Backdoors, Trust and Scalable Solutions

In this presentation I talk primarily about backdoors by going into to the details of Dual EC DRBG works and the underline math behind it. I then discuss the difficulty of detecting such backdoors, its effects on trust.

Transcript

1. Backdoors, Trust and scalable solutions Ashutosh Ahelleya
2. Encryption
3. Some questions 1. How is the key generated? 2. What are the consequences of seed being predictable? 3. Where does the security of PRNGs lie in? 4. How do we generate “true random” values for the seed? a. How do modern systems compute “true random” values? 5. Why not generate the key using a TRNG? 6. Properties of PRNGs a. How can these features be useful? 7. Assumption that forms the basis of our model for security 8. How is the key sent over the same insecure channel?
4. How is the key generated?
5. How is the key generated? Deterministic function 1. Outputs random looking numbers using an algorithm 2. Takes few “true random” bits and outputs much larger “pseudo random” bits Trivial examples: log(x), sin(x) etc.
6. Pseudo Random Number Generators: PRNGs
7. PRNG trivial example: log(x) Every PRNG requires some initial “true random” number as input → SEED
8. What happens if the seed is predictable?
9. What happens if the seed is predictable? PRNG is public → Attacker also has access Conclusion: seed should not be predictable
10. Security-decisive factors of a PRNG 1. Seed should be truly random and should not be predictable 2. Given a set of outputs of a PRNG, attacker should not be able to guess seed
11. How do we generate “true random” values for the seed?
12. True Random Number Generators A function that generates outputs from physical sources such as temperature, thermal noise etc. that are known to produce non-predictable outputs. Hence, they are called True Random Number Generators.
13. How to modern systems compute “true random” values?
15. Present day: usage of lava lamps in Cloudflare More about it here Youtube video: https://www.youtube.com/watch?v=1cUUfMeOijg
17. Why not generate the key using a TRNG?
18. Why not generate the key using a TRNG? Neither feasible nor scalable
19. Property of RNGs
21. How can this feature be useful? Synchronised RN generation: 2FA
22. Assumption that forms the basis of our model for security “Nature is truly random”
23. How are cryptography standards created? CFP by NIST → Testing → standard = Winner! CSPRNG
24. Dual EC DRBG: abstract CSPRNG published by NIST SP 800-90A in 2006 P, Q → Points on an EC S 0 → Seed S 2 → Updated seed 1. Operation: S i * P is computationally hard to reverse, similar to large integer multiplication 2. Security model: P, Q should be generated independently
25. What if P, Q are not independent?
26. What if P, Q are not independent? Assertion: Only we know the relation between P and Q, no one else does
27. Inference Dual EC PRNG has the potentiality of a backdoor if P, Q are not generated independently
28. What is a backdoor?
30. What is a backdoor? An intentional flaw in an algorithm or an implementation that allows an individual to bypass the security mechanism the system was designed to enforce.
31. Why would someone do it? Organisations can inject a backdoor without others knowing about it, for reasons of importance to their state, such as surveillance etc.
32. The Objection
33. Independency?
35. Allegation → confirmation
39. End of Dual EC DRBG 1. Withdrawn in 2014 2. Stayed a CSPRNG for 7 years!
40. How difficult of a problem it is to detect backdoors? Why?
41. Let’s talk about trust
42. Different meanings of trust 1. Personal and intimate trust 2. Less personal trust
43. What breaks trust?
44. What breaks trust? Overcommit, underdeliver Hype >>> value
46. What happens when a bubble bursts? Theranos: Silicon Valley’s greatest disaster The Big Short
47. Systems species use to incentivize trustworthy behaviour
48. Incentivising trustworthy behaviour 1. Morals 2. Reputation 3. Institutions 4. Security Systems Reference: Liars and Outliers by Bruce Schneier
49. NO! Reference: Reflections on Trusting Trust Can we trust the code we did not create?
50. Types of system we can adopt 1. Fully transparent 2. Fully private 3. Trustless
51. Cambridge-Analytica aftermath
54. Can we have a trustless system?
55. Who do we trust then?
56. “We, as humans, trust the established rules that induce organisations to behave according to a norm, imposing sanctions to those who do not, more than the system itself.” Reference: There’s no good reason to trust blockchain technology Who do we trust then?