How to secure a mobile or web application in Algeria (2027): it holds our customers' data and nobody can say whether it is safe, one digit changed in the address shows someone else's order, what to check, fix, and ask the contractor

“Our application holds the names, phone numbers and orders of our customers. A contractor built it, and nobody here can say whether it is safe. A customer showed us that by changing one digit in the address, he could see someone else's order. How do you secure an application, where do you start, and how do you check what was done?” An application's security is not decided on its screens: it is decided on the server, which must check every request. The rest follows an order: access, passwords, secrets, updates.

Symloop13 min read
How to secure an application what is checked on the serverAccess checked on every request, passwords hashed, secrets out of the code, updatesAPPLICATION SECURITY · ALGERIA · 2027How to secure an applicationwhat is checked on the serverAccess checked on every request, passwords hashed, secrets outof the code, updatesONE DIGIT CHANGEDanother customer’s orderREQUEST REFUSEDchecked on the serversymloop.com

Short answer. An application is safe when the server checks every request: who are you, and is this yours. That is the most widespread flaw: the 2025 edition of the OWASP ranking keeps it in first place. Start with four checks, in this order: access (changing a number in the address must show nothing), passwords (hashed, never readable), secrets (no key in the code on the phone), updates (system, components, application). Then close unused accounts, keep a log, and have the tests replayed by a third party. Six weeks in the example, for an existing application.

This page is for the manager whose application, mobile or web, holds customer data without anyone in the company being able to say how, and for the founder based in France who had a platform built for Algeria and wants to know what they have in their hands. The rules quoted come from public sources: the OWASP ranking of the ten main risks of web applications, 2025 edition, its list for mobile, 2024 edition, its sheet on password storage, and encyclopaedia definitions. The ordering application, its numbers 1042 and 1043, its four accounts and its calendar are a constructed example. This page does not cover legal obligations on personal data: see your adviser. The security of the whole company, computers, network, mail, belongs to another guide.

Changing one digit in the address: the most widespread flaw

The figure shows the same request, received by a fragile application then by a safe one.

Figure 1 · A signed-in customer changes one digit in the address of their order. The fragile application checks only the sign-in, and returns someone else’s order. The safe application also checks that the order is theirs, refuses, and records the attempt. Constructed example.

The name of the flaw. Access control enforces policy such that users cannot act outside of their intended permissions. The variant in the figure, viewing someone else's account or order by providing its identifier, is called insecure direct object reference. The most widespread. OWASP, which publishes a ranking of the ten main risks of web applications, keeps broken access control in first place in its 2025 edition, and writes that 100% of the applications tested were found to have some form of it.

Where the check is made. Access control is only effective in trusted server-side code. Hiding a button on the screen protects nothing: a request can be written by hand, without going through the screen. The building rule fits in three words, deny by default: except for what is public, everything is forbidden until a rule allows it. Who sees what. Those rules still have to be written, by role, by branch, by customer: that is the subject of the guide to roles, rights and the audit log.

«Hiding a button protects nothing: a request can also be written by hand.»
01

A safe application, in five layers

The accented layer is the server.

Figure 2 · The accented layer is the server: that is where each request is accepted or refused. The screens on the phone protect nothing by themselves; they only show what the server agreed to answer.

The phone. OWASP also keeps a list specific to mobile applications: its 2024 edition puts improper credential usage first, and counts insecure data storage and insecure communication among its ten risks. In plain words: no key or password written in the code, because an installed application can be taken apart and read; and on the device, only what the screen needs. The path. TLS is a cryptographic protocol designed to provide secured communications over a network: it is the padlock in the address. It protects the journey, not what arrives at the end: an application can be entirely on HTTPS and return someone else's order.

The server and the database. The principle of least privilege says that each component must be able to access only the information and resources necessary for its purpose: the account the application uses to read the database does not need to be able to erase everything. Signing in. “Who are you?” is settled by a code sent to the phone or by a password, as in the guide to sign-in by code; admin accounts add a second check. Partners. If other systems connect to yours, each key has its rights and its limits, as described in the guide to opening an API.

02

What the thief of the database finds

The figure shows the accounts table, in clear then hashed and salted.

Figure 3 · The accounts table, as a thief would read it. In clear, everything is read, and two customers visibly share a password. Hashed and salted, the four passwords become four unreadable and different strings. Constructed example, strings shortened.

Hashed, unreadable. Hashing is a one-directional mathematical operation, quick to calculate yet hard to reverse. OWASP recommends hashing passwords with modern, slow algorithms such as Argon2id, bcrypt or PBKDF2. So the server does not keep your password: it keeps enough to check that you know it. The salt. A salt is a unique, randomly generated string added to each password as part of the hashing process: two customers who chose “123456” do not have the same fingerprint, and a ready-made list of fingerprints is no longer of any use.

The test. Click on “forgot password”. If your old password is sent back to you, it is kept readable; a well-built application can only make you choose a new one. The other secrets. The keys of payment, text-message or mapping services live on the server, out of the code, and are changed when someone leaves the team. Backups. A copy of the database contains everything the database contains: it is protected just as much, as in the guide to backups.

03

Securing an existing application, in six steps

The second step is accented: test access.

Figure 4 · The accented step is the second: test access, screen by screen, by changing a number. It is the test that finds the most, the fastest. What is no longer used is then closed, before fixing what is used.

Inventory, test, close. What data does the application keep, who has admin access, which keys exist, which components are installed? Then the access test, screen by screen. Then closing: accounts of former employees or former contractors, keys no longer used, admin pages open to all. The 2025 edition of the OWASP ranking puts security misconfiguration in second place, and software supply chain failures in third: the components the application carries and nobody tracks.

Fix. Access first. Then inputs: SQL injection is a code injection technique used to attack data-driven applications; the protection is never to mix what the user types with the commands given to the database. Then sign-in attempts: rate limiting is used to control the rate of requests, so that nobody can try a thousand passwords in one night. Watch. The same ranking puts logging and alerting failures in ninth place: an attempt nobody recorded will be made again. Updates are planned by the year, as described in the guide to keeping an application alive.

«An application that is safe at delivery does not stay safe by itself.»
04

Six questions to put to whoever built the application

For each question, an answer that reassures and an answer that worries.

Figure 5 · Six questions, and for each an answer that reassures and an answer that worries. They are put in writing, to the contractor or to your team. A vague answer is already an answer.

In writing. Put these six questions in writing, to your contractor or your team, and ask to see, not only to hear: the list of admin accounts, the date of the last update, the result of the access test. What must be yours. The source code, the app store accounts, the domain name, access to the server: in the company's name, not a person's. If that is not the case, start there, with the guide to taking an application back.

Having it checked. A penetration test is an authorized simulated cyberattack on a computer system, performed to evaluate its security. It comes after the fixes, not before: it should not have to find what the access test finds in an hour. What it looks at, and what its report must contain, is described in the guide quoted above on the security audit and the penetration test.

05

Six weeks for an existing application, in 2027

Six weeks in the example, the first to inventory and test.

  • What the manager receives: the inventory of data, accounts and keys; the result of the access test, screen by screen; the list of what was closed; the fixes, in order of severity; a new version published; a report a third party can read.
  • What the company no longer does: learn of a flaw from a customer, keep a shared admin account, leave an application two years without an update, believe that a padlock in the address is enough.
  • What is measured: screens tested, access flaws found then fixed, named admin accounts, age of the last update, refused attempts written to the log.
  • What Symloop does: the inventory, the access test, closing unused accounts and keys, fixing access on the server, redoing the storage of passwords and secrets, updates and publication, then monitoring and preparing a test by a third party.
Figure 6 · Six weeks in the example, for an existing application. The accented period is the first: inventory and test, before touching the code. Access fixes take the longest, because they are done request by request. Indicative durations of a constructed example.

The order. Inventory and tests in the first week: at the end, you know what is serious. What is no longer used is then closed, which needs no code. Access fixes take three weeks in the example, because each request to the server is gone over one by one. The new version. On mobile, fixes that touch the application go through a new version on the stores; as long as old versions are in circulation, it is the server that must refuse what they should no longer ask for.

Afterwards. The same tests are replayed, then entrusted to a third party. And security becomes a line of the year: monthly updates, a review of accounts at each departure, an access test for each new screen. A new application is built this way from the first day: access rules are written with the screens, not after.

How do you secure an application?
In this order: check on the server that every request is allowed, keep passwords hashed, take the keys out of the code, update the system and components, close unused accounts, then log and watch. The first step finds the most flaws.
How do you secure a mobile application?
Like a web application, with three more rules: no secret in the application's code, as little data as possible on the phone, and a server that never trusts what the application sends it. An installed application can be taken apart and read.
How do we know whether our application is safe?
Run three tests on your own: change a number in an address, click on “forgot password”, and ask for the date of the last update. If another order shows, if the old password comes back or if the date is two years old, there is work to do.
What is access control?
It is the rule that stops a user acting outside their permissions: seeing someone else's order, changing a price, opening an admin page. It is checked on the server, on every request, and denies by default.
How should passwords be kept?
Hashed, with a slow algorithm such as Argon2id, bcrypt or PBKDF2, and a salt of its own for each account. Nobody, not even the administrator, should be able to read them. If the application can send you your password, it is badly kept.
Can a key be put in the application's code?
No. An application installed on a phone can be taken apart, and everything written in it can be read. The keys of payment, text-message or mapping services stay on the server, and the application goes through it.
Is HTTPS enough to secure an application?
No. It protects the journey between the phone and the server, which is essential. But it says nothing about what the server agrees to answer: an application entirely on HTTPS can return another customer's data.
Do we need a penetration test?
Yes, after the fixes, and by someone other than the person who wrote the application. Before, it spends time finding what a few simple tests already show. Ask for a report ranked by severity, with the way to reproduce each flaw.
How often should an application be updated?
Every month for the system and components, and without waiting when a serious flaw is announced. An application delivered two years ago and never updated carries two years of flaws known to everyone.
How long does it take to secure an existing application?
Six weeks in this page's example: one to inventory and test, one to close, three to fix access, then the data, the updates and the check. The duration depends mostly on the number of screens and requests.
Tell us what the application keeps, who built it and when it was last updated: we start with the inventory and the access test

An application holding your customers' data without anyone knowing how, a flaw reported by a customer, a contractor who has left, or a founder in France who wants to know what the platform they had built is worth: we inventory, we test, we fix in order

Inventory, access test screen by screen, closing unused accounts and keys, access checked on the server, passwords hashed, secrets out of the code, updates, a log, a check by a third party. We work from Algiers and Sétif, in French, Arabic and English.