Six patients with rectal cancer who had undergone laparoscopic rectal cancer resection plus enterostomies were injected with 10 mL of ICG(2.5 mg/mL)on postoperative day 1.Their ostomy fluids were collected 24 h after ICG injection to identify the possibility of the ICG excreting from the peripheral veins to the enterostomy stoma.Participants who had undergone colectomy or rectal cancer resection were enrolled in the diagnostic test.The peritoneal fluids from drainage were collected 24 h after ICG injection.The ICG fluorescence test was conducted using OptoMedic endoscopy along with a near-infrared fluorescent imaging system.RESULTS The peritoneal fluids from the GL group showed ICG-dependent green fluorescence in contrast to the sham group.Six samples of ostomy fluids showed green fluorescence,indicating the possibility of ICG excreting from the peripheral veins to the enterostomy stoma in patients.The peritoneal fluid ICG test exhibited a sensitivity of 100%and a specificity of 83.3%for the diagnosis of GL.The positive predictive value was 71.4%,while the negative predictive value was 100%.The likelihood ratios were 6.0 for a positive test result and 0 for a negative result.CONCLUSION The postoperative ICG test in a drainage tube is a valuable and simple technique for the diagnosis of GL.Hence,it should be employed in clinical settings in patients with suspected GL.

Establishment of a Rat Model of Capillary Leakage Syndrome Induced by Cardiopulmonary Resuscitation After Cardiac Arrest

Objective:Cardiopulmonary resuscitation(CPR)after cardiac arrest(CA)is one of the main causes of capillary leakage syndrome(CLS).This study aimed to establish a stable CLS model following the CA and cardiopulmonary resuscitation(CA-CPR)model in Sprague-Dawley(SD)rats.Methods:We conducted a prospective,randomized,animal model study.All adult male SD rats were randomly divided into a normal group(group N),a sham operation group(group S),and a cardiopulmonary resuscitation group(group T).The SD rats of the three groups were all inserted with 24-G needles through their left femoral arteries and right femoral veins.In group S and group T,the endotracheal tube was intubated.In group T,CA induced by asphyxia(AACA)was caused by vecuronium bromide with the endotracheal tube obstructed for 8 min,and the rats were resuscitated with manual chest compression and mechanical ventilation.Preresuscitation and postresuscitation measurements,including basic vital signs(BVS),blood gas analysis(BG),routine complete blood count(CBC),wet-to-dry ratio of tissues(W/D),and the HE staining results after 6 h were evaluated.Results:In group T,the success rate of the CA-CPR model was 60%(18/30),and CLS occurred in 26.6%(8/30)of the rats.There were no significant differences in the baseline characteristics,including BVS,BG,and CBC,among the three groups(P>0.05).Compared with pre-asphyxia,there were significant differences in BVS,CBC,and BG,including temperature,oxygen saturation(SpO_(2)),mean arterial pressure(MAP),central venous pressure(CVP),white blood cell count(WBC),hemoglobin,hematocrit,pH,pCO_(2),pO_(2),SO_(2),lactate(Lac),base excess(BE),and Na+(P<0.05)after the return of spontaneous circulation(ROSC)in group T.At 6 h after ROSC in group T and at 6 h after surgery in groups N and S,there were significant differences in temperature,heart rate(HR),respiratory rate(RR),SpO_(2),MAP,CVP,WBC,pH,pCO_(2),Na+,and K+among the three groups(P<0.05).Compared with the other two groups,the rats in group T showed a significantly increased W/D weight ratio(P<0.05).The HE-stained sections showed consistent severe lesions in the lung,small intestine,and brain tissues of the rats at 6 h after ROSC following AACA.Conclusion:The CA-CPR model in SD rats induced by asphyxia could reproduce CLS with good stability and reproducibility.

Three-dimensional forward modeling and response characteristics analysis of foundation pit leakage electric-field considering electrokinetic effect

The traditional ground direct current method is not suitable for leakage detection of underground diaphragm walls in foundation pits because of its low accuracy and poor anti-noise ability.Here,we propose a joint surface-borehole observation device for leakage electric fi eld detection to achieve rapid measurement of the electric fi eld distribution characteristics at ground level in the foundation pit,thus enabling rapid localization of leakage points.We first establish the mechanism and basic equation of the leakage electric field response by combining the electric field formed by electrokinetic effect(EK)and the stable electric fi eld formed by conduction current in a combined leakage channel.Then,the fi nite–infi nite element coupling method is used to solve the electric fi eld equation to simulate the responses of a three-dimensional foundation pit leakage model.Furthermore,we conduct numerical simulations of diff erent pit models to investigate the infl uencing factors of the detection device and response characteristics of the change in the properties of the leakage channel.The results demonstrate that the proposed joint surface-borehole observation device can effi ciently reveal anomalous potential caused by leakage,and the amplitude of the electric fi eld generated by EK can eff ectively strengthen the leakage electric fi eld signal at the leakage,thus improving detection accuracy and effi ciency.

Leakage Rate Model of Urban Water Supply Networks Using Principal Component Regression Analysis

To analyze the factors affecting the leakage rate of water distribution system, we built a macroscopic "leakage rate–leakage factors"(LRLF) model. In this model, we consider the pipe attributes(quality, diameter,age), maintenance cost, valve replacement cost, and annual average pressure. Based on variable selection and principal component analysis results, we extracted three main principle components—the pipe attribute principal component(PAPC), operation management principal component, and water pressure principal component. Of these, we found PAPC to have the most influence. Using principal component regression, we established an LRLF model with no detectable serial correlations. The adjusted R2 and RMSE values of the model were 0.717 and 2.067, respectively.This model represents a potentially useful tool for controlling leakage rate from the macroscopic viewpoint.

Modified model of gate leakage currents in AlGaN/GaN HEMTs

It has been reported that the gate leakage currents are described by the Frenkel-Poole emission(FPE) model,at temperatures higher than 250 K.However,the gate leakage currents of our passivated devices do not accord with the FPE model.Therefore,a modified FPE model is developed in which an additional leakage current,besides the gate(ⅠⅡ),is added.Based on the samples with different passivations,the ⅠⅡcaused by a large number of surface traps is separated from total gate currents,and is found to be linear with respect to(φB-Vg)0.5.Compared with these from the FPE model,the calculated results from the modified model agree well with the Ig-Vgmeasurements at temperatures ranging from 295 K to 475 K.

ESTABLISHMENT AND VERIFICATION OF MATHEMATICAL MODEL FOR CURRENT LEAKAGE IN BIPOLAR MULTI－COMPARTMENT CELL

ＥＳＴＡＢＬＩＳＨＭＥＮＴＡＮＤＶＥＲＩＦＩＣＡＴＩＯＮＯＦＭＡＴＨＥＭＡＴＩＣＡＬＭＯＤＥＬＦＯＲＣＵＲＲＥＮＴＬＥＡＫＡＧＥＩＮＢＩＰＯＬＡＲＭＵＬＴＩ－ＣＯＭＰＡＲＴＭＥＮＴＣＥＬＬＹａｎｇ；ＪｉａｎｈｏｎｇＺｈａｎｇ；ＱｉｎｇｘｉａｎｇＷａｎ...

Labyrinth seal leakage and dynamics characteristics analysis based on Black Model

One of the important problems to be tackled in turbo machines is the leakage dynamics characteristics of labyrinth seals. In this paper we analyzed the effect of labyrinth seal structure and the change in fluid flow pressure on the leakage characteristics of seal. Computational fluid dynamics (CFD) model for 3D labyrinth seal was built which provides a basis for reducing steam flow excitation. The streamline pattern and the pressure drop characteristics for leakage of steam through a labyrinth seal was investigated. Simulations of internal flow and leakage characteristics had been performed by CFD software and Black-Child model. The results showed that the amount of leakage is directly proportional to the tooth gap and inlet pressure and inversely proportional to the cavity depth and outlet pressure. The proposed CFD model provides a feasible method to predict the leakage characteristics of labyrinth seal in response to the structure of seal and the change in inlet-outlet pressures.

An improved lumped parameter model predicting attenuation of earmuff with air leakage

Since air leakage is inevitable when earmuffs are worn improperly or together with safety glasses in factory or military,it is required to be considered to accurately predict earmuff attenuation.Besides unwanted air leakage,under controlled air leakage is introduced to earmuff to achieve adjustable attenuations in different signal-to-noise ratios(SNRs)and balance between attenuation and speech intelligibility.This work is to develop an improved lumped parameter model(LPM)to predict earmuff attenuation with consideration of air leakage.Air leakage paths are introduced into conventional LPM without air leakage,and air leakage path impedance is analytically described by Maa’s microperforated tube impedance.Earmuff passive attenuation behavior can be analytically described and analyzed with the improved LPM.Finally,the validity of improved LPM is verified experimentally.The results indicate that the improved LPM can predict earmuff attenuation with air leakage,and air leakage deteriorates earmuff attenuation and turns resonance frequency higher.

Evaluation of Inverse Modeling Techniques for Pinpointing Water Leakages at Building Constructions

The location and nature of the moisture leakages are sometimes difficult to detect. Moreover, the relation between observed inside surface moisture patterns and where the moisture enters the construction is often not clear. The objective of this paper is to investigate inverse modeling techniques as a tool for the detection of moisture leakage locations in building constructions from inside surface moisture patterns. It is concluded that although the presented methodology is promising, more research is needed to confirm its usability.

A Review on Distribution Model for Mobile Agent-Based Information Leakage Prevention

With the continuous use of cloud and distributed computing, the threats associated with data and information technology (IT) in such an environment have also increased. Thus, data security and data leakage prevention have become important in a distributed environment. In this aspect, mobile agent-based systems are one of the latest mechanisms to identify and prevent the intrusion and leakage of the data across the network. Thus, to tackle one or more of the several challenges on Mobile Agent-Based Information Leakage Prevention, this paper aim at providing a comprehensive, detailed, and systematic study of the Distribution Model for Mobile Agent-Based Information Leakage Prevention. This paper involves the review of papers selected from the journals which are published in 2009 and 2019. The critical review is presented for the distributed mobile agent-based intrusion detection systems in terms of their design analysis, techniques, and shortcomings. Initially, eighty-five papers were identified, but a paper selection process reduced the number of papers to thirteen important reviews.

Study on Stack Effect of Stairwell by Numerical Model of Leakage Flow through Gap of Door

Since stack effect that occurs in high-rise buildings has an effect on the indoor environment of the buildings, energy loss and smoke control in case of a fire, there is a need to conduct research on this. For an analysis of the stack effect, analysis methods on the leakage flow through gap of interior door shall be formulated. Until now, studies related to the gap leakage flow in buildings have mainly analyzed flow field and pressure in the buildings one-dimensionally using pressure difference-leakage flowrate relations of Orifice Equation and a network numerical analysis algorithm that as- sumes each compartment in the buildings as a single point. In this study, the Momentum Loss Model which enables pressure drop to be proportional to the flow velocity through the gap of door in computational domain of 3-dimensional numerical analysis was proposed to reflect the gap flow phenomenon effectively in 3-dimensional numerical analysis. Using the proposed model, 3-dimensional numerical analysis of the stack effect on the stairs in buildings was performed, and the effects of separation door and lobby between stair and accommodation on the stack effect were investigated.

Mathematical model analysis of new-type magnetic integrated CRT considering magnetic leakage

A novel magnetic integrated controllable reactor of transformer type(CRT)has the advantages of simple structure,flexible assembly,convenient maintenance and practicability.To analyze its operation characteristics accurately,we establish corresponding equivalent mathematical model considering magnetic leakage based on magnetic circuit and circuit dualistic transformation method.The distribution of magnetic leakage field of each winding is analyzed qualitatively,and the analytical calculation formulas of magnetizing inductance and leakage inductance of each winding are derived.Based on this,the analytical calculation formulas of short-circuit impedance and winding current of CRT under different working conditions are derived.The field-circuit coupling finite element model of the magnetic integrated CRT is established to simulate the current of each winding under different working conditions.The results show that the analytical calculation results of each winding current have good consistency with the finite element calculation results,indicating the validity of CRT equivalent mathematical model and correctness of the analytical formulas of leakage inductance,short-circuit impedance and winding current of CRT.The working winding current of CRT is increasing gradually with the operation of control winding in turn to realise the transition of CRT compensation capacity from zero to a rated value.

Fine-Grain Sleep Transistor Insertion for Leakage Reduction

A fine-grain sleep transistor insertion technique based on our simplified leakage current and delay models is proposed to reduce leakage current. The key idea is to model the leakage current reduction problem as a mixed-integer linear programming （MLP） problem in order to simultaneously place and size the sleep transistors optimally. Because of better circuit slack utilization, our experimental results show that the MLP model can save leakage by 79.75%, 93.56%, and 94.99% when the circuit slowdown is 0%, 3%, and 5%, respectively. The MLP model also achieves on average 74.79% less area penalty compared to the conventional fixed slowdown method when the circuit slowdown is 7%.

Leakage Prediction Method for Contacting Mechanical Seals with Parallel Faces

Since the beginning of the 20th century, many researches on the sealing characteristic of mechanical seals were carried out broadly and in depth by various methods and some leakage models were built. But due to the lack of the way to characterize the main factors of influence on the leakage, most of the early researches were based on the assumptions that the seal faces topography and the frictional conditions were invariant. In the early built models, the effect of the surface topography change of the seal face on the leakage rate was neglected. Based on the fractal theory, the contact of end faces of the rotary and stationary rings was simplified to be the contact of a rough surface and an ideal rigid smooth surface, and the contact interface＇s cavity size-distribution function as well as the fractal characteristic of the cavity profile curve was discussed. By analyzing the influence of abrasion on the seal face topography and the leakage channel, the time-correlation leakage prediction model of mechanical seals based on the fractal theory was established and the method for predicting the leakage rate of mechanical seals with parallel plane was proposed. The values of the leakage rate predicted theoretically are similar to the measured values of the leakage rate in the model test and in situ test. The experimental results indicate that the leakage rate of mechanical seals is a transient value. The surface topography of the end faces of the seal tings and its change during the frictional wear of mechanical seals can be accurately characterized by the fractal parameters. Under the work conditions of changeless frictional mechanism, the fractal parameters measured or calculated based on the accelerated testing equation can be used to predict the leakage rate of mechanical seal in service. The proposed research provides the basis for determining the leakage state and predicting working life of mechanical seal.

PILOT LEAKAGE'S INFLUENCES ON THE PERFORMANCES OF EXTRA HIGH PRESSURE PROPORTIONAL PNEUMATIC VALVE

A mathematic model is built up to analyze the influences of a pilot valve'sleakage on the performances of pneumatic pressure proportional valve, and the performances aresimulated by using MATLAB. The results indicate that using slide pilot valve in the valve system isfeasible, but the leakage's influences can not be neglected, especially it may induce instability ina low output pressure situation. A pilot valve using too large throttle window will cause the valveoscillate. To improve the working condition of pilot valve, a method adopting different widths oftwo throttle window is proposed. According to our simulation, this method balances the pressure dropbetween the two stage throttle ports, and reduces the influences of pilot valve's leakage.

Prediction of Leakage Rates Through Sealing Connections with Nonmetallic Gaskets

In this work, a model of gas leakage through nonmetallic gaskets was developed in order to predict leakage rate of gasket sealing connections. The model was verified by the leakage experiments on two types of gaskets： compressed non-asbestos fiber gasket and flexible graphite gasket reinforced with tanged metal sheet. The coefficients in the leakage rate formula were obtained by regression of experimental data for each type of gasket, The model was also validated against the experimental leakage data by other researchers and shown to produce accurate predications. Furthermore, the model was applied to a bolted flanged connection in service in order to assess the tightness of the connection.

Tunnel SAM adapter:Adapting segment anything model for tunnel water leakage inspection

Water leakage inspection in the tunnels is a critical engineering job that has attracted increasing concerns.Leakage area detection via manual inspection techniques is time-consuming and might produce unreliablefindings, so that automated techniques should be created to increase reliability and efficiency. Pre-trainedfoundational segmentation models for large datasets have attracted great interests recently. This paper proposes a novel SAM-based network for accurate automated water leakage inspection. The contributions of thispaper include the efficient adaptation of the SAM (Segment Anything Model) for shield tunnel water leakagesegmentation and the demonstration of the application effect by data experiments. Tunnel SAM Adapter hassatisfactory performance, achieving 76.2 % mIoU and 77.5 % Dice. Experimental results demonstrate that ourapproach has advantages over peer studies and guarantees the integrity and safety of these vital assets whilestreamlining tunnel maintenance.

Numerical simulation of liquid aluminum leakage in casting process

A model was built to simulate liquid aluminum leakage during the casting process,including transient trough flow,orifice outflow,and spread,to prevent the explosion.A comparison between the simulation data and the theoretical calculation results verifies that the model has remarkable adaptability and high accuracy.Although the height of liquid aluminum in the mixing furnace and outlet radius are changed,the molten aluminum will not leak during the casting process.The aluminum in the trough moves forward in a wave-like motion and causes a leakage.The spread of the leaked aluminum resembles a long strip on the ground.The leakage amount and spread area of liquid aluminum increase with increasing the height of liquid aluminum in the mixing furnace.

Groundwater Leakage and River Runoff in a Catchment Influenced by Tectonic Movement

In order to clarify how groundwater leakage and river runoff occur in a catchment under tectonic movement, the water balance was estimated in the forested (88.3% in area) Oikamanai River catchment (area, 62.6 km2), Hokkaido, Japan. The catchment’s geology is early Miocene to Pliocene sedimentary bedrock of partly high permeability, accompanied by currently active faults. Daily evapotranspiration, E, in water balance was calculated by applying the one-layer model to meteorological data in the rainfall season of 2011 and 2012, with the topographic influence on heat balance of the catchment considered. The coupling with the short-term water balance method for river runoff events allows us to estimate groundwater leaking to the other catchments through the faults and bedrock. As a result, the leakage corresponded to 50% - 80% of effective rainfall (=P - E: P, rainfall) in 2011, whereas it was lower or negative in 2012. The estimate of leakage then included variability of ca. 80%. In 2012, shallow groundwater storage seems to retain high baseflow during non-rainfall.