EFBW

Disruptive Innovation in Information and Communication Technologies

Wed, 11 Mar 2026 01:03:09 +0100

On February 24^th, 2026, I was recently invited to the French Chamber of Commerce and Industry (CCI) in Dubai to present so;e of the disruptive technological innovations possible in the field of infor;ation and communication security. This included data protection in general and in the specific case of data processing (Machine Learning, Big Data and AI).

The presentation material is available here.

The Gostcrypt Project Goes Private

Sun, 05 Mar 2023 10:57:09 +0100

Since January 2020 and version 2.0 (Linux/Unix/Mac), the Gostcrypt project is no longer public and is restricted to a limited number of users and contributors.

The official project website was closed on March 22^nd, 2023. The Gostcrypt project, since version 3.0 has become a private R&D project for a restricted community. Gostcrypt version 2.0 was a transitional version (rewriting of the GUI in Qt from scratch, addition of new cryptographic standards in mutatinbg mode) and was never made public.

The whole archive up to version 1.3 is available here

We have received a great deal of feedback on this project, much of it constructive and we thank all contributors.

The latest version of Windows 10 does not allow to install Gostcrypt version 1.3 despite a valid certificate, due to changes in Microsoft’s security policy. Microsoft’s policy for obtaining digital certificates to sign Windows binaries, besides being expensive beyond reason, has become too complicated. We have therefore given up developing for Windows.

At last, the conclusion is that the Windows environment is not a trusted enough environment to develop a tool like Gostcrypt and more generally any encryption tool. It is like building a house on ground that you do not own, that you cannot explore and that you can always suspect to be undermined.

An Cryptographic Algorithm Design Framework for Effective Deniable Cryptography

Sat, 22 Feb 2020 10:47:34 +0100

Building effective encryption algorithm (deterministic algorithms) to realize practical deniable cryptography is still an open problem. The only known case is trivial and relies on onetime pads. Since the “key” is as long as the two (or more) plaintexts one may intend to deal with, this solution is not interesting.

I have solved this problem and I have built a design framework to build encryption algorithms enabling effective deniable cryptography. This framework is a C library. It is currently in the industrialization phase and should be presented at a international hacking conference (submission pending).

Let C be a ciphertext of length N, a unique algorithm E and any two different arbitrary plaintexts P₁ and P₂. Is it possible to build E with the following constraints:

E is a deterministic encryption algorithm (stream cipher or block cipher). It is supposed to be public and therefore resistant to known cryptalysis techniques. The key K is k-bit long.
k is far smaller than N (so one-time pad is not considered).
We have C = E(K₁, P₁) = E(K₂, P₂). Moreover P₁ and P₂ have the same size as the ciphertext C.
The scheme can be extended to a finite number of plaintexts P_i

The number of applications is awesome:

Code protection (malware or legitimate program) against static and dynamic analysis. The code first determines whether it is executed in a non-cooperative virtual environment (for dynamic analysis) or not. Depending on the environment, either K₁ (fooling code) or K₂ (real code) is provided by environmental key. This has been used i in blockchain-based malware technology at Forse 2020
Antiforensics techniques
Multiple communication channels in a single one

When considering our operational constraints, the problem of designing E was still an open problem. The first version of the framework deals with E as a stream cipher (second version of the framework aims at addressing the case of block ciphers). Stream ciphers E with k ranging from 128 to 256 and up to three ciphertexts have been produced (less than 2’ of computing time). The cryptographic security analysis of these algorithms have confirmed the resistance against the following attacks:

guess P₁ and P₂ from the ciphertext C (in other words, retrieving keys K₁ and/or K₂)
find P₁ knowing P₂ and conversely
when analysing the encryption algorithm, detecting plausible deniability is not possible.

Google Censorship

Sun, 01 Dec 2019 09:10:51 +0100

Enough is enough. I have decided to stop using Google. I was already living very well without Facebook, Twitter and other GAFAM (I assure you it is possible) but for convenience I was still using Google for Blogger and Google Sites.

For some time now, one of the pages of my site (the one referencing all my hacking conferences, without any dangerous content) was blocked and made inaccessible by Google for months. Deliberate choice or poor quality Google’s AI? I cannot say. In the first case, it is more than concerning and in the second why still use the products of a technology giant who is not able to produce a quality AI.

By doing a quick search, it became clear that GAFAM (Facebook in particular) have moved up a gear and intend to practice an increasingly strong informational eugenics: github accounts closed by Microsoft, Google Play accounts of developers closed by Google, applications deleted from stores by Google or Apple, accounts deleted by Instagram, photos censored by Facebook (in the name of their “morality”)…, without any objective reason or law infringement. The number of cases is increasing day after day.

The purpose of this censorship is to develop social cooling and to shape a world according to the wishes and “morality” of these GAFAMs. Wich is of course seldom ours.

I am neither an extremist (left or right) nor an excited anti-state activist. We need strong states, which means respectful and protective of their citizens. We need to return to a new humanist age, in short a world at the opposite of the one imposed by the GAFAM.

So I am leaving Google. This site contains my web page and some posts from my former blog.

Interesting links (among many others):

OSINT on the TOR Foundation (Update)

Mon, 19 Mar 2018 11:11:00 +0100

At this date (on March 18^th, 2018), we would like to make an update about The TOR Project Inc. As we shown on our paper OSINT Analysis of the TOR Foundation, we worked on the funds and proved that the US government is deeply involved by bringing approximatively 85% of the funds in 2015.

But since this year, we do not have any data to study. Be searching information, we found that the Form 990 which is mandatory for every 501(c)(3) research-education nonprofit organization as The TOR Project Inc is, is not available. This form is required to be released the 15th of the 5th month after the fiscal year with a possible extension of 3 month according to several sources (for instance). Even with this extension, the form 990 of the fiscal year of 2016 of The TOR Project Inc. is supposed to be due on August 15^th 2017. However, we are still not able to find it. It is neither present on the website of The TOR Project Inc. unlike all others form 990 and neither on diverse sources like GuideStar or ProPublica.

Furthermore, by doing some complementary OSINT work, we found that Andrew Lewman was part of the CyAN (Cybersecurity and cybercrime Advisors Network) organization. Just a few reminders: first, Andrew Lewmann was the former Executive Director of The TOR Project Inc. Secondly, we had some reasons to believe that the US government has strong links with The TOR Project Inc. via Roger Dingledine who made an internship in NSA and with some presentations in front of high authorities like the White House and the FBI.

On the CyAN website, we have found the profile of Andrew Lewman and a link to his Linkedin page. On both sources, he claims to have connections with the US government such as several intelligence agencies, NSA or the Department of Homeland Security. We do not know since when he has these connection but doubt is allowed.

To summarize, since the beginning of The TOR Project Inc. in 2006, both of the creator (Roger Dingledine) and the Executive Director (Andrew Lewman) had connections to the US government. Andrew Lewman quit the board in 2015, such as Roger Dingledine in 2016.

List of TOR Relays for Optimal Correlation Attack

Tue, 17 Oct 2017 08:01:00 +0200

Following our study whose preliminary results have been published in a previous post, we have obtained additional results as far as correlation attack and partial anonymity removing are concerned. People who have made feedbacks on our study were interested in the consequences of possible attacks that could be performed on the TOR network.

We will not give data on relay bridges but as far as simple ORs are concerned, here is the list of the most sensitive ORs (preeminent weight in the three-OR circuits) that could be targeted

to correlate aproximatively 40 % of the traffic
to remove anonymity of approximatively 9 % of the traffic.

These ORs should have a boosted/strengthened security.

By the way, we give here the Maltego graph of the authorities with respect to the version evolution. Detailed explanation is contained in a paper recently submitted in an international conference.

Preliminary Results on TOR Routing Protocol Statistical and Combinatorial Analysis

Mon, 04 Sep 2017 16:41:00 +0200

In the present document we share the preliminary results of a 4-month study about the TOR routing protocol from a statistical and combinatorial point of view. For the moment we only give the final data not the algorithms and technical/mathematical approaches (to be fair with conferences where we will present them first).

We have modeled exhaustively all possible routes while taking different parameters into account with the data provided by the TOR foundation only. We have then confronted our theoretical model with the reality on the ground. To do this, we generated thousands of roads on the TOR network and compared the results obtained with those predicted by the theory. A last step of combinatorial analysis has enabled us to identify critical subsets of Onion routers (ORs). We have also managed to extract most of the relay bridges and give an initial list of nearly 2500 relay bridges.

Important notice: we do not claim to have broken TOR in any way. These results are just preliminary results data that will be presented in two parts, each in an international security event. Right after those events, this paper will be extended to provide most of technical details, algorithms, protocols we have set up and used. From those results anyone can figure out the interest, the impact and consequences of them.

The New GOST Standard from the Russian Federation: GOST Grasshopper

Thu, 01 Jan 2015 12:59:00 +0100

The Russian Federation has recently published the project of new standard for block encryption algorithm. This is still a project which has not been formally validated and approved yet.

This algorithm called Gost-Grasshopper intends to supersede (in the future) the current GOST 28147-89 algorithm (64-bit block and 256-bit key, Feistel structure).

I have translated the text from Russian which relates to the Grasshopper algorithm and implemented this new algorithm in C language, using the test vectors sets provided in the reference document. The translation is available here while the source code (under GPLv3) is given here

I have performed a first quick analysis of the new GOST Grasshopper algorithm and here are the first observations I have made (of course this analysis needs to be pushed on further):

Contrary to the GOST 28147-89 algorithm, GOST Grasshopper belongs to the SPN family (Substitution Permutation Network), a block encryption algorithm family which contains the AES (Rijndael).
The main features are: 128-bit blocks (plaintext, ciphertext), a 256-bit master key from which 10 128-bit subkeys are derived according to highly nonlinear process, 10 rounds. The general structure is then very common to SPN and is sketched as follows
<p align='center'>
The statistical analysis of the cipher does not reveal any statistical bias and has passed all the tests (performed with NIST STS SP800-22 revision 1 and L’ecuyer & Simard’s TESTU01 suite). Gost Grasshopper seems to have far better algebraic and combinatorial complexities that the existing SPN and especially than the AES (whose relatively weak algebraic complexity has been pinpointed by several cryptographers [see Harris Nover 2009 for a summary]). Further analyses have to be conducted to confirm this initial result.

Gost Grasshopper seems to be a very promising algorithm both for its speed encryption and for its high cryptogaphic security. Upon confirmation, it should be included soon in the Gostcrypt suite with the same approach and settings (variable S-Box [Substitution S in the reference document], S-Box mutation based on the user’s master key, 512-byte cluster ID used as salt value).

Have a nice day.

E.F.

Focus on my situation with EICAR

Sat, 20 Sep 2014 19:45:00 +0200

A number of authors who have submitted to the EICAR conference have recently contacted because the relevant conference website still mention me as the Scientific Director of EICAR and as the EICAR conference Program Chair. The issue is that they absolutely go no notification regarding their paper. I am sorry for that situation but I am no longer the Scientific Director nor the EICAR Conference Chair since the end of 2013. I have officially announced that in October 2013.

I have held these two positions on a strict voluntary basis. This was a great experience which have provided a lot of intellectual satisfaction. I tried to increase the scientific level of the EICAR conference and had the occasion to meet nice people, authors and attendees. After six years I decided to stop because first I think that no one should occupy the same positions too longer in order to remain efficient and second I was disagreeing with the lack of real and total independence with respect to the Antivirus vendors community. Moreover, being myself in charge of developping the sovereign antimalware products for France (called DAVFI and commercially available under the brand Uhuru), staying with EICAR would have been a contradiction with my own principles.

So I am very sorry for the authors who have been misled by seeing my name as the EICAR Program Chair. I am no longer involved in this conference. For any claims, please contact this address

Have a nice week end

Les incohérences de Microsoft

Tue, 08 Apr 2014 12:31:00 +0200

Bonjour à tous

Microsoft ne soutient plus XP. Donc à part la gendarmerie nationale que cela fait sourire (une fois de plus nos gendarmes ont été clairvoyants), les autres utilisateurs (environ 30 à 40 % du parc) vont souffrir:

soit financièrement : il faut passer à Windows 7 (ou pire à Windows 8) mais aussi changer beaucoup d’applications, en particulier métier, voire changer le matériel (juste passer un scanner Canon Canoscan de XP à Windows 7, par exemple, pour madame Michu c’est impossible). Bref cela va avoir un cout exorbitant ;
soit se dire tant pis je ne fais rien et tant pis pour les attaques (je frémis à l’idée de ce choix que certains vont faire : rappel le parc informatique sous XP est encore important dans bien des secteurs et infrastructures critiques, donc beaucoup connectées sur Internet).

Microsoft annonce qu’en payant très cher (voir par exemple ce lien), il est possible de bénéficier d’une protection prolongée. C’est une gestion en mode Titanic : seuls les passagers de première classe peuvent espérer embarquer dans les canots de sauvetage, les autres peuvent se noyer. Or, ce matin le Directeur technique de Microsoft sur France Inter, Journal de 07:30, index à 10 minutes 50, déclarait : “on n’arrive plus à protéger XP.” (écouter la suite tout aussi sidérante).

La question est alors : pourquoi faire payer pour un service impossible à assurer (dixit le responsable sécurité de Microsoft en France ce matin sur France Inter) ?

Cela appelle donc deux commentaires :

Il est vraiment temps de passer à une alternative crédible dont ne serons plus les otages et les vaches à lait, à savoir le logiciel libre. Si on doit payer, autant que ce soit à des entreprises françaises qui assurerons un soutien plus réactif, moins opaque et à un coût réduit et qui pourront assurer rapidement la migration des applications métier en quelques mois. La Gendarmerie Nationale y est parvenue brillamment, c’est donc possible et ce, sans perte de capacité opérationnelle. Il y a là une occasion en or de créer un véritable choc économique, de compétitivité et des opportunités de création de richesses, et, pour la France de redevenir un exemple (et donc un leader) éclairé en Europe. Mais pour cela, il faut du courage.
Espérons que l’État français ne paiera pas, comme l’Angleterre et les Pays-Bas, pour un service qui n’existe pas.

Bonne journée E.F.