Presentation on theme: "I Know your PIN I Know Your PIN Jolyon Clulow Prism"— Presentation transcript:
I Know your PIN I Know Your PIN Jolyon Clulow Prism
Introduction What this talk is not about: The Internet, SSL, VPNs What this talk is about: Bank, Credit and Debit cards Banks, Financial Networks and Switches PINs, PANs, ATMs, TRSMs, POS, Mobile And… …(some of the) ways that one can recover PINs from such supposedly secure systems!
So why are we interested? Justification Driver for modern cryptography – the so called ‘killer app’ of cryptography The concept of a ‘PIN’ is internationally understood and accepted Scale of use Bank, Credit and Debit Cards Card issuing banks Card Associations (Visa, MasterCard,etc) Amount of money protected by these operations Guaranteed that (almost) everyone who reads this, relies on the security thereof to protect their own personal finances.
Talk Outline Introduction Background info: What is PIN security? The attacks Some remedies Real world scenarios The road ahead? Conclusion
Background info Financial Security 101: An Introduction to PINS
Terminology PIN: Personal Identification Number PAN: Personal Account Number ATM: Automatic Teller Machine (cash machine) API: Application Programming Interface (the set of functions exposed/available) API attack: an attack which uses(or abuses) the existing/available functions to compromise the security of the system
What is a TRSM? Tamper Resistant/Responding Security Module (TRSM) Host Security Module (HSM) Hardware Security Module (HSM) Crypto Coprocessor Provides a secure, trusted environment to perform sensitive operations Detects and responds to physical, electronic (or other) attempts to recover key material or sensitive data. Typical measures include: physical tamper envelope/membrane temperature, radiation sensors power supply monitoring and filtering Trigger causes erasure of protected data
Financial Network Architecture
Key Zones Each connected pair of entities share a common key to form a key zone
PIN Encryption e.g. PIN is 1234, Key is ABCDEF 1. Start with an empty PIN block 2. Insert PIN 3. Pad 4. Encrypt the clear PIN block It’s that simple! FFFFFFFFFFFF 2580D0D6B489DD1B
PIN Formats (some examples) VISA Format 3 PIN Block = PPPPFXXXXXXXXXXX IBM 3624 PIN Block = PPPPxxxxxxxxXXXX ISO-1 PIN Block = CLPPPPrrrrrrrrRR where C = X‘1`, L = X‘4` to X’C` r is either P or R VISA Format 2 PIN Block = LPPPPzzDDDDDDDDD
ANSI X9.8 Format (ISO-0) E.g. For a 4 digit PIN P1 = 04PPPPFF FFFFFFFF P2 = 0000AAAA AAAAAAAA Where AAAAAAAAAAAA represents 12 digits of the PAN PB = P1 P2 EPB = e k (PB) Binds the account number to the PIN Diversifies the encrypted PIN block
PIN Translate Translate between different zone keys Question: What if different actors/entities use different formats? Additional operation required PIN Reformat Supports change in PIN formats and PANs
PIN Verification Exist multiple different approaches Simple Offsets PIN Verification Values(PVV) Simple Compare the customer supplied PIN with a reference PIN
PIN Verification (Offsets) 1. Validation data is encrypted under PIN generation (verification) key. 2. Ciphertext is ‘decimalised’ to form IPIN by means of a table. 3. Calculate the offset as OFFSET = PIN-IPIN (where ‘-’ is subtraction modulo 10)
PIN Verification (Offsets) IBM PIN Offset Algorithm Allows user to choose own PIN (also to change it easily) Validation data is typically customer and financial institution specific (e.g. PAN) ‘Decimalization’ by means of a table ABCDEF
The Attacks Attack #1a: ANSI X9.8 Attack Attacks the PIN translate function. Attack #1b: Extended ANSI X9.8 Attack Attacks the PIN translate and reformat functions. Attack #2: The Decimalization Attack Attack against PIN verification algorithm using offsets.
The Attacks Attack #3: Key Separation #1 Attack against PIN verification functions based on failure to enforce key separation between verification and translation(encryption). Attack #4: Key Separation #2 Attack against PIN verification functions based on failure to enforce key separation for different verification algorithms. Attack #5: Check Value Attack Attack against PIN verification algorithm using the check value of a key
Attack #1 ANSI X9.8 (ISO-0) Attack Attacks the PIN translate/reformat function
ANSI X9.8 (ISO-0) Attack Input Parameters Encrypted PIN Block (EPB) PAN Encrypted ‘In’ Key Encrypted ‘Out’ Key Attack Strategy: In an iterative manner, we make a modification to the PAN and observe the effects
ANSI X9.8 (ISO-0) Attack Under normal operation: Inputs ( EPB, P2 ) PB = d k (EPB) P1= PB P2 = 04PPPPFFFFFFFFFF Extract PIN as PPPP Test that PPPP is valid PIN (i.e. each P is a valid decimal digit)
ANSI X9.8 (ISO-0) Attack Instead of supplying the correct PAN (P2) to a call, use a modified PAN (P2’ = P2 ) Inputs ( EPB, P2’ ) PB = d k (EPB) P1’= PB P2’ = (P1 P2) (P2 ) = P1 Say = 0000x P1’= 04PPPPFFFFFFFFFF 0000x
ANSI X9.8 Attack Q: What happens if (P x) is a decimal digit? A: The call passes. Q: What happens if (P x) is not a decimal digit? A: Typically, the call FAILS! We have a test for (P x) < 10.
ANSI X9.8 Attack Building a simple algorithm to identify P 1. Try all possible values of x, yielding a unique * pattern of ‘passes’ and ‘fails’ allowing you to identify P. 2. A decision tree
Attack #2 The Decimalization Attack Attacks the PIN Verification using offsets function
Decimalization Attack Input Parameters Encrypted PIN Block (EPB) Validation Data Decimalization Table Offset Encrypted Key Attack Strategy: In an iterative manner, we make a single change to an entry in the decimalization table and observe the effects
Decimalization Attack PIN = 6598 PIN Ver Key = Val. Data= Ciphertext= E481FC Dec. Table= IPIN= 4481 Offset= 2117
Decimalization Attack Dec. Table (0)= IPIN= 4481 Offset= 2117 (will pass) Dec. Table (1)= IPIN= 4482 Offset= 2117 (will fail) = 2116 (will pass) Thus far we have identified that the 4 th digit in the original IPIN is a 1 and hence that the 4 th PIN digit is 1+7 = 8 (IPIN + Offset).
Decimalization Attack Work factor Initial search for (an unknown) offset requires at most (n-4)10 queries Each change in the dec. table requires at most (n-4) queries At most need to try 15 of the 16 entries in the table for a total of 15(2 4 + n-4) queries. Attack time dependant on TRSM speed Typical values (dependent on speed of TRSM): Known initial offset: 1 – 20 seconds Unknown initial offset: seconds
Properties How efficient are these attacks? Computationally trivial Extremely fast Requires just a few seconds on a Pentium I Typically limited by performance of TRSM What are the requirements? Requires query access to the device, implying either: Physical access to the device/switch/trust center –Special case: Stolen device Access to the network transporting transaction traffic and the ability to inject messages
What about in the ‘Real World’? Real world systems should be following standard industry best practices that if implemented correctly and enforced should limit a potential hacker’s ability to perform such attacks. Physical access control to restricted area. Some thoughts and counter arguments. Attacker can attack at weakest point. One institution’s account holder can be compromised on another institution’s network. Hence must guarantee that all potential networks through which the PIN may travel to be secure. So why did you buy an expensive TRSM in the first place if your defense rests on physical access control? Multi-lane Retail Stores
So what went wrong? Some functions are just badly thought out and insecure. Individually secure functions were added to the API in a manner to make entire system insecure. Insufficient attention was given to the possible interplay between functions. Absence of a single standard to which everyone completely adheres to (many different formats and algorithms exist due to historical reasons). Different customers want different functionality from the same product.
Solutions - Cryptographic Remove ‘weaker’ algorithms/functions (leave only the strongest) Parameter(data) Integrity MAC the PIN block and data PAN, PIN block format, etc MAC any verification/generation data Decimalization table, Validation data, TSP, etc A better PIN Block Format? Key Separation Format (PIN Block Variance) Algorithms Other data (e.g. PAN)
Solutions – Access Control Electronic access control Fine grained, allowing the individual enablement/disablement of Formats Algorithms Functions Limit functionality. Only enable what is required. Disable everything else. Useful to allow a function to be disabled should it later be shown insecure. True split control
Hackers and Threats? Real world scenarios: Risk, Reward and Liability
Disclaimer This material is made available as a courtesy, purely for educational and informative purposes only for an intended audience of responsible individuals with a genuine interest in improving the security of financial networks. Prism makes no claim as to the accuracy or completeness of this information. Prism accepts no responsibility or liability arising from the use of this material.
Insider attack 1. Extract the PIN number for a given account (or accounts) 2. Create a duplicate ‘white card’ (or multiple duplicate cards) 3. Distribute to accomplices to perform a random tour of ATMs
Insider Attack - Reward Let N be the number of compromised accounts, P the average period before unnatural transaction behavior is noticed and L the daily withdrawal limit. Total Fraud Value = NPL Example: N = 5000 P = 2 L = $1000 Total Fraud = $ 10 M
Account Holder Attack 1. Produce a number of duplicate ‘white cards’ of your own card 2. Distribute to multiple accomplices, preferably in different geographical locations to perform a random tour of ATMs. 3. Report the ‘unauthorized’ activity on your account and dispute the transactions.
Account Holder Attack (cont.) It may be advisable to perform a valid transaction “simultaneously” with a fraudulent one since this ‘proves’ you are in possession of your card and preferably in a different location. Best done by multiple card holders from a given institution since: Not an isolated incident Questions the security of the institution Gives the impression of a possible insider attack
Account Holder Attack - Reward Let N be the number of conspiring account holders, P the average period before unnatural transaction behavior is noticed and L the daily withdrawal limit. Total Fraud Value = NPL Average return = PL Example: N = 100 P = 10 L = $1000 Total Fraud = $ 1 M Average return per account holder = $ 10 K
The Repudiation Attack Just deny a transaction Dispute procedure leading to possible litigation Argue the insecurity of the system Best if security of institution already questioned Scenario: Following a successful account holder/insider attack being made public – other account holders (acting individually) may dispute valid transactions that occurred during the attack period (or after) Financial risk is great due to the possible scale (e.g. 0.1 % of an institution’s 1,000,000 customers each disputing a $1000 transaction = $1 M) Loss of confidence in the given institution could well be more damaging
Other Ideas The Competitor Attack Use own network to compromise a competitor institution (could even choose to use administrator privileges to effect this) Reward not the stolen money but the ‘after effects’ Less of a connection between accomplices and institution (no cash trail leading back) The Stock Market Attack ‘Short’ the stock prior to any attack (no cash trail) The Terrorist Attack All/any combinations of all the previous attacks
What now? Q: What should you do now if you are a bank? Contact your vendor, request any best practices information and implement it. Be vigilant. Increase your auditing. Reassure your clients. Wait. Positive pressure on the role players. Q:Is that all? The nature of the problem is such that it is not yours alone (unless you disconnect from the network). The entire network must be secured and until that happens you and your account holders are potentially vulnerable.
The road ahead? Process driven by Card Associations? Due to role and influence over the infrastructure Revise the standards New design/security requirements. Prescriptive requirements limiting what functionality is allowed. Vendors will then update products based on revised standards Expecting (and hoping) for more uniformity and collaboration between different vendor product offerings. (Makes business sense for institutions) Card associations will mandate new requirements to institutions.
The unanswered question? Who is liable in the event of such an attack leading to fraud?
Summary A set of API attacks which allow PIN recovery Design criteria/suggestions to combat the attacks Some potential attack scenarios
The final comment… The most concerning aspect of these attacks, is that you can be attacked on someone else’s network – a network over which you have little or no control.