A dream of Soaring Higher! We are thrilled to introduce you to Team Maverick India, a dynamic aero team from Pimpri Chinchwad College of Engineering, Pune. Over the past 3 years, we are a dedicated group of students who have been soaring towards the skies ...

A Flight to Victory!

Chandigarh is home to some of the best cancer hospitals and medical centres in India. These hospitals are equipped with… Read More »Top 10 Cancer Hospitals in Chandigarh

Darshan Supekar

In recent years, crowdfunding services have exploded in popularity. In the wake of the global economic crisis, crowdsourcing has become an important source of fundraising. Crowdfunding is a powerful tool for empowering people and organizations to make a ...

Online Donations Through Crowdfunding And Fundraising

Mumbai, a metropolitan city known for its world-class hospitals, offers a range of options for cancer treatment. The city is renowned for its exceptional specialists and advanced therapies, making it a highly sought-after destination for the best cancer ...

Top 10 Cancer Hospitals in Mumbai To Get the Best Cancer Treatment

APPEAL BY PARENTS FOR THEIR SON “MASTER. RAJ HANUMANT DUKARE” FOR HIS TREATMENT, HE IS SUFFERING FROM “LARGE OSTIUM SECUNDUM ASD”. My Son Master. Raj is just 11 Years old boy and he is diagnosed with “Large Ostium Secundum ASD” RAJ has to ...

Support Master. Raj To Get Healthy Heart

We are the health and disease group from CP Goenka International School, Mumbai. Though medicine had advanced, there are many diseases which have no cure. We aim to raise Rs. 50,000 to help make kids, who are suffering from incurable or life-threatening ...

School kids come together to raise funds to help make wishes come true

Fuel Dreams, Ignite Change: Empower Girls Now! Unleash girl power like never before. Your donation isn't just a drop in the bucket – it's a tidal wave of transformation, unlocking a world of possibilities: Championing Potential: Your support fuels high ...

Fuelling the dreams of 2000 athletes!

I am Aditya Reckwar, a 2022 Electrical Engineering graduate from of Delhi Technological University and lived in JCB Hostel . Today I am seeking your support to help me survive tragic health conditions and I am hardly making my ends meet. Through is fund ...

Aditya have no family support, he needs our support

A true embodiment of strength and compassion, Haveena founded "Her Holidays," a women's only travel agency, with the vision of empowering women through immersive and life-changing travel experiences. Beyond her entrepreneurial pursuits, Haveena ventured ...

Embracing Hope: Help Haveena’s Journey to Recovery

32-year-old Mohit Yadav, a bus conductor, passed away, leaving behind his parents, wife, four-year-old son and sister. A loving husband, a doting father, a supportive brother, and a caring son, Yadav has left behind a void that is impossible to fill.

Mohit Yadav, A Bus Conductor, Passed Away, Leaving Behind His Loving Family. They Need Your Support.

Ketto

ketto

You got a perfect product, been used by about a million of users, so does it mean your product is successful? Well, if you think right now it’s on the path of success, but it might just take a small turn to go down the road to failure. Product just can’t be successful and keep its game up for a longer run unless it exploit the behavioural domain of its users. Every product which became the epitome of human creation, for instance Facebook or Google, which left this era completely imprinted by its existence, have cleverly utilised cognitive and behavioural psychology of its user base. Oh! yes, being a manager or a product designer you might have noticed it, but still didn’t apply it to your product. Another factor is you might be unknowingly doing it but are not able to get optimum return from it. Let’s first talk about the behavioural psychology importance in product success and its efficiency in getting the product to the next level. Early this December, my mother was in a shopping mall when a couple of girls approached her and gifted her handmade Christmas greeting cards and flowers, plus the perfunctory request for a donation. She typically complains that this kind of stuff happens every year, but then she reaches for her purse. When I tell her she can simply ignore the request, she shakes her head: “But they already gave it to me!” What compelled her to donate to a charity she isn’t really passionate about? Reciprocity Technique, We Give Tit for Tat What happened with my mother is a simple case of : “I’ll do it because you already gave me something.” A moral obligation instilled by social pressure to do something back for those children who made those season cards. We are naturally wired to want to return favours. Reciprocity If you do something nice for someone, or give them something as a gift, they will feel indebted to you for it. A tried-and-tested way in a typical E-Commerce industry is to give something extra to instil a delight in the customer, which in return gets them coming back to you again and again and sometime even much more. For instance, a renowned e-payments and e-commerce brand in India, , occasionally rewards its users with cashback credited to their e-wallet. While a closer look shows it provides such to delight a specific set of users. PayTm The same way at my previous company , a premium Rental service for branded & designer clothing for Men & Women in India (similar to ), we provided surprise bundled accessories to instil delight and along with that provide a placard with social coordinates to share their experiences on their social media tagging us. Flyrobe Rent the Runway And it generated huge social confirmations, branding for us, with loyal customers as cherry on the cake. Door-in-the-Face or Rejection-then-retreat technique The other technique used sparingly is “ technique or otherwise known as the “ technique. The persuader attempts to convince the respondent to comply by making a large request that the respondent will most likely turn down, much like a metaphorical slamming of a door in the persuader’s face. The respondent is then more likely to agree to a second smaller request. The-door-in-the-face” Rejection-then-retreat” Popup shown when user tries to exit the page - Image Source: Ketto A quick example for this is used by popular charity, social funding sites. For instance: At my current company , a very popular online Crowdfunding Platform and Website in India for fundraising of Social, Charity, Personal and Creative causes, we ask the user to at least share the fundraiser page to their social walls, if they are not willing to donate to the fundraiser, which they agree to as this is a smaller request than the first one which involved making a donation. Ketto (a crowdfunding site similar to GoFundMe ) The way it’s presented completely utilises the “Rejection-then-retreat technique”. Visitors on the fundraiser page are shown an exit- intent (when they try to leave the site- refer the above screenshot) with a message from the campaigner asking them to at least share their fundraiser, if not making a donation. The same technique is utilised at the payment page where they are provided with preset amount slabs to choose from, which they wanted to donate. In case user tries to leave the site, they are shown message that even a contribution as small as INR 100 ($15 for people outside India) can help their cause. If you look at the screenshot above (the pop up on Ketto’s page), there’s no close button present on it. the way to close the pop up is by clicking “I don’t wish to share”. This is a cleverly deployed tactic, where the directive actionable text plays an important role in changing user behavior. User experiences a and to minimise that they click the share button. The use of specific negate term “don’t” in the clickable text, makes the user think twice and create a sense of uneasiness through guilt. cognitive dissonance Content will adhere the throne if only you chose the right one. As it played an important part in altering user behavior above. The fable of “ ”, can be seen as an example of . The fox held two incompatible thoughts simultaneously, desire and frustration. The Fox and the Grapes cognitive dissonance The criticism used by the fox was a way to neutralize this dis-balance and reduce the mental conflict. Humans subconsciously strive for internal consistency. Experiencing this leads to discomfort. This leads to alteration of behavior, so that we can stay in balance and be happy. Foot-in-the-Door technique Another powerful compliance technique is called “ ” technique. It aims at getting a person to agree to a large by having them agree to a modest first. Foot-in-the-door request request Imagine, working on the onboarding flow of your product. Before asking the users for a big favor ( ), you first ask them for a small favor ( ), a favor so small, they’ll almost surely do it. become a paid member register for free with one click Once they’ve set foot in the door and you wait for a bit, you can ask them for a larger favor. In the end, you can ask them the big favor to become a paid member of your product. Use of various social psychology technique in UX from one of the most popular e-commerce site eBay Let’s go through the above screenshot, and unravel some of social psychology techniques used by various E-commerce sites. People are more likely to complete tasks when they see other people doing them. is a psychological phenomenon where people reference the behavior of others to guide their own behavior. The idea: “I’ll do it because everyone seems to be doing it too.” Social Proof or Conformity This tendency is driven by our natural desire to behave “correctly” under most circumstances — whether making a purchase, deciding where to dine, determining where we should go, what we say, who we say it to, and so on. One of the best examples of social proof, in real life, is the long line in front of an Apple Store on the day a new iPhone is released. The fact that a group of people find the new phone so desirable as to invest considerable time standing (or sleeping!) in line impacts our perception of the phone value (and makes us covet one, too). Apple Store UK Opening Day photo by Lucius Kwok. Used via Creative Commons License. Looking towards the web we can see the social proof principle used frequently — for example ratings and review options on many e-commerce sites. “people who bought this item also bought” is a good example of this. Twitter also shows what’s “Trending”. Amazon’s People see these popular links and often click on stuff they otherwise would never have thought about. The most prevalent example of this is probably ability to sort by “best selling” which is available on many e-commerce sites. Going back to the above screen shot of , “add to watchlist” and “number of views per hour” are examples of this. eBay Scarcity: Value what is rare, we react to urgent deadlines & dwindling supplies. Image Source: Souq So my younger brother recently happen to visit a very popular E-Commerce website, to buy iPhone X. He was on a trip to Dubai, and his intentions were to buy it just before he gets back. Souq But, he was inadvertently forced to place the order instantly after visiting the page. When looking closer to what changed his behavior was a small text placed below the add to cart button, “Only 1 left in stock!”. He was a victim of a very well known social phenomenon “ ”. The scarcity principle The idea: “I’ll do it because it’s my last chance.” The causes people to assign high value to things they perceive as being less available. In real life, Black Friday is a good example of scarcity: a sale that occurs on only one day of the year (the day after Thanksgiving in the United States) and consists of a limited number of products offered at discounted prices. scarcity principle Ways to effectively utilise the scarcity principle: : Create price deals that last for a limited time, whether this means making a purchase on your app, or signing up for a promotion. Providing discounts : We’ve all seen the countdown timers on , ticking away the seconds before the deal disappears. Display it and place that urgency right in your users’ faces. Use countdown timers Amazon : For events or physical goods, you can display a specific number of units or reserved seats remaining. This is one reason why travel apps always show “Only X seats left” when you’re trying to book an airline ticket. Display units remaining Make it a special customer/buyer group with additional benefits and offer limited seats through subscription or signups.Now keep that scoreboard ticking, and keep crossing each Level! Limited Inclusion: Now that you have an arsenal of tactics you can use to persuade users to engage. Just remember that it is your duty as an ethical Product designer, UX designer or marketer to use these principles responsibly. Don’t exploit them in the short run for profit and dispense them when you’re done. Play the long game and persuade your users toward action while building true loyalty for — and engagement with — your brand. Call to action If you found this article useful share with others so that they can enjoy it, too. Thanks for your time! Connect with me on LinkedIn and Medium . Stay tuned for more UX and cognitive psychology posts till then Happy Reading!

Levelling Up Your Conversion Metrics Via Behavioural Psychology

Things To Make by Archibald Williams is part of the HackerNoon Books Series. You can jump to any chapter in this book . QUICK BOILING KETTLES. here XIX. QUICK BOILING KETTLES. [Transcriber’s note: Do not use lead solder on articles associated with human or animal consumption.] The principles of increasing the area of heating surface in model boilers may be applied very practically to the common kettle. The quick-boiling kettle is useful for camping out, for heating the morning tea water of the very early riser, and for the study “brew,” which sometimes has to be made in a hurry; and, on occasion, it will be so welcome in the kitchen as to constitute a very useful present to the mistress of the house. As the putting in of the tubes entails some trouble, it is worth while to select a good kettle for treatment. Get one that is made of thick tinned sheet iron (cast-iron articles are unsuitable), or even of copper, if you are intent on making a handsome gift which will last indefinitely. The broad shallow kettle is best suited for tubing, as it naturally has a fair heating surface, and its bottom area gives room for inserting plenty of tubes. Also, the tubes can be of good length. Let us, therefore, assume that the kettle will be of at least 8 inches diameter. In Figs. 96 (a) and 96 (b) are shown two forms of fire-tube kettles (a and b) and two of water-tube (c and d). For use over a spirit or Swedish petroleum stove the first two types are most convenient; the third will work well on a stove or an open fire; and the last proves very efficient on an open fire. One may take it that, as a general rule, areas of heating surface being equal, the water-tube kettle will boil more quickly than the fire-tube. Fire-tube Kettles. The tubing of Figs. 96 (a) and 96 (b) presents a little difficulty in each case. The straight tube is the more difficult to insert, owing to the elliptical shape of the ends; whereas the bent tube requires only circular holes, but must be shaped on a template. The tubing used for (a) should have at least 5/8-inch internal diameter, for (b) 1/2 inch, and be of thin copper. Hot gases will not pass willingly through tubes much smaller than this, in the absence of induced or forced draught. For convenience in fitting, the tubes should run at an angle of 45 degrees to the bottom and side of the kettle, as this gives the same bevel at each end. Find the centre of the bottom, and through it scratch plainly four diameters 45 degrees apart. From their ends draw perpendiculars up the side of the kettle. [Illustration: FIG. 96 (a).] Now draw on a piece of paper a section of the kettle, and from what is selected as a convenient water-level run a line obliquely, at an angle of 45 degrees, from the side to the bottom. Measuring off from this diagram, you can establish the points in the side and bottom at which the upper and longer side of the tubes should emerge. Mark these off. Next bevel off a piece of tubing to an angle of 45 degrees, cutting off roughly in the first instance and finishing up carefully with a file till the angle is exact. Solder to the end a piece of tin, and cut and file this to the precise shape of the elliptical end. Detach by heating, scribe a line along its longest axis, and attach it by a small countersunk screw to the end of a convenient handle. Place this template in turn on each of the eight radii, its long axis in line with it, being careful that the plate is brought up to the marks mentioned above, and is on the bottom corner side of it. Scratch round plainly with a fine steel point. To remove the metal for a tube hole, it is necessary to drill a succession of almost contiguous holes as near the scratch as possible without actually cutting it. When the ring is completed, join the holes with a cold chisel held obliquely. Then file carefully with a round file, just not cutting the scratch. As the side of the hole nearest to the bottom corner should run obliquely to enable the tube to pass, work this out with the file held at an angle. As soon as a pair of holes (one in the bottom, the other in the side) have been made, true up the side hole until a piece of tubing will run through it at the correct angle. Then bevel off the end to 45 degrees and pass the tube through again, bringing the bevel up against the bottom hole from the inside. If it is a trifle difficult to pass, bevel off the edge slightly on the inside to make a fairly easy driving fit. (Take care not to bulge the bottom of the kettle.) Mark off the tube beyond the side hole, allowing an eighth of an inch extra. Cut at the mark, and number tube and hole, so that they may be paired correctly later on. When all the tubes are fitted, “tin” the ends with a wash of solder before returning them to their holes. If there is a gap at any point wide enough to let the solder run through, either beat out the tube from the inside into contact, or, if this is impracticable, place a bit of brass wire in the gap. Use powdered resin by preference as flux for an iron kettle, as it does not cause the rusting produced by spirit of salt. If the latter is used, wipe over the solder with a strong ammonia or soda solution, in order to neutralize the acid. As the hot gases may tend to escape too quickly through large tubes, it is well to insert in the upper end of each a small “stop,” x—a circle of tin with an arc cut away on the bottom side. To encourage the gases to pass up the tubes instead of along the bottom, a ring of metal, y, may be soldered beyond the bottom holes, if an oil or spirit stove is to be used. This ring should have notches cut along the kettle edge, so as not to throttle the flame too much. [Illustration: FIG. 96—(b), (c), and (d).] As the tubes for these require bending to shape in each case, the three types may be grouped together. The tubes of c and d, which require bending to somewhat sharp curves, may be of 3/8-inch internal diameter. In the last two cases the direction of the water travel is shown. The up-flow end, which projects farther through the bottom than the down-flow, is nearer the centre, where, if a gas stove is used, the heat is more intense than at the circumference of the bottom. (Note.-If type c is for use on a three-support stove, increase the number of tubes to 9, equally spaced, 40 degrees apart, so that the kettle may be adjusted easily.) The copper tubing should be annealed or softened by heating to a dull red and plunging in cold water. Cut a wooden template of the exact outline of the inside line of the shape that the tube is to assume, and secure this firmly to a board. Fill the tube with melted resin, to prevent, as much as possible, “buckling” or flattening on the curves. The tube must be kept up to the template by a stop of hard wood, at the end at which bending commences. Don’t cut the tube into lengths before bending, as short pieces are more difficult to handle. When a piece sufficient for a tube has been bent, cut it oft, and remove the resin by heating. The fitting of the tubes is an easy matter, as the holes are circular. Pair off a tube with its holes and number it. A fluted reamer will be found invaluable for enlarging them to the correct size. Tin all tubes at points where they are to be attached to the kettle. In Fig. 96 (c) and (d) care should be taken to make all the tubes project the same distance, so that the kettle may be level when resting on them. About HackerNoon Book Series: We bring you the most important technical, scientific, and insightful public domain books. This book is part of the public domain. Archibald Williams (2005). Things To Make. Urbana, Illinois: Project Gutenberg. Retrieved https://www.gutenberg.org/cache/epub/14664/pg14664-images.html This eBook is for the use of anyone anywhere at no cost and with almost no restrictions whatsoever. You may copy it, give it away or re-use it under the terms of the Project Gutenberg License included with this eBook or online at , located at . www.gutenberg.org https://www.gutenberg.org/policy/license.html

QUICK BOILING KETTLES.

The Story-book of Science by Jean-Henri Fabre, is part of the HackerNoon Books Series. You can jump to any chapter in this book here . THE KETTLE CHAPTER XI. THE KETTLE NOW, that day, Mother Ambroisine was very tired. She had taken down from their shelves kettles, saucepans, lamps, candlesticks, casseroles, pans, and lids. After having rubbed them with fine sand and ashes, then washed them well, she had put the utensils in the sun to dry them thoroughly. They all shone like a mirror. The kettles particularly were superb with their rosy reflections; one might have said that tongues of fire were shining inside them. The candlesticks were a dazzling yellow. Emile and Jules were lost in admiration. “I should like to know what they make kettles of, they shine so,” remarked Emile. “They are very ugly outside, all black, daubed with soot; but inside, how beautiful they are!” “You must ask Uncle,” replied his brother. “Yes,” assented Emile. No sooner said than done: they went in search of their uncle. He did not have to be entreated; he was happy whenever there was an opportunity to teach them something. “Kettles are made of copper,” he began. “And copper?” asked Jules. “Copper is not made. In certain countries, it is found already made, mixed with stone. It is one of the substances that it is not in the power of man to make. We use these substances as God has deposited them in the bosom of the earth for purposes of human industry; but all our knowledge and all our skill could not produce them. “In the bosom of mountains where copper is found, they hollow out galleries which go down deep into the earth. There workmen called miners, with lamps to light them, attack the rock with great blows of the pick, while others carry the detached blocks outside. These blocks of stone in which copper is found are called ore. In furnaces made for the purpose they heat the ore to a very high temperature. The heat of our stove, when it is red-hot, is nothing in comparison. The copper melts, runs, and is separated from the rest. Then, with hammers of enormous weight, set in motion by a wheel turned by water, they strike the mass of copper which, little by little, becomes thin and is hollowed into a large basin. “The coppersmith continues the work. He takes the shapeless basin and, with little strokes of the hammer, fashions it on the anvil to give it a regular shape.” “That is why coppersmiths tap all day with their hammers,” commented Jules. “I had often wondered, when passing their shops, why they made so much noise, always tapping, without any stop. They were thinning the copper; shaping it into saucepans and kettles.” Here Emile asked: “When a kettle is old, has holes in it and can’t be used, what do they do with it? I heard Mother Ambroisine speak of selling a worn-out kettle.” “It is melted, and another new kettle made out of the copper,” replied Uncle Paul. “Then the copper does not wear away?” “It wears away too much, my friend: some of it is lost when they rub it with sand to make it shine; some is lost, too, by the continual action of the fire; but what is left is still good.” “Mother Ambroisine also spoke of recasting a lamp which had lost a foot. What are lamps made of?” “They are of tin, another substance that we find ready-made in the bosom of the earth, without the power of producing it ourselves.” About HackerNoon Book Series: We bring you the most important technical, scientific, and insightful public domain books. This book is part of the public domain. Jean-Henri Fabre (2018). The Story-book of Science. Urbana, Illinois: Project Gutenberg. Retrieved https://www.gutenberg.org/cache/epub/56795/pg56795-images.html This eBook is for the use of anyone anywhere at no cost and with almost no restrictions whatsoever. You may copy it, give it away or re-use it under the terms of the Project Gutenberg License included with this eBook or online at www.gutenberg.org , located at https://www.gutenberg.org/policy/license.html.

THE KETTLE

Nowadays the engineering community has many products for authentication in their frameworks. Lots of them have built-in features for authentication and a lot of libraries available for social sign-in. We have the Django framework, Flask, and python-social-auth to build almost everything we need to authenticate users in the pythonic world. In this article, I'll show you an example of how to add everything we need for the user's authentication without writing lots of lines of code. The code used in this blog post is available on . We'll use , , , , , and . GitHub Flask flask cookie-cutter docker docker-compose Postgres Ory Kratos Ory Keto Let's take a look at the login flow of our application using Ory Kratos and Ory Keto. What we will use in our project is a great tool to bootstrap our project structure. It's always a great idea to have ready-to-use linters, Dockerfile, and package management tools out of the box. Flask cookiecutter Postgres as an RDBMS. We will have two Postgres services running in two containers in this example. I think that it's a great idea to keep it simple without using custom scripts to have multiple databases available in a single docker-compose service. Ory Kratos with UI to authenticate users. Ory Keto as an access control service. Setting up Ory Kratos Ory Kratos will be responsible for storing identity data such as email/login and password. Using the guide we need to copy the contents of to the root of your project and then add the following content to the docker-compose: quickstart contrib/quickstart/kratos/email-password postgres-kratos:
    image: postgres:9.6
    ports:
      - "5432:5432"
    environment:
      - POSTGRES_USER=kratos
      - POSTGRES_PASSWORD=secret
      - POSTGRES_DB=kratos
    networks:
      - intranet

  kratos-migrate:
    image: oryd/kratos:v0.8.0-alpha.3
    links:
      - postgres-kratos:postgres-kratos
    environment:
      - DSN=postgres://kratos:secret@postgres-kratos:5432/kratos?sslmode=disable&max_conns=20&max_idle_conns=4
    networks:
      - intranet
    volumes:
      - type: bind
        source: ./kratos
        target: /etc/config/kratos
    command: -c /etc/config/kratos/kratos.yml migrate sql -e --yes

  kratos:
    image: oryd/kratos:v0.8.0-alpha.3
    links:
      - postgres-kratos:postgres-kratos
    environment:
      - DSN=postgres://kratos:secret@postgres-kratos:5432/kratos?sslmode=disable&max_conns=20&max_idle_conns=4
    ports:
      - '4433:4433'
      - '4434:4434'
    volumes:
      - type: bind
        source: ./kratos
        target: /etc/config/kratos
    networks:
      - intranet
    command: serve -c /etc/config/kratos/kratos.yml --dev --watch-courier

  kratos-selfservice-ui-node:
    image: oryd/kratos-selfservice-ui-node:v0.8.0-alpha.3
    environment:
      - KRATOS_PUBLIC_URL=http://kratos:4433/
      - KRATOS_BROWSER_URL=http://127.0.0.1:4433/
    networks:
      - intranet
    ports:
      - "4455:3000"
    restart: on-failure


  mailslurper:
    image: oryd/mailslurper:latest-smtps
    ports:
      - '4436:4436'
      - '4437:4437'
    networks:
      - intranet

  postgres-keto: Setting up Ory Keto You can get familiar with the concepts of Ory Keto reading the guide. These articles can give you a brief introduction to it. Since we need to manage access to the home page, we need to create a folder at the root of our project and have a file with the following content: quickstart keto keto/keto.yml version: v0.7.0-alpha.1

log:
  level: debug

namespaces:
  - name: app
    id: 1

serve:
  read:
    host: 0.0.0.0
    port: 4466
  write:
    host: 0.0.0.0
    port: 4467 We need the following containers: postgresd-auth is the database for Ory Keto. keto-migrate that takes care of database migrations. keto-perms is a wrapper to work with permissions using a command-line interface. keto runs the server. version: "3.7"

x-default-volumes: &default_volumes
  volumes:
    - ./:/app
    - node-modules:/app/node_modules
    - ./dev.db:/tmp/dev.db

services:
  oathkeeper:
    image: oryd/oathkeeper:v0.38
    depends_on:
      - kratos
    ports:
      - 8080:4455
      - 4456:4456
    command:
      serve proxy -c "/etc/config/oathkeeper/oathkeeper.yml"
    environment:
      - LOG_LEVEL=debug
    restart: on-failure
    networks:
      - intranet
    volumes:
      - ./oathkeeper:/etc/config/oathkeeper

  flask:
    build:
      context: .
    image: "kratos_app_example-development"
    environment:
      - FLASK_APP=autoapp.py
      - FLASK_ENV=development
    networks:
      - intranet
    restart: on-failure
    volumes:
      - type: bind
        source: ./
        target: /app

  postgres-kratos:
    image: postgres:9.6
    ports:
      - "5432:5432"
    environment:
      - POSTGRES_USER=kratos
      - POSTGRES_PASSWORD=secret
      - POSTGRES_DB=kratos
    networks:
      - intranet

  kratos-migrate:
    image: oryd/kratos:v0.8.0-alpha.3
    links:
      - postgres-kratos:postgres-kratos
    environment:
      - DSN=postgres://kratos:secret@postgres-kratos:5432/kratos?sslmode=disable&max_conns=20&max_idle_conns=4
    networks:
      - intranet
    volumes:
      - type: bind
        source: ./kratos
        target: /etc/config/kratos
    command: -c /etc/config/kratos/kratos.yml migrate sql -e --yes

  kratos:
    image: oryd/kratos:v0.8.0-alpha.3
    links:
      - postgres-kratos:postgres-kratos
    environment:
      - DSN=postgres://kratos:secret@postgres-kratos:5432/kratos?sslmode=disable&max_conns=20&max_idle_conns=4
    ports:
      - '4433:4433'
      - '4434:4434'
    volumes:
      - type: bind
        source: ./kratos
        target: /etc/config/kratos
    networks:
      - intranet
    command: serve -c /etc/config/kratos/kratos.yml --dev --watch-courier

  kratos-selfservice-ui-node:
    image: oryd/kratos-selfservice-ui-node:v0.8.0-alpha.3
    environment:
      - KRATOS_PUBLIC_URL=http://kratos:4433/
      - KRATOS_BROWSER_URL=http://127.0.0.1:4433/
    networks:
      - intranet
    ports:
      - "4455:3000"
    restart: on-failure


  mailslurper:
    image: oryd/mailslurper:latest-smtps
    ports:
      - '4436:4436'
      - '4437:4437'
    networks:
      - intranet

  postgres-keto:
    image: postgres:9.6
    ports:
      - "15432:5432"
    environment:
      - POSTGRES_USER=keto
      - POSTGRES_PASSWORD=secret
      - POSTGRES_DB=keto
    networks:
      - intranet

  keto-migrate:
    image: oryd/keto:v0.7.0-alpha.1
    volumes:
      - type: bind
        source: ./keto
        target: /home/ory
    environment:
      - LOG_LEVEL=debug
      - DSN=postgres://keto:secret@postgres-keto:5432/keto?sslmode=disable&max_conns=20&max_idle_conns=4
    command: ["migrate", "up", "-y"]
    restart: on-failure
    depends_on:
      - postgres-kratos
    networks:
      - intranet

  keto-perms:
    image: oryd/keto:v0.7.0-alpha.1
    volumes:
      - type: bind
        source: ./keto
        target: /home/ory
    environment:
      - KETO_WRITE_REMOTE=keto:4467
      - KETO_READ_REMOTE=keto:4466
      - LOG_LEVEL=debug
      - DSN=postgres://keto:secret@postgres-keto:5432/keto?sslmode=disable&max_conns=20&max_idle_conns=4
    depends_on:
      - postgres-kratos
    networks:
      - intranet

  keto:
    image: oryd/keto:v0.7.0-alpha.1
    volumes:
      - type: bind
        source: ./keto
        target: /home/ory
    ports:
      - '4466:4466'
      - '4467:4467'
    depends_on:
      - keto-migrate
    environment:
      - DSN=postgres://keto:secret@postgres-keto:5432/keto?sslmode=disable&max_conns=20&max_idle_conns=4
    networks:
      - intranet
    command: serve


volumes:
  node-modules:
  kratos-sqlite:

networks:
  intranet: Working with policies Keto has configured namespace to use in Flask application. Following the guide I decided to implement simple permission policy for the demo project: app Check whether a User has Access to Something Use keto-cli managing permissions. Use email for without symbol. subjects @ Pros Easy to use and maintain. Can easily be automated using CI/CD pipelines. Cons Lack of UI can be dealbreaker for non-engineering staff This permission policy can violate , or any other compliances due to personal data usage. GDPR HIPAA Flask part HTTP_STATUS_FORBIDDEN = 403


@blueprint.route("/", methods=["GET", "POST"])
def home():
    """Home page."""

    if 'ory_kratos_session' not in request.cookies:
        return redirect(settings.KRATOS_UI_URL)

    response = requests.get(
        f"{settings.KRATOS_EXTERNAL_API_URL}/sessions/whoami",
        cookies=request.cookies
    )
    active = response.json().get('active')
    if not active:
        abort(HTTP_STATUS_FORBIDDEN)

    email = response.json().get('identity', {}).get('traits', {}).get('email').replace('@', '')

    # Check permissions

    response = requests.get(
        f"{settings.KETO_API_READ_URL}/check",
        params={
            "namespace": "app",
            "object": "homepage",
            "relation": "read",
            "subject_id": email,
        }
    )
    if not response.json().get("allowed"):
        abort(HTTP_STATUS_FORBIDDEN)

    return render_template("public/home.html")

@blueprint.route("/oathkeeper", methods=["GET", "POST"])
def oathkeeper():
    """ An example route to demo oathkeeper integration with Kratos """
    return {"message": "greetings"} Nota bene Consider having and packages that use and . Instead of just calling some magic endpoints, your code will be more readable with SDKs. authorization authentication Kratos SDK Keto SDK Please pay attention to configure and . login session cookies It's better to use instead of having Ory Kratos managed by your team just because Ory manages it, and you don't need to enable observability/logging/metrics for your service. Ory Cloud

How to Secure Your Flask Application with Ory Kratos and Ory Keto

Table of Contents 1. Introduction to Intermittent Fasting and Keto 2. What Intermittent Fasting and Keto Is 3. Advantages of Intermittent Fasting and Keto 4. When Intermittent Fasting and Keto is not right for You 5. All You Need to Know About Intermittent Fasting and Keto Introduction to Intermittent Fasting and Keto Intermittent fasting and keto are very much talked about in the fitness industry, and for a good reason: the combination of fasting and diet has helped some people lose a significant amount of weight. While some people swear by intermittent fasting, others say it is completely unhealthy. Working in tech has its effects on the body: long hours spent at a computer screen, a sedentary lifestyle, and a junk food-based diet can spiral into weight gain. If you are a techie, intermittent fasting and keto may be what you need to improve your health and maintain a good weight. On the other hand, some health conditions can be worsened by intermittent fasting and a keto diet. In this article, with my experience as a medical student, I will be telling you more about intermittent fasting and keto and what the combination can do for you. What is Intermittent Fasting and Keto? is an eating pattern that cycles between periods of fasting and eating. It is not a diet per se; it is a method of eating within a particular time period and fasting for the rest of the day. During the fast period, you may drink water, coffee or a broth to stay energized. Intermittent fasting For example, in the 16:8 method, you fast for 16 hours and set your eating time to that 8 hour period, say 1 pm-9 pm, after skipping breakfast. Other popular methods include the Eat-stop-eat method where fasting is done for 24 hours once or twice a week and the 5:2 method where a 500-600 calorie diet is given for two days a week. , on the other hand, is a diet plan that involves eating a large proportion of healthy fats, an adequate amount of protein, and very few carbohydrates. Keto The keto diet aims to put the body in a state of ketosis: where fat is broken down for energy. There are many variations of the keto diet and they include the Standard Ketogenic Diet and the Cyclical Ketogenic diet. A good number of techies are embracing the keto trend, and is a good example: the tech investor, podcaster, and author of "The Four Hour Work Week" says that the keto diet helped him with his joint pain. Tim Ferris involves combining the fasting program with the high fat keto diet: the combination helps the body burn fat faster by reaching ketosis earlier. The intermittent fasting and keto program Advantages of Intermittent Fasting and Keto : The intermittent fasting and keto combination is great for people who want to lose weight. is faster than just intermittent fasting, but keep in mind that healthier meal options will keep the weight off while fasting. 1. Weight Loss Weight loss : Intermittent Fasting and Keto are both linked with a reduction in cholesterol levels and blood pressure: they will help your stay healthy and strong. 2. Reduced Blood Pressure heart : Intermittent Fasting and Keto help reduce blood by reducing carb intake and lowering insulin levels. This results in more stable blood glucose and proper usage of fat for energy production. 3. Reduced Blood Sugar and Insulin Levels sugar levels : Fasting is reported to delay the aging process and reduce the risk of having cancers. It is that a keto diet will slow the growth of tumors and improve treatment in people with cancer. 4. Reduced Risk of Cancers speculated : Studies have shown that the state of ketosis significantly improves . Intermittent fasting and keto encourage resistance of the nervous system to injury and disease. 5. Better Brain Health brain health When Intermittent Fasting and Keto is not Right for You In some cases, intermittent fasting and keto may produce like fatigue, headaches, bad breath, low blood sugar, loss of muscle mass, and weight gain in the long run. This is why some experts believe that fasting and keto diets are best used for short-term weight loss plans: for example, in obese individuals before a surgery. negatives In addition, there are some conditions in which intermittent fasting and keto may be detrimental to your health. If you are diabetic, have chronic kidney disease, or are undergoing cancer therapy, you shouldn't fast or diet. Women who are pregnant or trying to conceive, children and teenagers, as well as people who have had eating disorders, should avoid fasting or dieting. However, you can discuss with a healthcare professional on a more controlled method of weight loss. All You Need to Know About Intermittent Fasting and Keto Weight loss is important but staying fit and healthy is the endpoint goal. Intermittent fasting and keto may help you lose weight, but sticking to the plan for a long period may be difficult. have shown that many people who lose weight gain it all back: it is better to reduce your calorie intake and work out, and then lose those pounds consistently. Studies If you want to fast and maintain the keto diet, avoid processed foods and plenty of carbs: they might make you gain weight despite fasting for hours. Stick to healthier fats like avocado, olive oil, and whole eggs. Stay hydrated by drinking enough water, teas, and broth while fasting and be consistent. Intermittent fasting and keto is a good short-term weight loss plan: but if you are in the category of the people with conditions that prevent you from fasting and dieting, do not beat yourself up. Many other methods will get you in shape and with your dream body. Just make a plan with your health care professional, stick to it and believe.

Intermittent Fasting and Keto: All You Need To Know

We all know a family member, friend, or co-worker who has at one point, talked about their transition to a structured form of dieting. If you are one of those individuals who have tried or are currently on a dietary program, how are you accurately keeping track of exactly what you are consuming (or not consuming)? For , an author, health advocate, and entrepreneur, she understands the very meaning of living a “hard diet.” She was diagnosed with Hashimoto’s disease, an autoimmune disease that has interfered with her thyroid. In turn, this made it difficult for her to lose weight, while suffering from low energy levels. Martina Slajerova Hopping from one traditional diet to the next, her condition began to impact her career and personal relationships. That’s when Slajerova turned to the Keto diet. One of the most popular, but among the most controversial diets, Keto requires the individual to skip grains, sugar, most fruits and some vegetables. Despite the debate in terms of long-term benefits of Keto, Slajerova began to find her energy levels begin to rise and even started to lose weight. However, there was something still missing for her. Staying On Track in Today’s Digital Age For Slajerova, the goal is simple: help people stay on track. Given how much of a difference the keto diet has made for her life, helping other people could make a huge difference. And while there are a ton of keto experts out there, few have the personal drive that Slajerova does. Then again, few have enjoyed such life-changing results since switching to keto. Slajerova, the author of multiple books, including The KetoDiet Cookbook and Keto Slow Cooker & One-Pot Meals, has taken her knowledge, experience, and passion to the next level—digitally. Her newest work is her —the . And according to , people are loving it. mobile application KetoDiet Women's Health Magazine To be wholly transparent, I have no vested interest in this story or company. Coming from someone who personally has a difficulty with difestion and dietary intake, as I suffer from diverticulitis, I found Slajerova's mission to be quite impactful. The KetoDiet app makes it easy for users to find information specific to their diet. Want to look up meal ideas or need to check out some ingredients? You can use the KetoDiet app and get all the information you need. You can even scan barcodes to look up nutritional information. Men's Health of its top thirteen best keto Instagram accounts, along with . ranked Slajerova’s app as number 9 Healthline rating the app as one of the best amongst paleo apps in 2019 For most dieters, committing themselves to an entirely new regiment is more than a transition—it’s a lifestyle change. For Keto-dieters, they need easier access to information. There’s no question as to the overwhelming complexity one feels scouring through voluminous books, blogs, and videos. But even after reading through dozens of these sources, there are always more questions left unanswered. So, why not create a community, where fellow Keto dieters can share their information and experiences with one another. For Slajerova, she admitted that countless times, she had found herself standing in the produce section, wondering how she could sub out an ingredient in a recipe, or if she could eat a particular vegetable. Often, she found the answer by using her smartphone to dig up her answer(s) on the internet. But, then the —what if you could skip the books, blogs, and videos entirely? Instead of digging through blogs or reading countless books, Keto Diet App users can start building their diet plan . And when questions do arise, say in the grocery store aisle, customers can simply scan the barcode to pull up data. answer presented itself right in the app The vision behind the Keto Diet App will hopefully set precedent for other mobile apps designed and tailored for other dietary regimes.

[Interview] ‘Keto’ Is Going Digital With This New App; It’s Time For You To As Well

Most people in the tech don’t really fit the “meathead” stereotype. industry These are people who spend a lot of time at their desks, working day and night in some cases. But recently, the Keto diet has been booming in the space. And it’s bringing a new definition to the term “meathead.” blockchain If you aren’t familiar with the diet, it’s based around vastly increasing your intake of good fats, lowering your intake of protein, and minimizing carb consumption as much as possible. It’s counterintuitive if you grew up using the food pyramid as a guide. But the diet has been around since the 1920’s, originally as a . natural treatment for epilepsy For true believers, putting your body in a ketosis state — burning fat for fuel instead of sugars — and mental clarity. improves thinking Recently, these benefits of the diet are a major topic of discussion in my Twitter feed and the Reddit communities in the blockchain industry. All the hype intrigued me, so I decided to try the diet out. I’m currently a month in, and I wanted to share some of my thoughts on why it’s become such a trend in tech: Both the blockchain and the Keto diet go against the norm. There’s a bit of a counterculture feel to the Keto diet, much like that of the blockchain space. Today, most people shy away from diets high in fat. That type of program is still considered unhealthy and unsustainable. But the Keto diet isn’t necessarily about eating tons of red meat and brie. There are plenty of ways to get good fats in your diet — olive oil, coconut oil, and avocados to name a few. The urge to go against the grain — quite literally in this case — is one of the fundamental traits in the blockchain and crypto space. We’ve created an entirely new financial system that isn’t controlled by one country or group. It’s for society, so it’s not surprising that the same people who work and create in that space are drawn to other “underground” or counterculture ways of living. a major paradigm shift So far, it’s really interesting to see these two outsider concepts collide. This isn’t the first time we’ve seen tech culture coalesce around nutrition. The craze over the Keto diet is similar to another dietary trend that caught on a few years ago. Back in 2013, Soylent became a huge fad in the tech world. Soylent is essentially just packets of powder that have all of the necessary nutrients you need to survive. People mix them with a number of different liquids and flavorings, but the point is that you don’t have to spend time making and eating a meal. It certainly seems ideal for people in tech who spend what most people consider an inordinate amount of time working. This group usually doesn’t want to spend time thinking about their food — they just want to keep fueling their bodies and minds. The rise of the Keto diet is similar, even though it does take more energy to decide what to eat and make sure you’re sticking with the diet. Adherents claim it makes them sharper and places them in a state of peak mental clarity, a necessary edge in a competitive industry. That seems to be the major motivation for most people on the diet. They’re building , and there’s a sense that this way of eating gives them an intellectual superiority. complex and ambitious systems A new community has formed around the diet and the lifestyle. It’s fascinating how communities form online, even though their users are spread across the globe. When Soylent came out, there was a really vibrant Reddit community of people “hacking” it with Tabasco, cinnamon, and other kinds of flavorings. An ecosystem formed around the product, and because all of the ingredients were posted online, it had a very open-source feel to it. People were providing feedback, trying new mixtures, and letting everyone know how it went. You can see a similar type of community building around the Keto diet. People are sharing recipes and exploring what works best. They’re creating a sense of belonging that’s based in part on whether or not you’re on the Keto diet. It does feel like a small paradigm shift, a movement to reinvent systems that were put in place ages ago and are no longer relevant to us — or were never relevant in the first place. To be clear, I’m not actually advocating for anyone to randomly go out and try this diet. You should always consult your doctor before making any radical changes in the way you eat. I will say that even though I don’t necessarily feel superhuman, I do feel more satisfied, full, and focused in a way that hasn’t happened in the past. I absolutely have more mental clarity and an ability to sit still for longer periods of time. I can’t say how long I’ll stick with it, but I love the idea of exploring these kinds of cultural myths and movements growing in the tech world and the blockchain industry. Plus, the superhuman mental abilities would be a welcome side effect. This article originally appeared on Forbes.com. Thanks for reading! Follow me on Twitter and Quora for more insights on blockchain technology. Or get in touch with the Chronicled team here .

Blockchain And The Keto Diet: How Tech Is Changing The Definition Of “Meathead”

How Mixtral 8x7B Sets New Standards in Open-Source AI with Innovative Design

Authors:
(1) Divyanshu Kumar, Enkrypt AI;
(2) Anurakt Kumar, Enkrypt AI;
(3) Sahil Agarwa, Enkrypt AI;
(4) Prashanth Harshangi, Enkrypt AI. Table of Links Abstract and 1 Introduction 2 Problem Formulation and Experiments 3 Experiment Set-up & Results 4 Conclusion and References A. Appendix 4 CONCLUSION Our work investigates the LLM’s safety against Jailbreak attempts. We have demonstrated how fine-tuned and quantized models are vulnerable to jailbreak attempts and stress the importance of using external guardrails to reduce this risk. Fine-tuning or quantizing model weights alters the risk profile of LLMs, potentially undermining the safety alignment established through RLHF. This could result from catastrophic forgetting, where LLMs lose memory of safety protocols, or the finetuning process shifting the model’s focus to new topics at the expense of existing safety measures. The lack of safety measures in these fine-tuned and quantized models is concerning, highlighting the need to incorporate safety protocols during the fine-tuning process. We propose using these tests as part of a CI/CD stress test before deploying the model. The effectiveness of guardrails in preventing jailbreaking highlights the importance of integrating them with safety practices in AI development. This approach not only enhances AI models but also establishes a new standard for responsible AI development. By ensuring that AI advancements prioritize innovation and safety, we promote ethical AI deployment, safeguarding against potential misuse and fostering a secure digital future. REFERENCES Zifan Wang Andy Zou. AdvBench Dataset, July 2023. URL https://github.com/ llm-attacks/llm-attacks/tree/main/data. Patrick Chao, Alexander Robey, Edgar Dobriban, Hamed Hassani, George J. Pappas, and Eric Wong. Jailbreaking Black Box Large Language Models in Twenty Queries. arXiv, October 2023. doi: 10.48550/arXiv.2310.08419. Edoardo Debenedetti, Giorgio Severi, Nicholas Carlini, Christopher A. Choquette-Choo, Matthew Jagielski, Milad Nasr, Eric Wallace, and Florian Tramer. Privacy Side Channels in Machine ` Learning Systems. arXiv, September 2023. doi: 10.48550/arXiv.2309.05610. Tim Dettmers, Artidoro Pagnoni, Ari Holtzman, and Luke Zettlemoyer. QLoRA: Efficient Finetuning of Quantized LLMs. arXiv, May 2023. doi: 10.48550/arXiv.2305.14314. Micah Gorsline, James Smith, and Cory Merkel. On the Adversarial Robustness of Quantized Neural Networks. arXiv, May 2021. doi: 10.1145/3453688.3461755. Kai Greshake, Sahar Abdelnabi, Shailesh Mishra, Christoph Endres, Thorsten Holz, and Mario Fritz. Not what you’ve signed up for: Compromising Real-World LLM-Integrated Applications with Indirect Prompt Injection. arXiv, February 2023. doi: 10.48550/arXiv.2302.12173. Bing He, Mustaque Ahamad, and Srijan Kumar. PETGEN: Personalized Text Generation Attack on Deep Sequence Embedding-based Classification Models. In KDD ’21: Proceedings of the 27th ACM SIGKDD Conference on Knowledge Discovery & Data Mining, pp. 575–584. Association for Computing Machinery, New York, NY, USA, August 2021. ISBN 978-1-45038332-5. doi: 10.1145/3447548.3467390. Edward J. Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Lu Wang, and Weizhu Chen. LoRA: Low-Rank Adaptation of Large Language Models. arXiv, June 2021. doi: 10.48550/arXiv.2106.09685. Shuhei Kashiwamura, Ayaka Sakata, and Masaaki Imaizumi. Effect of Weight Quantization on Learning Models by Typical Case Analysis. arXiv, January 2024. doi: 10.48550/arXiv.2401. 17269. Siwon Kim, Sangdoo Yun, Hwaran Lee, Martin Gubri, Sungroh Yoon, and Seong Joon Oh. ProPILE: Probing Privacy Leakage in Large Language Models. arXiv, July 2023. doi: 10.48550/arXiv. 2307.01881. Aounon Kumar, Chirag Agarwal, Suraj Srinivas, Aaron Jiaxun Li, Soheil Feizi, and Himabindu Lakkaraju. Certifying LLM Safety against Adversarial Prompting. arXiv, September 2023. doi: 10.48550/arXiv.2309.02705. Thai Le, Suhang Wang, and Dongwon Lee. MALCOM: Generating Malicious Comments to Attack Neural Fake News Detection Models. IEEE Computer Society, November 2020. ISBN 978-1- 7281-8316-9. doi: 10.1109/ICDM50108.2020.00037. Yi Liu, Gelei Deng, Zhengzi Xu, Yuekang Li, Yaowen Zheng, Ying Zhang, Lida Zhao, Tianwei Zhang, and Yang Liu. Jailbreaking ChatGPT via Prompt Engineering: An Empirical Study. arXiv, May 2023. doi: 10.48550/arXiv.2305.13860. Anay Mehrotra, Manolis Zampetakis, Paul Kassianik, Blaine Nelson, Hyrum Anderson, Yaron Singer, and Amin Karbasi. Tree of Attacks: Jailbreaking Black-Box LLMs Automatically. arXiv, December 2023. doi: 10.48550/arXiv.2312.02119. Long Ouyang, Jeff Wu, Xu Jiang, Diogo Almeida, Carroll L. Wainwright, Pamela Mishkin, Chong Zhang, Sandhini Agarwal, Katarina Slama, Alex Ray, John Schulman, Jacob Hilton, Fraser Kelton, Luke Miller, Maddie Simens, Amanda Askell, Peter Welinder, Paul Christiano, Jan Leike, and Ryan Lowe. Training language models to follow instructions with human feedback. arXiv, March 2022. doi: 10.48550/arXiv.2203.02155. Xiangyu Qi, Yi Zeng, Tinghao Xie, Pin-Yu Chen, Ruoxi Jia, Prateek Mittal, and Peter Henderson. Fine-tuning Aligned Language Models Compromises Safety, Even When Users Do Not Intend To! arXiv, October 2023. doi: 10.48550/arXiv.2310.03693. Traian Rebedea, Razvan Dinu, Makesh Narsimhan Sreedhar, Christopher Parisien, and Jonathan Cohen. NeMo Guardrails: A Toolkit for Controllable and Safe LLM Applications with Programmable Rails. ACL Anthology, pp. 431–445, December 2023. doi: 10.18653/v1/2023. emnlp-demo.40. Alexander Wei, Nika Haghtalab, and Jacob Steinhardt. Jailbroken: How Does LLM Safety Training Fail? arXiv, July 2023. doi: 10.48550/arXiv.2307.02483. Martin Weyssow, Xin Zhou, Kisub Kim, David Lo, and Houari Sahraoui. Exploring ParameterEfficient Fine-Tuning Techniques for Code Generation with Large Language Models. arXiv, August 2023. doi: 10.48550/arXiv.2308.10462. Yisong Xiao, Aishan Liu, Tianyuan Zhang, Haotong Qin, Jinyang Guo, and Xianglong Liu. RobustMQ: Benchmarking Robustness of Quantized Models. arXiv, August 2023. doi: 10.48550/ arXiv.2308.02350. Andy Zhou, Bo Li, and Haohan Wang. Robust Prompt Optimization for Defending Language Models Against Jailbreaking Attacks. arXiv, January 2024. doi: 10.48550/arXiv.2401.17263. Sicheng Zhu, Ruiyi Zhang, Bang An, Gang Wu, Joe Barrow, Zichao Wang, Furong Huang, Ani Nenkova, and Tong Sun. AutoDAN: Interpretable Gradient-Based Adversarial Attacks on Large Language Models. arXiv, October 2023. doi: 10.48550/arXiv.2310.15140. Andy Zou, Zifan Wang, Nicholas Carlini, Milad Nasr, J. Zico Kolter, and Matt Fredrikson. Universal and Transferable Adversarial Attacks on Aligned Language Models. arXiv, July 2023. doi: 10.48550/arXiv.2307.15043. This paper is available on arxiv under CC BY-NC-SA 4.0 DEED license. Authors: (1) Divyanshu Kumar, Enkrypt AI; (2) Anurakt Kumar, Enkrypt AI; (3) Sahil Agarwa, Enkrypt AI; (4) Prashanth Harshangi, Enkrypt AI. Authors: Authors: (1) Divyanshu Kumar, Enkrypt AI; (2) Anurakt Kumar, Enkrypt AI; (3) Sahil Agarwa, Enkrypt AI; (4) Prashanth Harshangi, Enkrypt AI. Table of Links Abstract and 1 Introduction Abstract and 1 Introduction 2 Problem Formulation and Experiments 2 Problem Formulation and Experiments 3 Experiment Set-up & Results 3 Experiment Set-up & Results 4 Conclusion and References 4 Conclusion and References A. Appendix A. Appendix 4 CONCLUSION Our work investigates the LLM’s safety against Jailbreak attempts. We have demonstrated how fine-tuned and quantized models are vulnerable to jailbreak attempts and stress the importance of using external guardrails to reduce this risk. Fine-tuning or quantizing model weights alters the risk profile of LLMs, potentially undermining the safety alignment established through RLHF. This could result from catastrophic forgetting, where LLMs lose memory of safety protocols, or the finetuning process shifting the model’s focus to new topics at the expense of existing safety measures. The lack of safety measures in these fine-tuned and quantized models is concerning, highlighting the need to incorporate safety protocols during the fine-tuning process. We propose using these tests as part of a CI/CD stress test before deploying the model. The effectiveness of guardrails in preventing jailbreaking highlights the importance of integrating them with safety practices in AI development. This approach not only enhances AI models but also establishes a new standard for responsible AI development. By ensuring that AI advancements prioritize innovation and safety, we promote ethical AI deployment, safeguarding against potential misuse and fostering a secure digital future. REFERENCES Zifan Wang Andy Zou. AdvBench Dataset, July 2023. URL https://github.com/ llm-attacks/llm-attacks/tree/main/data. Patrick Chao, Alexander Robey, Edgar Dobriban, Hamed Hassani, George J. Pappas, and Eric Wong. Jailbreaking Black Box Large Language Models in Twenty Queries. arXiv, October 2023. doi: 10.48550/arXiv.2310.08419. Edoardo Debenedetti, Giorgio Severi, Nicholas Carlini, Christopher A. Choquette-Choo, Matthew Jagielski, Milad Nasr, Eric Wallace, and Florian Tramer. Privacy Side Channels in Machine ` Learning Systems. arXiv, September 2023. doi: 10.48550/arXiv.2309.05610. Tim Dettmers, Artidoro Pagnoni, Ari Holtzman, and Luke Zettlemoyer. QLoRA: Efficient Finetuning of Quantized LLMs. arXiv, May 2023. doi: 10.48550/arXiv.2305.14314. Micah Gorsline, James Smith, and Cory Merkel. On the Adversarial Robustness of Quantized Neural Networks. arXiv, May 2021. doi: 10.1145/3453688.3461755. Kai Greshake, Sahar Abdelnabi, Shailesh Mishra, Christoph Endres, Thorsten Holz, and Mario Fritz. Not what you’ve signed up for: Compromising Real-World LLM-Integrated Applications with Indirect Prompt Injection. arXiv, February 2023. doi: 10.48550/arXiv.2302.12173. Bing He, Mustaque Ahamad, and Srijan Kumar. PETGEN: Personalized Text Generation Attack on Deep Sequence Embedding-based Classification Models. In KDD ’21: Proceedings of the 27th ACM SIGKDD Conference on Knowledge Discovery & Data Mining, pp. 575–584. Association for Computing Machinery, New York, NY, USA, August 2021. ISBN 978-1-45038332-5. doi: 10.1145/3447548.3467390. Edward J. Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Lu Wang, and Weizhu Chen. LoRA: Low-Rank Adaptation of Large Language Models. arXiv, June 2021. doi: 10.48550/arXiv.2106.09685. Shuhei Kashiwamura, Ayaka Sakata, and Masaaki Imaizumi. Effect of Weight Quantization on Learning Models by Typical Case Analysis. arXiv, January 2024. doi: 10.48550/arXiv.2401. 17269. Siwon Kim, Sangdoo Yun, Hwaran Lee, Martin Gubri, Sungroh Yoon, and Seong Joon Oh. ProPILE: Probing Privacy Leakage in Large Language Models. arXiv, July 2023. doi: 10.48550/arXiv. 2307.01881. Aounon Kumar, Chirag Agarwal, Suraj Srinivas, Aaron Jiaxun Li, Soheil Feizi, and Himabindu Lakkaraju. Certifying LLM Safety against Adversarial Prompting. arXiv, September 2023. doi: 10.48550/arXiv.2309.02705. Thai Le, Suhang Wang, and Dongwon Lee. MALCOM: Generating Malicious Comments to Attack Neural Fake News Detection Models. IEEE Computer Society, November 2020. ISBN 978-1- 7281-8316-9. doi: 10.1109/ICDM50108.2020.00037. Yi Liu, Gelei Deng, Zhengzi Xu, Yuekang Li, Yaowen Zheng, Ying Zhang, Lida Zhao, Tianwei Zhang, and Yang Liu. Jailbreaking ChatGPT via Prompt Engineering: An Empirical Study. arXiv, May 2023. doi: 10.48550/arXiv.2305.13860. Anay Mehrotra, Manolis Zampetakis, Paul Kassianik, Blaine Nelson, Hyrum Anderson, Yaron Singer, and Amin Karbasi. Tree of Attacks: Jailbreaking Black-Box LLMs Automatically. arXiv, December 2023. doi: 10.48550/arXiv.2312.02119. Long Ouyang, Jeff Wu, Xu Jiang, Diogo Almeida, Carroll L. Wainwright, Pamela Mishkin, Chong Zhang, Sandhini Agarwal, Katarina Slama, Alex Ray, John Schulman, Jacob Hilton, Fraser Kelton, Luke Miller, Maddie Simens, Amanda Askell, Peter Welinder, Paul Christiano, Jan Leike, and Ryan Lowe. Training language models to follow instructions with human feedback. arXiv, March 2022. doi: 10.48550/arXiv.2203.02155. Xiangyu Qi, Yi Zeng, Tinghao Xie, Pin-Yu Chen, Ruoxi Jia, Prateek Mittal, and Peter Henderson. Fine-tuning Aligned Language Models Compromises Safety, Even When Users Do Not Intend To! arXiv, October 2023. doi: 10.48550/arXiv.2310.03693. Traian Rebedea, Razvan Dinu, Makesh Narsimhan Sreedhar, Christopher Parisien, and Jonathan Cohen. NeMo Guardrails: A Toolkit for Controllable and Safe LLM Applications with Programmable Rails. ACL Anthology, pp. 431–445, December 2023. doi: 10.18653/v1/2023. emnlp-demo.40. Alexander Wei, Nika Haghtalab, and Jacob Steinhardt. Jailbroken: How Does LLM Safety Training Fail? arXiv, July 2023. doi: 10.48550/arXiv.2307.02483. Martin Weyssow, Xin Zhou, Kisub Kim, David Lo, and Houari Sahraoui. Exploring ParameterEfficient Fine-Tuning Techniques for Code Generation with Large Language Models. arXiv, August 2023. doi: 10.48550/arXiv.2308.10462. Yisong Xiao, Aishan Liu, Tianyuan Zhang, Haotong Qin, Jinyang Guo, and Xianglong Liu. RobustMQ: Benchmarking Robustness of Quantized Models. arXiv, August 2023. doi: 10.48550/ arXiv.2308.02350. Andy Zhou, Bo Li, and Haohan Wang. Robust Prompt Optimization for Defending Language Models Against Jailbreaking Attacks. arXiv, January 2024. doi: 10.48550/arXiv.2401.17263. Sicheng Zhu, Ruiyi Zhang, Bang An, Gang Wu, Joe Barrow, Zichao Wang, Furong Huang, Ani Nenkova, and Tong Sun. AutoDAN: Interpretable Gradient-Based Adversarial Attacks on Large Language Models. arXiv, October 2023. doi: 10.48550/arXiv.2310.15140. Andy Zou, Zifan Wang, Nicholas Carlini, Milad Nasr, J. Zico Kolter, and Matt Fredrikson. Universal and Transferable Adversarial Attacks on Aligned Language Models. arXiv, July 2023. doi: 10.48550/arXiv.2307.15043. This paper is available on arxiv under CC BY-NC-SA 4.0 DEED license. This paper is available on arxiv under CC BY-NC-SA 4.0 DEED license. available on arxiv

Stories, News & Mentions about KETTO

Ketto’s Stories on HackerNoon

Ketto's News & Mentions