PHYSICAL REVIEW A 107, 029902(E) (2023)

Erratum: Fundamental limitations on the device-independent quantum conference

key agreement [Phys. Rev. A 105, 022604 (2022)]

Karol Horodecki, Marek Winczewski, and Siddhartha Das

(Received 12 January 2023; published 16 February 2023)

DOI: 10.1103/PhysRevA.107.029902

We recently learned that the bound presented in Sec. IV of our paper is redundant. This fact motivated us, apart from
describing which parts should be disregarded (see below), to also revise some of the text. It is important to note that there
is no change with respect to the results and proofs of the original version of our paper. We describe below essential changes.
We have learned that the two measures of multipartite entanglement–multipartite squashed entanglement Esq and its dual Ẽsq
in our paper are, in fact, equal to each other. The equality follows from Theorem 7 in Ref. [1]. Precisely the two definitions given
in our paper below are equivalent [1].
Definition 4 [from Ref. [23] (of the original paper)]. “For an N-partite state ρA1 ,...,AN ,
 
q
Esq ρA1 ,...,AN := inf I (A1 :A2 : . . . :AN |E )σ , (20)
σ

where the infimum is taken over states σA1 ,...,AN E that are extensions of ρA1 ,...,AN , i.e., Tr E [σA1 ,...,AN E ] = ρA1 ,...,AN .”
and

Definition 7. “For an N-partite state ρA1 ,...,AN ,

sq (ρN (A) ) := inf DN (σN (A)E ),
E (49)
where the infimum is taken over states σN (A)E that are extensions of ρN (A) , i.e., Tr E [σN (A)E ] = ρN (A) .”
Here,

N
I (A1 : . . . : :N |E )ρ = S(Ai |E )ρ − S(A1 , . . . , AN |E )ρ .
i=1

and
DN (ρN (A)E ) := I (A1 :A2 · · · AN |E )ρN (A)E + I (A2 :A3 · · · AN |A1 E )ρN (A)E + I (A3 : A4 · · · AN |A1 A2 E )ρN (A)E
+ · · · + I (AN−1 :AN |A1 · · · AN−2 E )ρN (A)E .
In turn, our upper bounds on the device-independent (DI) key based on the dual measure given in Theorem 5 and
Corollary 3 in Sec. IV of the paper equals the multipartite reduced c-squashed entanglement bounds given in Theorems 2
and Corollary 2 in Sec. III of our paper, respectively. For this reason, any reference to the dual function, including these in the
discussion section, should be skipped in reading. As a result, Fig. 2 also should not contain the plot of a dual bound, which
appeared to be not equal to the c-squashed due to lack of optimization. Figure 2 should look like Fig. 2 here.
We have also noted typographical errors that could make unclear Eq. (68) and text around it. The new text reflects the numerics
that was actually performed in our paper, hence, the modification presented below does not affect the plot given in Fig. 2 there.
It was
“Pνattack (a, b1 , b2 , f |020) = E →F PνCC (a, b1 , b2 , e|020)
= (1 − ν)3 PGHZ (a, b1 , b2 |x, y1 , y2 )δe,? + [1 − (1 − ν)3 ]PνL (a, b1 , b2 |x, y1 , y2 )
× δe,(a,b1 ,b2 ) [δa,b,c δ f ,a + (1 − δa,b,c )δ f ,? ], (68)
where δa,b,c is 1 if all three indices have the same value and 0 otherwise. The above attack strategy is, therefore, a direct
three-partite generalization of strategy proposed in Ref. [17]. The eavesdropper aims to be correlated only with the events
(a, b1 , b2 ) = (0, 0, 0) or (a, b1 , b2 ) = (1, 1, 1), which mimic outputs of the honest strategy of the Greenberger- Horne-Zeilinger
(GHZ) state. By applying the above attack strategy, we are ready to plot an upper bound on the reduced cc-squashed entanglement
Corollary 2.”

2469-9926/2023/107(2)/029902(3) 029902-1 ©2023 American Physical Society

ERRATA PHYSICAL REVIEW A 107, 029902(E) (2023)

KDI
1.0

0.8

0.6

0.4

0.2

Q
0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14

FIG. 2. Plot of upper and lower bounds on the device-independent conference key agreement (DI-CKA) of Ref. [2]. The yellow dashed
line represents an upper bound (not fully optimized) on the upper bound N−11
I[N (A) ↓ E ] from Eq. (39) with the attack strategy in Eq. (68).
The red dashed-dot curve is the trivial upper bound obtained in Corollary 5 via the relative entropy of entanglement bound (1 − ν). The blue
solid line represents the lower bound from Ref. [2].

It should be now as follows:

“Pνattack (a, b1 , b2 , f |020) = E →F PνCC (a, b1 , b2 , e|020)
   
= (1− ν)3 PGHZ (a, b1 , b2 |020)δ f ,? + [1 − (1 − ν)3 ]PνL (a, b1 , b2 |020) δa,b1 ,b2 δ f ,a + 1 − δa,b1 ,b2 δ f ,? ,
(68)
where δa,b1 ,b2 is 1 if all three indices have the same value and 0 otherwise. The above attack strategy is, therefore, a direct three-
partite generalization of strategy proposed in Ref. [3]. The eavesdropper aims to be correlated only with the events (a, b1 , b2 ) =
(0, 0, 0) or (a, b1 , b2 ) = (1, 1, 1), whenever they originate from the local behavior PνL , and maps all other events to f =?. By
applying the above attack strategy, we are ready to plot an upper bound on the reduced c-squashed entanglement shown in
Corollary 2. The latter bound is a multipartite version of the intrinsic information [4,5], used first for the bipartite case in Ref. [6]
against a nonsignaling adversary (see in this context Refs. [7–9]). Here, the strategy of Eve to process her classical variable E to
F is based on Ref. [3] as shown above.”
Proposition 3 originally read as follows:
Proposition 3. “For any N-partite quantum behavior (ρN (A) , M) there is

iid
KDI,dev (ρN (A) , M)  min min KDI,dev
iid
(ρP (N (A)) ) min inf KDD (σP (N (A)) ) , (72)
P P (σP (N (A)) ,L)=(ρP (M(A)) ,M)

where P is any nontrivial partition of the set of systems A1 , . . . , AN .”

It should now be as follows (sign × exhanged for the comma and without a bracket ]):

Proposition 3. “For any N-partite quantum behavior (ρN (A) , M) there is


iid
KDI,dev (ρN (A) , M)  min min KDI,dev
iid
(ρP (N (A)) ), min inf KDD (σP (N (A)) ) , (72)
P P {σP (N (A)) ,L}={ρP (M(A)) ,M}

where P is any nontrivial partition of the set of systems A1 , . . . , AN .”

The errors listed here, and corrected typographical errors, do not affect the results and proofs in our paper. An updated version
has been made available [10].

S.D. acknowledges M. E. Shirokov for pointing out that the dual bound is redundant.

[1] N. Davis, M. E. Shirokov, and M. M. Wilde, Energy- [3] M. Farkas, M. Balanzó-Juandó, K. Łukanowski, J. Kołodyński,
constrained two-way assisted private and quantum capac- and A. Acín, Bell Nonlocality Is Not Sufficient for the Secu-
ities of quantum channels, Phys. Rev. A 97, 062310 rity of Standard Device-Independent Quantum Key Distribution
(2018). Protocols, Phys. Rev. Lett. 127, 050503 (2021).
[2] J. Ribeiro, G. Murta, and S. Wehner, Fully device-independent [4] U. Maurer and S. Wolf, The intrinsic conditional mutual infor-
conference key agreement, Phys. Rev. A 97, 022307 mation and perfect secrecy, in Proc. 1997 IEEE Symposium on
(2018). Information Theory (Abstracts) (1997), p. 88.

029902-2
ERRATA PHYSICAL REVIEW A 107, 029902(E) (2023)

[5] U. Maurer and S. Wolf, Unconditionally secure key agreement [8] E. Kaur, M. M. Wilde, and A. Winter, Fundamental limits on
and the intrinsic conditional information, IEEE Trans. Info. key rates in device-independent quantum key distribution, New
Theor. 45, 499 (1999). J. Phys. 22, 023039 (2020).
[6] A. Acín, S. Massar, and S. Pironio, Efficient quantum key [9] A. Philip, E. Kaur, P. Bierhorst, and M. M. Wilde, Multipartite
distribution secure against no-signalling eavesdroppers, New J. intrinsic non-locality and device-independent conference key
Phys. 8, 126 (2006). agreement, Quantum 7, 898 (2023).
[7] M. Winczewski, T. Das, and K. Horodecki, Limitations on a [10] K. Horodecki, M. Winczewski, and S. Das, Fundamental lim-
device-independent key secure against a nonsignaling adversary itations on the device-independent quantum conference key
via squashed nonlocality, Phys. Rev. A 106, 052612 (2022). agreement, arXiv:2111.02467.

029902-3